What is the best portable split ergonomic keyboard?

It depends on whether you want one piece or two. For a true split with separate halves, the Elim Elytra: 420g, wireless, a standard row-staggered layout, rated 4.5 out of 5 after seven weeks of daily use, at $249. For a single-piece Alice layout at a lower price, the Keychron K15 Max from around $75, built on the same platform as the retired Keychron K11 Max.

What is the cheapest split ergonomic keyboard?

The JOMAA Ergonomic Split Foldable Keyboard is available for around USD 22 to 30 on AliExpress. Despite its low price, it offers a comfortable laptop-like typing feel, a full function row, and folds in half for portability.

Can you travel with a split keyboard?

Yes. All the keyboards we review are selected for portability. The lightest option, the ProtoArc XK03, weighs 400g and folds in half to fit in a folder. All our picks fit comfortably in a standard backpack or laptop bag.

Do I need a mechanical keyboard for ergonomics?

No. Both mechanical and scissor switch keyboards can provide excellent ergonomic typing experiences. Keyboard design, the split angle, key spacing, and form factor, matters more than switch type.

What is an Alice layout keyboard?

An Alice layout, also called a unibody split, is a keyboard where the two halves are angled apart but remain in a single connected unit. This promotes a more natural wrist position while keeping the keyboard compact and portable. The Keychron K15 Max is a popular example.

Which split keyboard is best for Mac?

The Keychron K15 Max. It ships with macOS keycaps in the box alongside the Windows set, so Option and Command keys are labelled correctly from day one. It connects via Bluetooth, 2.4GHz, and USB-C, all of which work natively on Mac.

What is the best wireless split keyboard?

The Elim Elytra is the best true wireless split: two separate halves with no cable between them, Bluetooth to three devices and USB-C wired. The Keychron K15 Max offers the most connection options on a single-piece board: Bluetooth, 2.4 GHz wireless and USB-C. The SpiderIsland is fully wired.

Which split keyboard is best for laptop use?

All of our reviewed keyboards pair well with laptops. For the lightest option, the ProtoArc XK03 folds to 400g. For better typing feel at a desk with your laptop, the Keychron K15 Max offers mechanical keys and wireless connectivity.

Can a split keyboard help with wrist pain?

Yes. Split keyboards allow you to position your hands at a more natural angle, reducing wrist pronation that causes strain. The Keychron K15 Max's Alice layout gently angles the keys for comfortable all-day use.

Which split keyboard is best for students?

The JOMAA Ergonomic Split Foldable Keyboard at $22 is ideal for students: affordable, portable, and comfortable for essays and notes. For heavier typing, the ProtoArc XK03 at $50 is a worthwhile step up.

Which split keyboard supports tenting?

None of the budget-friendly portable keyboards in our current guide have built-in tenting. The NocFree & offers tenting as an optional accessory. A good chair with adjustable armrests can compensate.

Which split keyboard is best for writers?

For extended writing, the Keychron K15 Max. The reason is key travel and fatigue over hours, not switch type. The ProtoArc and JOMAA are excellent on the go but their flat profile becomes tiring in the second hour of sustained writing.

What is the best ergonomic keyboard for a desk?

The Microsoft/Incase Sculpt Ergonomic Keyboard. More than a decade after release it is still the best all-round desktop ergonomic keyboard, thanks to its negative-riser design that keeps the wrists neutral, a comfortable split-dome layout, and a separate numpad. Its only real weakness is that it connects via a USB-A dongle, with no Bluetooth or wired option.

Is the Microsoft Sculpt still worth buying in 2026?

Yes. Although Microsoft stopped making it in 2023, Incase now manufactures it, and the design has aged remarkably well. It remains our top pick for a desktop ergonomic keyboard. The main caveat is availability: it is in stock at some resellers and backordered at others, so check current stock before you commit.

What is the best split keyboard for a desktop?

For two genuinely separate halves you can position shoulder-width apart, the NocFree offers an excellent wireless dual-body split, and at a desk its weight is a non-issue. For a more affordable wired option, the SpiderIsland Tech Split delivers deep mechanical key travel. Keychron's own Q11 is a brand-name alternative worth watching.

What is a negative riser or negative tilt on a keyboard?

A negative riser slopes the keyboard gently downward, away from you, rather than propping the back up. This keeps the wrists in a neutral position instead of bending them upward, which reduces the strain that causes typing pain. The Sculpt is the standout example, and it is a big part of why it remains our top desktop pick.

Should a desktop ergonomic keyboard be wired or wireless?

At a fixed desk it matters less than for travel. Wireless keeps the desk tidy, but a wired board never runs out of battery and has no dongle to lose. The Sculpt is dongle-only, the SpiderIsland is fully wired, and the Keychron K15 and NocFree offer full wireless. Choose based on how much you value a clean desk versus never charging anything.

Do I need a numpad on an ergonomic keyboard?

Only if you do heavy number entry. The Sculpt includes a separate numpad, which is ideal because it lets you centre the main keyboard and keep the mouse close, better ergonomics than a built-in numpad that pushes the mouse further away. Most of the split keyboards here omit the numpad to stay compact.

Why are columnar or ortholinear keyboards not recommended here?

Columnar and ortholinear boards like the ZSA Moonlander and Glove80 arrange keys in straight vertical columns rather than the staggered rows everyone learns to type on. They can be excellent for people who commit to them, but they require relearning how to type, and they cost far more than the boards in this guide. This guide sticks to familiar layouts you can use immediately.

Karn Bulsuk: Full Speed Ahead

Goodnight, Kyoto | Photography

noreply@blogger.com (Unknown) — Thu, 29 Oct 2026 12:19:40 +0000

Kyoto sits in a valley, and if you're there at the right time of day at the right elevation, the sun disappears behind the mountains, leaving a golden glow in contrast with the now dark city. Taken from Higashiyama.

東山より、夕映えの京都盆地。

90mm · f/8 · 1/320s · ISO 400

Best Split Keyboards 2026: Portable Ergonomic Picks Reviewed & Compared

noreply@blogger.com (Unknown) — Wed, 28 Oct 2026 11:58:51 +0000

Last updated: October 2026.

This is a guide to keyboards you can carry. Not the big ergonomic slabs that Logitech, Microsoft, and Lenovo make for a fixed desk, but the split and Alice-layout boards light and slim enough to live in a backpack and move between offices, cafes, and hotel rooms without weighing you down.

Portability is the whole point. I went looking, years ago, for a portable split keyboard I could carry between offices without noticing the weight, and I have been testing them ever since. The boards here are judged first on that question: can you actually take it with you? Carry weight, slimness, and whether it survives being thrown in a bag matter as much as how it types. Everything is tested in real use, works straight out of the box, and needs no soldering or programming.

Looking for a desk keyboard instead? If your board lives permanently on a desk, where weight and size does not matter, that is a different guide. See our companion roundup of the best ergonomic and split keyboards for desktop use.

For years the standout portable split was Keychron's, at a price nobody else came close to. That board, the one we recommended for three years, has quietly vanished from sale, and its successor is priced like a mainstream board rather than the bargain it replaced. The picks now sort into three rough tiers. At the budget end, foldable scissor-switch boards do the job for very little. In the middle sits the mainstream default, a mechanical, wireless, Alice-layout split, which is still where Keychron is hard to beat, though no longer the cheapest thing here. At the top, a handful of small makers ask two to three times as much for lighter, better-built boards.

This guide sticks to familiar, row-staggered layouts (the type you see on any laptop keyboard). We leave out the enthusiast columnar and ortholinear boards, such as the ZSA Voyagers of the world. They are fine devices, but they ask you to relearn how to type, and that is a different guide for a different reader.

The Full Collection. Our expert ergonomic keyboard and mouse guides, reviews, and previews: https://www.bulsuk.com/p/all-our-ergonomic-keyboard-reviews.html

This site is a one-person show, with keyboards bought and tested on our own dime. If a guide here has helped you, consider buying us a coffee.

☰ In this guide

Our top picks
Full comparison table
Which keyboard is right for you?
In-depth reviews
Upgrade picks
Retired benchmark
How we test
FAQs

The Best Portable and Travel Split Keyboards in 2026

What to buy now

Elim Elytra

The Elytra is a true wireless split – two separate halves, no cable between them – that weighs 420g and keeps the row-staggered layout you already type on, so there is nothing to relearn. After months of daily use it is the lightest split we have tested, the halves have never lost each other, and a single charge lasted six working weeks. It is also an excellent typing experience. It is not cheap at $249, has no Right Shift, and early units needed a firmware update before Bluetooth would hold. Everything since has been right. Rated 4.5 out of 5 in our full review.

Orderable direct from Elim

Best Alice layout

Keychron K15 Max

If you want a single-piece Alice-layout keyboard rather than two separate halves, and you do not want to spend enthusiast money, the K15 Max is the sensible default. The board we recommended for three years, the K11 Max, is now discontinued, and the K15 is built on the same platform, so the typing feel we trusted for years carries over. It is heavier than the old K11 (700g against 510g) and no longer the bargain that board was, but it remains the best all-round mechanical wireless split at this end of the market, and unlike the K11 you can actually order one. Full reasoning in our review below.

Orderable direct from Keychron

Runner up

SpiderIsland Tech Split Macro Keyboard

This keyboard gives you the feel of a full-sized board, with the added ability to position both halves at any angle and as far apart as the cable allows. Despite its no-brand pedigree, the keys are comfortable and it is easy to maintain accuracy and speed. As a fully wired keyboard you will need to spend time connecting both halves and then connecting it to the computer, and it is a thick combination. But if you want the desktop keyboard feel in a package you can still carry around, this is the one. At $82 it sits firmly in the affordable tier.

Available

Most portable

ProtoArc XK03

Want the smallest possible keyboard without sacrificing the ability to type on it? The ProtoArc folds in half yet still offers full-sized keys and bouncy scissor switches. This is the one if you need the ultimate in portability. You can type accurately, though key travel is lower to keep the package slim, so it suits emails and messages more than writing hundreds of pages. At $50, it is a no-brainer backup keyboard.

Available

Quick Comparison: All Keyboards

Scroll → to see all columns

Keyboard	Price (USD)	Rating	Weight	Switches	Connectivity	Best For
Elim Elytra Buy now	$249	●●●●● 4.5	420g (0.9 lbs)	Mechanical	BT + USB-C	Lightest true wireless split
Keychron K15 Max Best Alice	$75–130	Not yet rated	700g (1.5 lbs)	Mechanical	BT + 2.4 GHz + USB-C	Best single-piece Alice board
Keychron K11 Max Retired	was $90	●●●●● 5.0	510g (1.1 lbs)	Mechanical	BT + 2.4 GHz + USB-C	The board that set the standard
SpiderIsland Tech Split Runner up	$82	●●●●○ 4.0	563g (1.3 lbs)	Mechanical	Wired only	Desktop feel, full key travel
ProtoArc XK03	$50	●●●●○ 4.0	400g (0.9 lbs)	Scissor	Bluetooth	Ultra-portable, travel
JOMAA Split Foldable	$22	●●●●○ 4.0	400g (0.9 lbs)	Scissor	Bluetooth	Budget, students, value
BOW Ergonomic Keyboard Testing	~$55	Testing	TBC	Scissor	BT + 2.4 GHz + USB-C	Budget, also a desktop board
Jiffy 75 Testing	$279	Testing	TBC	Mechanical	Wireless	Premium 75% true split
NocFree & Upgrade pick	$269–349	●●●●○ 3.5	~1.03kg (2.3 lbs)	Mechanical	Wireless	Premium, weight no object

Prices approximate as of October 2026. USD throughout.

 The Elim Elytra and Keychron K15 Max are ordered directly from their makers. The ProtoArc XK03, JOMAA, and SpiderIsland are available on AliExpress with worldwide delivery, and on some Amazon stores.

Which split keyboard is right for you?

The comparison table tells you what each keyboard is. These answers tell you which one fits your situation.

Weight is your only criterion

If you are optimising purely for carry weight, the ProtoArc XK03 and the JOMAA are the two boards to choose between. Both weigh 400g, fold in half, and fit in a folder. The difference is build and key feel: the ProtoArc has the edge on both, while the JOMAA costs $22 and is genuinely hard to fault at that price. Neither is a keyboard you want for eight-hour writing sessions – scissor switches and a flat profile tire the fingers over time – but for emails, notes, and moving between desks, either is a no-brainer.

You want the best typing feel you can still carry

The Keychron K15 Max is the answer if you want a single-piece board, and the reason to trust it is not the spec sheet but the platform. If you want the halves apart, the Elim Elytra types just as well at 420g, and our review covers the switch feel in detail. The K11 Max, which this board replaces, was our top pick for three years and logged thousands of hours of daily use without a single complaint about the typing feel. The K15 is the same switches, the same keycaps, the same firmware, in a larger shell. It is heavier than the foldables at 700g, but that weight disappears into a bag, and the mechanical low-profile Gateron switches do not disappear into the background after an hour the way scissor switches do.

You need your hands genuinely far apart

An Alice-layout keyboard – the K15, the K11, the ProtoArc – angles the two halves but keeps them joined. If your shoulders are broad or you have been told by a physiotherapist to position your hands wider than a single unit allows, an Alice board will not solve the problem. You need a true split, where the halves are separate pieces.

Three boards in this guide do that. The Elim Elytra is the one to buy if you carry it: two wireless halves at 420g, no cable between them, and a layout you already know. The SpiderIsland does the same job wired and thick for $82, and gives you full-size key travel. The NocFree & is the best typist of the three but weighs a kilogram, so it belongs on a desk.

You are on a tight budget

The JOMAA at $22 is the honest answer. It folds in half, has a full function row, and the typing feel is considerably better than the price implies. The caveat worth naming: the build is plastic throughout, and it does not feel like a keyboard that will last five years of daily use the way the Keychron platform does. If you are a student, a light user, or someone who wants a backup board to leave in a bag, the JOMAA is outstanding value. If you type heavily every day and are simply choosing the cheapest option, the ProtoArc XK03 at $50 is a meaningfully more durable step up.

You use a Mac

The Elim Elytra. Elim ships the board with two sets of keycaps in the box – one for Windows, one for macOS – so the Option and Command keys are labelled correctly from day one without any remapping. It connects via Bluetooth, 2.4GHz, and USB-C, all of which work natively on Mac. The ProtoArc XK03 and JOMAA also pair via Bluetooth without issues. For a daily driver that looks right as well as works right, the Elim Elytra is the only choice.

You write for hours at a stretch

Reach for the Keychron K15 Max over any of the foldables, and the reason is not switch type – it is key travel and fatigue. The ProtoArc and JOMAA are flat, slim boards by design. That flatness is what makes them portable, and it is also what makes them tiring: after an hour or two of sustained writing, the shallow travel and hard base start to register in the fingers. The K15's low-profile Gateron switches have enough travel to provide genuine tactile feedback on every keystroke without bottoming out hard. You will feel the difference somewhere in the second hour. If you are writing thousands of words a day and the keyboard is a tool you rely on, spend the extra and get the mechanical board.

Looking for a desk keyboard, not a travel one? If portability is not your priority, our companion guide covers the best ergonomic and split keyboards for desktop use, including the Microsoft/Incase Sculpt and heavier premium splits.

On the horizon

The Compose Keyboard – $300+

Our preview: https://www.bulsuk.com/2026/04/the-compose-keyboard-split-keyboard.html

A wireless ergonomic split built around the Apple ecosystem, currently on Kickstarter and slated for 2027. It is slim, uses scissor switches, and replicates the design choices of the Apple Magic Keyboard. Tracking only, not a pick. We will continue to follow its progress.

Keyboards we are reviewing

I buy and test keyboards on a rolling basis, and makers occasionally send units on the same terms as everything else here: no copy approval, no pre-publication review, rating not negotiable. A few are in the queue right now.

Jiffy 75: in hand, testing since September. A 75% true wireless split from Jezail Funder, the maker of the Cornix, with CNC aluminium halves, a 37g silent switch, two rotary knobs and a 2,800mAh battery. At $279 it sits directly against the Elytra, and the trade it makes is weight for battery and a function row. Unit supplied by the maker. Full review to come.
BOW Ergonomic Keyboard: in hand, testing not yet started. A cheap (~$55) unibody split, slim at 20mm, light, and wireless (Bluetooth, 2.4GHz dongle, and wired USB-C), so it may earn a place here as a budget travel option as well as in our desktop guide.

Reviews

Keychron K15 Max

Why we recommend it before testing: Our full reasoning here.

Price: USD 75–130 depending on config and discounts

Rating: Not yet rated – recommended on the basis of the shared K11 platform.

The K15 Max is what we now point most buyers to, and the reason is simple: it is the K11 in a larger shell. It uses the same low-profile Gateron switches, the same LSA keycaps, the same QMK firmware, and the same aluminium body, scaled up to a 75% Alice layout that adds a function-and-number row. We spent three years with that platform, so everything that made the K11 good to type on is a known quantity. We recommend the K15 on the strength of that shared platform rather than a separate test of our own. Read our full reasoning.

SpiderIsland Tech Split Macro Keyboard

Read full review: https://www.bulsuk.com/2024/12/spiderisland-split-keyboard-review.html

Price: USD 82

Rating: ●●●●○ 4.0/5.0 Excellent

Pros

Gives you the full-sized keyboard experience with good key travel
Able to separate both sides of the keyboard and position them in the way you like
Software customisable
Fairly light coming in at 563g (1.3 lbs)
Can be shipped to anywhere in the world

Cons

Rather thick when put together, and is bulky when travelling
No wireless connectivity
Minimal documentation on how to use keyboard customisation software

When it comes to wish fulfilment, the SpiderIsland split keyboard will satisfy all those who miss the key travel of IBM keyboards and wished they had a portable ergonomic version they could bring on-the-road. The keyboard is split into two separate sides, and users can position both halves at any angle they want, as close or far apart as they prefer. This supreme customisability makes it a very comfortable keyboard to type on. I can average 83 WPM, which is the same as the speed on the Microsoft/Incase Sculpt Ergonomic keyboard.

Weighing in at 563g (1.3 lbs), this is a fully wired device, and requires a cable to connect both sides together, and another cable from the left half into your computer. While set up time is increased by 30 seconds compared to a Bluetooth keyboard, it does mean you never need to worry about running out of battery.

While it remains portable and will easily fit into a backpack, it is naturally thicker than the other reviewed options, given the amount of key travel offered. The keys are customisable using the Bootmapper software, and it has a white backlight in case you're ever using it in the dark.

ProtoArc XK03

Read full review: https://www.bulsuk.com/2024/04/protoarc-xk03-foldable-ergonomic.html

Price: USD 50

Rating: ●●●●○ 4.0/5.0 Excellent

Pros

Incredibly portable, folds in half, and is super slim
Has full sized keys with decent key travel, and comfortable ThinkPad-esque concave keys, providing an great typing experience
Very affordable compared to alternatives
Portable and reasonably priced
High quality build

Cons

Longer typing sessions may be tiring due to the slim form factor
If set up incorrectly, slim form factor, and lack of negative tenting, may place pressure on your wrists

For road warriors looking for a travel-friendly, super compact, split ergonomic keyboard with the smallest practical form factor, the ProtoArc XK03 is an excellent choice. The XK03 has full sized keys despite its form factor, comes with wireless connectivity via Bluetooth, and is a delight to use. It uses the standard qwerty layout and requires no retraining. The keys clearly draw inspiration from Lenovo's ThinkPads, with concave keys which cup the fingers, good key spacing, and a surprising amount of key travel given its size, ensuring typing accuracy and speed.

It selling point is that it folds in half, making it easy to throw into a backpack (or even a folder) and forget it's even there, and weighs only 400g (0.9 lbs).

Although it does a lot right, the form factor results in some minor gripes. The keys aren't exactly centred, and when using it for the first week, you may end up hitting the hard sides of the keyboard. Because it's so flat, longer typing sessions may strain your fingers as it can feel like you're typing on wood after a while. Nevertheless, the ProtoArc XK03 remains an excellent choice if you need the smallest split keyboard possible.

Best Budget Split Keyboards Under $50

JOMAA Ergonomic Split Foldable Keyboard

Full review: https://www.bulsuk.com/2026/01/jomaa-portable-folding-split-keyboard.html

Price: USD 22

Rating: ●●●●○ 4.0/5.0 Excellent

The JOMAA Ergonomic Split Keyboard is a flat, ultraportable, foldable split keyboard, available from AliExpress for between USD 22-30. It is the cheapest keyboard we've come across in this form factor, and it punches far above its weight, offering incredible value for the price, with a comfortable, laptop feel, a full function row, and good ergonomics. If you're looking for a budget version of the Protoarc XK03, the JOMAA Ergonomic Split Keyboard is practically on par, giving you way more than what you paid for.

[message]

ProtoArc XK03 vs JOMAA Ergonomic Split Keyboard: which is best? Check out our comparison here: https://www.bulsuk.com/2026/02/protoarc-xk03-vs-jomaa-ultraslim.html

Upgrade picks

Under $150, Keychron is effectively the only game in town for a good standard-layout split. Everything genuinely a step up comes from a small maker and costs two to three times as much. There is very little in between, which is why these are a separate tier rather than the next rung on a single ladder. If you want to spend more, here is what money buys.

NocFree &

Read full review: https://www.bulsuk.com/2026/08/nocfree-ampersand-full-review.html

Price: USD 269-349

Rating: ●●●●○ 3.5/5.0 Good

Pros

Outstanding, cushioned low-profile typing feel with hot-swappable switches
True standalone wireless split (no cable between the halves)
Substantial build weight stops any desk sliding
Quick, highly accessible support team via Discord
Clean browser-based key remapping and layer support

Cons

Heavy, and an oversized case negates true portability
Severe battery drain with backlighting on (1–2 days max)
Out-of-the-box QC issues (missing feet and screws) and risky firmware updates
Keycap printing looks soft and fuzzy
No battery indicator on the board itself
Expensive

A capable dual-body split (two fully separate wireless halves) that is genuinely backpack-portable, but heavy. At 1.03 kg it weighs about as much as a Lenovo X1 Carbon laptop, so it straddles portable and desktop rather than being a grab-and-go board. It earns a place here for the reader who wants a premium split and does not mind carrying laptop-weight, but if portability is your priority the lighter picks above win. Its natural home is really the desk, and we cover it more fully in the desktop ergonomic roundup. After several months living with it, our verdict is 3.5 out of 5: the typing feel, ergonomics and build are outstanding, but out-of-the-box quality control issues and a risky firmware update held it back.

Elim Elytra – $249

Read full review: https://www.bulsuk.com/2026/09/elim-elytra-review.html

Price: USD 249 (case USD 39)

Rating: ●●●●● 4.5/5.0 Outstanding

Pros

True wireless split at 420g, the lightest we have tested
Familiar row-staggered layout means almost no adaptation time
Four weeks of daily use on a single charge
Hot-swappable low-profile switches with a cushioned feel
Halves never lost each other in seven weeks

Cons

Shipped on firmware that could not pair over Bluetooth, since fixed
Padded case erases most of the weight advantage in a bag
No Right Shift
No battery percentage, only a low-battery blink
Expensive, especially with accessories

An ultralight wireless dual-body split from a small Hong Kong manufacturer, and after seven weeks of daily use our top pick for anyone who wants the halves apart and needs to carry them. It weighs 420g on our scale, keeps the row-staggered layout every laptop uses, and we were at full typing speed within an hour. The two halves have never dropped their link to each other. Battery matched Elim's four-week claim on 300mAh cells, which is good power management rather than a big battery. The catches are real but historical or minor: early units shipped on firmware that could not hold Bluetooth until updated, the layout quietly omits Right Shift, and the padded case the halves need takes the packed weight to 584g, so a single-piece board in a sleeve still packs flatter. The halves also snap together magnetically into an Alice shape, though we did not find that comfortable. 4.5 out of 5. Our full review covers the switch comparison, the firmware history and Elim's answers to our questions.

Disclosure: I backed the Elytra on Kickstarter and paid for the unit. Elim had no input into the review.

Retired benchmark

The board below is no longer sold, and we have retired it from the recommendations. It stays on the page because it set the standard every other board here is measured against, and because readers still ask about it.

Keychron K11 Pro/Max – retired benchmark

Retired, October 2026. Keychron discontinued the K11 Max in mid-2026 and we have now retired it from our recommendations. It shows sold out with a notify-me only on Keychron's own store, and has gone from AliExpress and Amazon. We keep this review as the benchmark record and because its successor, the Keychron K15 Max, is built on the same platform. If you already own a K11, hold on to it.

Buying one second-hand. If you find a K11 in good condition, it is still worth having – the typing feel has not been bettered at the price. Two caveats. There is no warranty, so check every key and both connection modes before paying. And the K11 Max carries a smaller battery than the K11 Pro, so a two-year-old Max will have lost more capacity: between the two, buy the Pro.

Read Full Review: https://www.bulsuk.com/2024/04/keychron-k11-pro-review.html

Read Long Term Use Review: https://www.bulsuk.com/2026/03/keychron-k11-promax-longterm-review.html

Price: USD 90

Rating: ●●●●● 5.0/5.0 Exemplary

Pros

Typing experience is great
Portable and reasonably priced
High quality build from a reputable manufacturer
Software to customise keyboard user friendly and well documented
Keycaps and switches can be replaced by the user
Fairly light at 510g (1.1lbs)

Cons

Typing feels a bit cramped, takes 2-3 weeks to get used to
Keys are slightly smaller than full sized keyboards

The Keychron K11 Pro/Max is an excellent split ergonomic keyboard, utilising an Alice layout, which is small and light enough to throw into a backpack and not really notice it. The build quality is solid and remains so even after years of usage (read long term review), and if you buy the RGB backlight version, it is fully customisable with replaceable keycaps and switches. It has Bluetooth, and supports USB-C for both wired connectivity and charging. At launch it was remarkably affordable at around USD 90, which was a large part of its appeal.

In our tests, the keyboard is responsive and the keys feel just right. I can reach around 83 WPM, which is on par with the Microsoft/Incase Sculpt Ergonomic keyboard which I use as my daily driver. More importantly, the keys cradle the fingers and has good travel to provide sufficient tactile feel on each keypress.

The customisation software is very user friendly, and available through a downloadable executable, or directly through a web browser. For USD 20 extra, you can also buy a matching travel pouch which fits the keyboard perfectly and looks good.

There are a few cons, mainly that the layout feels a bit cramped despite having full sized keys, with typing accuracy taking a hit at first. The Keychron K11 Pro/Max also has a positive, as opposed to a negative raiser, which reduces the ergonomic nature of the keyboard. These are minor flaws and you can easily adapt to them after a short while.

The Keychron K11 Pro and Max are virtually identical keyboards. The key differences are that the K11 Max adds acoustic foam, supports 2 GHz wireless connectivity, but has a smaller battery.

Keyboards we considered, but did not pass screening

One of the things we do is perform a paper review over the options before deciding to purchase a keyboard for review, and the following are the ones which did not pass initial screening:

Keychron B11 Pro

The B11 Pro looks like a dream traveller's keyboard: foldable, slim enough to fit into a manila folder, and unlike the ProtoArc XK04, you can remap the keys. But that's its only edge, and it's most of what the US$65 price buys over the ProtoArc XK03 (US$50) and JOMAA Portable Folding Split Ergonomic Keyboard (US$30). On a compact board, remapping solves a problem you don't have, and the muscle memory it rewards isn't worth building on a keyboard you only use on the move. It's a feature for tinkerers, not travellers. We couldn't justify the premium, and won't bring it in for review.

NocFree Lite Wireless Split Keyboard

This keyboard looks beautiful, portable, and well designed. It even has built-in tenting, which is unique among the ones we've looked at so far. There are a few unfortunate issues with the NocFree Lite though. Firstly, the "?" key is weirdly positioned to the right of the arrow keys, inviting mistaken key presses until you retrain muscle memory. It also relies on a dongle for wireless connectivity, and does not have Bluetooth. Finally, it is higher than our budget, priced at USD 179. Its successor, the NocFree & [our review], addresses the connectivity and layout issues.

R-Go Split Keyboard

At first, this appeared to be a serious contender, as the form factor, size, and chiclet keys all tick the right boxes. Unfortunately, the reviews on Amazon were mixed (3.2 as of April 2026), with many talking about quality problems and high failure rates. It is difficult to justify spending USD 125 on a keyboard which may break in a few months.

ZSA's Voyager

The Voyager keyboard ticks off most of the boxes, being small, thin, light, customisable and well-built for travel. The fact it uses slim keys, and can be tented won it extra points. However, the lack of certain dedicated keys, such as arrow keys, may appeal to enthusiasts who are looking for the most efficient keyboard layout possible, but we felt it would not be appreciated by an everyday user coming from standard keyboards. Although a beautifully designed and engineered piece of hardware, its price of USD 365 was also beyond our budget. With cheaper and just as functional options available, we decided to skip the Voyager.

Goldtouch Go!2 Bluetooth Wireless Mobile Keyboard

When looking at this keyboard, there were some glaring issues which quickly removed it from the review list. Firstly, the layout looked strange and incredibly cramped. The arrow keys seemed to be squished in as an afterthought, and the home, end, page up and down keys on the left looks like it invites constant accidental presses. Secondly, it is 6.35 cm (7 inches) in height when folded, which is even thicker than the (already thick) SpiderIsland Split Keyboard reviewed above. As portability and compactness is a key requirement to ensure travel friendliness, the Go!2 keyboard quickly forfeited its candidacy.

Keyboards we reviewed, but can't recommend

Generic Foldable Split Ergonomic Keyboard

Rating: ○○○○○ 0.0/5.0 Dismal

Full review: https://www.bulsuk.com/2025/01/generic-foldable-split-ergonomic.html

This is a keyboard manufactured by an OEM and commonly sold under a variety of brand names. Despite its small size and apparent similarity to the ProtoArc XK03 reviewed above, it is a terrible keyboard. The keys are too small and fingers keep slipping off, negatively impacting both accuracy and feel. Worse, our review unit kept dropping keystrokes, meaning it fails one of the basic functions of a keyboard.

Looking for a desktop split ergonomic keyboard?

If your keyboard lives on a desk, where weight does not matter, you have a wider field to choose from, including boards too heavy for this travel guide. Our dedicated desktop roundup covers them, led by the Microsoft/Incase Sculpt Ergonomic Keyboard.

See our desktop guide: The Best Ergonomic and Split Keyboards for Desktop 2026
Or read our full review of the Sculpt: https://www.bulsuk.com/2025/09/sculpt-ergonomic-keyboard-review.html

How We Test Split Keyboards

We've formed the list of requirements based on a user who is regularly on the move, travelling between offices and client sites, and needs to carry around a portable, yet ergonomic set up, which will fit into a backpack. As a regular rider of public transport, we prefer lighter weight keyboards. In addition, this user persona is someone who primarily types out reports and emails, and while does the occasional spreadsheet, does not need a number pad.

Based on this, we've split the selection criteria into three categories:

Must-haves:

Affordability is a core principle: our main picks aim to stay around or under US$150, the sweet spot balancing price and performance. That spans everything from budget foldables up to the mainstream mechanical splits. We also keep a clearly-labelled upgrade tier above this for the small number of premium boards worth knowing about, so price guides our recommendations rather than acting as a hard cut-off.
Comes pre-built and useable straight out-of-the-box: While soldering your own keyboard offers infinite customisability, we're looking for something which just works out-of-the-box.
Is mobile and travel friendly: it needs to be slim and compact enough to fit into a medium backpack. We favour lighter boards, ideally under 500g, because we carry these on public transport every day. This is a guiding preference rather than an absolute limit: a heavier board can still earn a place if it is the best available option, but we will always tell you what it weighs and who it is not for.
Feels right on the fingers: We're big fans of the Lenovo ThinkPad keyboards, the Microsoft/Incase Sculpt Keyboard (our review), and the old school IBM Model M keyboard. These are all examples of great keys which provide excellent tactile feel and support on your fingers. They are shaped using the right materials to prevent your fingers from easily slipping off, resulting in comfortable, accurate and satisfying typing.
Has enough key travel: Based on the Goldilocks principle, key travel needs to be just right. Too little, and it feels like you're typing on a plank of wood. Too much, and it feels like you're typing on a 1960s manual typewriter. Either extreme leads to fatigue, and pain.
Has a standard row staggered QWERTY layout: This will help to reduce the amount of retraining compared to other enthusiast keyboards (such as those with otholinear layouts). We're not big fans of key layers, which requires users to use a combination of key presses to access basic functions. Number row keys, a dedicated escape key, and real arrow keys, are non-negotiables.
Does not have a number pad: A traditional full sized keyboard has a number pad on the right, forcing users to reach further from the centre to use the mouse. This is poor ergonomics.
Is a 65%-80% keyboard: This is a good balance between reducing unnecessary keys to reduce the overall keyboard size, while still allowing seamless switching from a desktop keyboard. It also means no number pad, which takes up too much space. [Read more: What's a 65%-80% keyboard?]

Nice-to-haves:

These are the requirements which would certainly be helpful, but does not push a keyboard out of consideration.

Bluetooth: We prefer a wireless keyboard to reduce set up time when moving from site to site, and to avoid tangled wires. We'd also take Bluetooth over a wireless dongle any day. Another dongle is something to potentially lose.
Negative tilt: This is where you can angle the keyboard forwards, so that the hands are in a more natural position. While this is comfortable, a good chair with adjustable arm rests can make up for the lack of negative tilt, by ensuring one's fingers can hover comfortably over the keyboard. A good example is the negative riser that comes with the Microsoft/Incase Sculpt Ergonomic Keyboard.
Tenting: This is where you can raise the middle of the keyboard higher, which is good for preventing hand pronation, or bending your wrists up when you type. While this is nice-to-have, similar to negative tilt, good adjustable armrests can help to make up for a lack of this.

Optional:

Mechanical switches: Both mechanical switches and scissor switches can offer excellent typing experiences, with the keyboard design being a more important factor in typing comfort.
Backlighting: RGB and other backlights might be nice to play around with, but working in bright offices means they are not essential.

FAQs

[accordion]

What is the best portable split ergonomic keyboard?

It depends on whether you want one piece or two. For a true split with separate halves, the Elim Elytra – 420g, wireless, a layout you already know, rated 4.5 after seven weeks of daily use. For a single-piece Alice layout at a lower price, the Keychron K15 Max. It uses a 75% Alice split layout, connects via Bluetooth, 2.4GHz, or USB-C, and starts around $75. It is built on the same platform as our long-time pick the Keychron K11 Pro/Max, which Keychron has now discontinued. If you already own a K11 Max, hold on to it.

What is the cheapest split ergonomic keyboard?

The JOMAA Ergonomic Split Foldable Keyboard is available for around USD 22–30 on AliExpress. Despite its low price, it offers a comfortable laptop-like typing feel, a full function row, and folds in half for portability. It punches well above its weight.

Can you travel with a split keyboard?

Yes. All the keyboards we review are selected for portability. The lightest option, the ProtoArc XK03, weighs 400g and folds in half to fit in a folder. The heaviest, the SpiderIsland, is 563g. All fit comfortably in a standard backpack or laptop bag.

Do I need a mechanical keyboard for ergonomics?

No. Both mechanical and scissor switch keyboards can provide excellent ergonomic typing experiences. Keyboard design – the split angle, key spacing, and form factor – matters more than switch type. Several keyboards in our list use scissor switches and score just as well for comfort as the mechanical options.

What is an Alice layout keyboard?

An Alice layout is a type of split keyboard where the two halves are angled apart but remain in a single connected unit. This promotes a more natural wrist position while keeping the keyboard compact and portable. The Keychron K11 Pro/Max is a popular example of an Alice layout keyboard.

Which split keyboard is best for Mac?

The Keychron K15 Max – it ships with macOS keycaps in the box, not just Bluetooth compatibility. Full reasoning in the use cases section above.

What is the best wireless split keyboard?

The Elim Elytra is the best true wireless split, with two separate halves and no cable between them. The Keychron K15 Max offers the most connection options on a single-piece board: Bluetooth, 2.4 GHz wireless, and USB-C. The ProtoArc XK03 and JOMAA Foldable Split connect via Bluetooth only. The SpiderIsland is fully wired.

Which split keyboard is best for laptop use?

All of our reviewed keyboards pair well with laptops – that's the core use case. For the lightest option, the ProtoArc XK03 folds to 400g. For better typing feel at a desk with your laptop, the Keychron K15 Max. See the use cases section above for the full breakdown.

Can a split keyboard help with wrist pain?

Yes. Split keyboards allow you to position your hands at a more natural angle, reducing the wrist pronation that causes strain with standard keyboards. All keyboards in this guide use a split layout to promote better wrist alignment. The Keychron K11 Pro's Alice layout gently angles the keys without requiring a full two-piece separation, which many users find comfortable for all-day use.

Which split keyboard is best for students?

The JOMAA Ergonomic Split Foldable Keyboard at $22 is ideal for students. It's the most affordable option in our guide, folds in half for portability, and offers a comfortable typing experience for essays and notes. For students who type heavily and want a longer-lasting keyboard, the ProtoArc XK03 at $50 is a worthwhile step up.

Which split keyboard supports tenting?

None of the budget-friendly portable keyboards in our current guide have built-in tenting. The Nocfree & offers tenting as an optional accessory, and the NocFree Lite has built-in tenting (though we did not recommend it due to other design issues). For the keyboards we do recommend, a good chair with adjustable armrests can compensate for the lack of tenting.

Which split keyboard is best for writers?

For extended writing sessions, the Keychron K15 Max. The reason is key travel and fatigue, not switch type. Full reasoning in the use cases section above.

If your wrist pain extends beyond the keyboard, a Wacom graphics tablet used as a mouse replacement is worth considering alongside a split keyboard. We cover that here.

Updates

[tab]

2026

1 Jan 2026: Completed review of JOMAA Portable Foldable Split Keyboard, and added details.
2 Feb 2026: Added comparison of JOMAA Portable Foldable Split Keyboard with the ProtoArc XK03
22 Mar 2026: Added Keychron K11 Max/Pro long term review, and how it continues to perform well. Changed expected arrival date of the Elim Elytra for review.
3 April 2026: Added link to compare the Sculpt Mouse vs Logitech Vertical MX.
9 April 2026: Added FAQs. Removed Keyboards We May Review section, as the list is pretty slim at the moment.
23 April 2026: Restructured article. Added content on use cases, additional FAQs.
25 April 2026: Added the Compose keyboard to a new list of keyboards we're tracking
9 May 2026: Updated hero image, and initial thoughts on the Nocfree &
6 June 2026: Updated Elytra Elim's shipping status.
20 July 2026: Added note about the Keychron B11 Pro, and why we will not bring it in for review.
2 August 2026: Published our full review of the NocFree & (rated 3.5/5.0). Updated its comparison-table entry and upgrade-pick write-up.
25 October 2026: Elim Elytra review finalised at 4.5/5.0 and made our top pick for a true wireless split. Keychron K15 Max relabelled as best Alice layout. Keychron K11 Pro/Max moved to a new retired benchmark section with second-hand buying advice, and its rating label corrected to Exemplary. Jiffy 75 added to boards under review. Use cases, comparison table, FAQs and schema updated.
16 August 2026: Major rebuild. Keychron K11 Max marked discontinued and replaced as top pick by the Keychron K15 Max. Added upgrade tier (Elytra, NocFree &), the BOW Ergonomic Keyboard to boards under review, availability column to the comparison table, and a companion desktop guide. Updated schema and FAQs.
17 August 2026: Expanded use cases section with six buyer scenarios. Brought all rating colours to site standard. Fixed NocFree & half-circle rendering. Shortened four FAQs to point to use cases section.
28 October 2026: Revised recommendation on best portable board from K11 Pro to Elim Elytra.

2025

27 Jan 2025: Originally published
3 Feb 2025: Added notes around the Keychron K11 Max
15 Feb 2025: Updated the Keychron K11 Pro/Max price
28 Apr 2025: Added the NocFree Lite into the list of untested keyboards
15 June 2025: Added reviewed keyboards which we can't recommend
27 July 2025: Updated price of the Spidersisland keyboard (decreased from $118 to USD 82)
29 July 2025: Added the Nocfree & keyboard as an upcoming contender
9 August 2025: Added omission of a number pad in the list of must-have requirements
6 September 2025: Added "Keyboards we may review" section
17 September 2025: Added our desktop keyboard recommendation
4 October 2025: Added the JOMAA Ergonomic Split keyboard to our list of keyboards we're currently testing
9 Oct 2025: Updated typing speeds for the Keychron K11 Pro/Max, and the SpidersIsland Split Keyboard
1 Nov 2025: Added initial thoughts on JOMMA keyboard
1 Dec 2025: Added the Elytra Split keyboard on the list of things we're tracking. Clarified that the Keychron K11 Pro and Max are virtually the same keyboard.
22 Dec 2025: Added details for the Elim Elytra

How to Make Hong Kong Lemon Tea (凍檸茶 Dong Ling Cha): Authentic Recipe

noreply@blogger.com (Unknown) — Sat, 24 Oct 2026 13:20:51 +0000

Making an authentic glass of Hong Kong lemon tea is easy, here's the full recipe.

Prep: 2 min · Boil: 7–10 min · Assembly: 4 min · Total: about 15 min · Makes: about 12–14 glasses · Serve: over ice

Dong ling cha (凍檸茶) is the drink sitting right next to milk tea on every cha chaan teng menu, made from the exact same tea. Most English-language recipes miss that connection. They treat it as a separate, generic iced tea, reaching for Earl Grey, tea bags, and a gentle steep. None of that is how it is actually made.

The tea base is identical to Hong Kong milk tea. Loose Ceylon and Assam black tea, boiled hard, not steeped, then diluted with water instead of milk. The dilution ratio varies from shop to shop, from roughly equal parts tea and water to about one part tea to two parts water. Lemon goes in fresh and sweetener is added at the glass. The tea itself is never brewed from a bag.

Only want one glass? Don't try to shrink the brewing step. A tea base this small doesn't extract reliably, and no cha chaan teng brews one cup at a time either. Make the full batch, keep the rest refrigerated for up to 2–3 days, and assemble a fresh glass whenever you want either HK lemon tea or HK milk tea.

Nobody sweetens this for you at a cha chaan teng, and everybody has an opinion about how many lemon slices is correct. This is the version I make at home, using the same tea base as the milk tea, poured cold and on the lighter side of sweet, which is how I like it.

→ Jump straight to the recipe

Ingredients

Recipe

Hong Kong Lemon Tea (凍檸茶, Dong Ling Cha)

Prep 2 min Boil 7–10 min Assembly 4 min Total ~15 min Makes ~12–14 glasses

Ingredients

1,200 ml (5 cups) water
15 g (0.5 oz) Ceylon black tea
10 g (0.35 oz) Assam
4 g (0.15 oz) bo lei (普洱, Pu'er)
Water, to dilute each glass (about 100–125 ml per glass)
3–5 slices fresh lemon per glass
Sugar or syrup, to taste (2–3 tsp sugar, or 1–2 small syrup portions, per glass, as a starting point)
Plenty of ice

Method

Boil the tea. Bring the 1,200 ml of water to a boil. Add the Ceylon, Assam and bo lei, and boil for 7–10 minutes at a full rolling boil, not a gentle steep. This is the same tea base as Hong Kong milk tea. It'll taste harsh on its own. That's correct.
Rest and strain. Turn off the heat, rest 2 minutes, then strain through a cloth sock strainer into a heatproof jug. This is your concentrate, about 1.2 litres.
Sweeten now if using sugar. Sugar needs to go in while the concentrate is still hot, or it won't dissolve properly, about 2–3 tsp per glass you plan to pour. If you're using syrup instead, skip this and add it at the glass.
Assemble. Fill a tall glass generously with ice. Pour in 75–100 ml of the tea concentrate and 100–125 ml of water. Shops vary this ratio, from about one-third tea to two-thirds water, up to closer to equal parts, so adjust to how strong you like it. Add 3–5 fresh lemon slices.
Sweeten to taste and serve. If you haven't already sweetened the concentrate, add syrup at the glass now, 1–2 small portions as a starting point, though this really is down to personal preference. Serve with a straw, and press the lemon gently against the glass as you drink to release its juice and oils, not hard enough to crush the bitter white pith underneath the peel.

Tea base shares the same method as our Hong Kong milk tea recipe. If you have found yourself here after a lemon tea gone wrong, the companion piece on where most recipes go wrong covers the mistakes in more depth.

Method

Boil the water. Bring all 1,200 ml (5 cups) to a boil.

Add the tea and boil for 7–10 minutes. A full rolling boil, not a gentle steep. It's the same tea base as Hong Kong milk tea, and the same logic applies: boiled hard, it'll taste harsh and tannic on its own. That's correct – the dilution and lemon are what balance it, the same way milk balances the hot version.

Rest 2 minutes, then strain through a cloth sock. A single strain is fine. This gives you your concentrate, about 1.2 litres.

Sweeten the concentrate now, while it's hot, if you're using sugar. Granulated sugar won't dissolve properly once the tea has cooled, so if sugar is your sweetener, stir it in at this stage, about 2–3 teaspoons for every glass you plan to pour. If you're using syrup instead, it dissolves fine cold, so you can leave it until the glass is assembled.

Assemble each glass. Fill a tall glass generously with ice, then pour in 75–100 ml of the concentrate and 100–125 ml of water. Shops vary this ratio quite a bit – some run closer to one-third tea to two-thirds water, others nearer equal parts – so treat this as a starting point and adjust to how strong you like it. Add 3–5 fresh lemon slices.

Sweeten and serve. If you haven't already sweetened the concentrate, add syrup now, to taste. Serve with a straw, and press the lemon gently against the glass as you drink – enough to release its juice and oils, not hard enough to crush the bitter white pith underneath the peel.

Make it ahead. The concentrate keeps in the fridge for 2–3 days and holds up well, so there's no reason to brew a smaller batch just for one glass. Mix each glass fresh rather than pre-diluting the whole batch – the concentrate on its own keeps better than the diluted tea does.

A note on hot lemon tea

Hot lemon tea (熱檸茶, yit ning cha) is made the same way, minus the ice. The tea base, dilution, and self-serve sweetening all carry over. This recipe, though, is built for the iced version, where the drink actually comes alive.

FAQ

What is dong ling cha?
Dong ling cha (凍檸茶) is Hong Kong-style iced lemon tea, a cha chaan teng staple made from the same black tea base as Hong Kong milk tea, diluted with water, served over ice with fresh lemon slices. Dong (凍) means iced, ling (檸, short for 檸檬) is lemon, and cha (茶) is tea. The hot version is yit ning cha (熱檸茶); the general term covering both is ning cha or ning mung cha (檸茶 / 檸檬茶).

Is Hong Kong lemon tea the same tea as Hong Kong milk tea?
Yes, at cha chaan teng they share the same Ceylon and Assam black tea concentrate (茶胆). The difference is what happens after brewing: milk tea is cut with evaporated milk, lemon tea is diluted with water and finished with lemon. The exact dilution ratio for lemon tea varies by shop, ranging from roughly equal parts tea and water to about one part tea to two parts water.

Should Hong Kong lemon tea be brewed with tea bags?
No. Authentic cha chaan teng lemon tea is made from loose, broken-grade Ceylon and Assam black tea, boiled hard, not steeped from bags. Bagged tea and gentle steeping produce a thinner, weaker cup than the real thing.

Do you boil or steep the tea for Hong Kong lemon tea?
Boil. The tea base is hard-boiled, exactly as for Hong Kong milk tea, not steeped gently at a controlled temperature. Many English-language recipes call for steeping, but that is not the traditional method.

Is Earl Grey used in traditional Hong Kong lemon tea?
No. Traditional cha chaan teng lemon tea uses a Ceylon and Assam black tea blend. Earl Grey lemon tea is a real but modern variation served at a small number of newer-style cafes, valued for its bergamot fragrance, but it is not the traditional cup.

Can I use Lipton tea for Hong Kong lemon tea?
It works as a fallback if you cannot access proper loose Ceylon or Assam tea, since Lipton's Ceylon leaf gets you reasonably close. But it is a compromise, not what cha chaan teng actually brew, they use unbranded wholesale loose tea, not a retail bagged brand.

Is Hong Kong lemon tea pre-sweetened?
Not always. Many shops serve it unsweetened with syrup on the side so you can add your own. This recipe follows that convention: sweeten at the glass, to taste, rather than pre-sweetening the batch.

How do you eat or use the lemon slices in Hong Kong lemon tea?
Press them gently against the side of the glass with a straw or spoon as you drink, to release juice and aromatic oils. Do not crush them hard. The bitterness some people get from homemade lemon tea usually comes from rupturing the white pith underneath the peel, not from the lemon itself.

Can I make just one glass of Hong Kong lemon tea?
Not by shrinking the brewing step. A tea base this small doesn't extract reliably at a tiny volume, and cha chaan teng never brew one cup at a time either. Make the full batch, refrigerate what you don't drink straight away, and assemble a fresh glass whenever you want one – the diluted tea keeps for 2 to 3 days in the fridge, though it's best not to push it much beyond that.

The 2005 Amagasaki Rail Disaster: What the Official Report Reveals About Japan’s Worst Train Crash Since 1963

noreply@blogger.com (Unknown) — Tue, 13 Oct 2026 12:41:31 +0000

A record of the Fukuchiyama Line derailment of 25 April 2005, drawn from the official investigation

I.

At 9:18 and 54 seconds on a clear Monday morning, the first carriage of a seven-car commuter train left the rails on a right-hand bend in the Hyogo city of Amagasaki and went over onto its side. The passengers who survived it described the same thing in different words. A teenage girl in the first car, asleep against the window, woke to the body of the train shaking hard and felt it tip, all at once, before it struck anything.¹ A man standing by the third door watched the curve come up too fast and thought, in the moment, that it did not look like a train that could make the turn. He felt the right side of the carriage go light – ‘floating,’ he said later – and then it never came back down.² A woman standing with a friend heard the windows rattle and then a long, high, metallic shriek she had never heard from a train before. The car tilted to perhaps forty-five degrees and the lights went out.³ From a balcony overlooking the line, a resident watching trains with a child heard a grinding that was not the usual sound, saw the right wheels of the lead car lift, saw a small puff of dust, and then watched the carriage slide off the track towards the apartment block in what looked like slow motion.⁴

The lead car struck the building’s ground-floor mechanical car park and folded. The second car, still coupled, swung in behind it and wrapped around a concrete pillar, its cross-section crushed from a rectangle into a diamond. Cars three, four and five derailed behind them. Six and seven stayed on the rails. The last carriage came to rest at 9:19:04, ten seconds after the first had gone over.⁵

One hundred and six passengers died. So did the driver. Five hundred and sixty-two people were injured.⁶ It was the deadliest railway accident in Japan in more than forty years, and it fell to West Japan Railway, the company that had inherited this stretch of line when the national railway was broken up in April 1987.^6a

The proximate cause was never in doubt. The train entered a curve with a limit of 70 km/h at approximately 116 km/h, and at that speed the physics were unforgiving.⁷ What the investigation spent two years and some 270 pages establishing was why a licensed driver, in good health, with no alcohol in his blood and a clean run that morning until Takarazuka, let a train he had driven many times run 46 km/h over the limit into a bend he knew was there. The answer the Aircraft and Railway Accidents Investigation Commission gave had little to do with machinery. It concerned where his attention had gone in the last fifty seconds, and the company that had built the reasons for it to go there.

II.

The driver was twenty-three. He had joined the company in April 2000, worked a station platform, passed the examination to become a conductor, and then the one to become a driver. He had received his licence in May 2004, eleven months before. He lived at home, the third of four children, and by his family’s account the most cheerful of them. He snowboarded until the snow ran out. He had told them he wanted, eventually, to drive the Shinkansen.⁸

He had spent the night at the company’s crew dormitory at Hanaten, in eastern Osaka, and reported for roll call at 6:08, thirteen minutes early. The supervisor who took it noticed nothing unusual: his voice was normal, he did not look tired.⁹ His roster that day was a long loop. He would take an empty train out to Matsui-yamate, bring it back in service to Kyobashi, continue under a new train number through the underground Tozai Line to Amagasaki, run it empty up the Fukuchiyama Line to Takarazuka, then change ends and drive it back as the 9:03 rapid service to Doshisha-mae: train 5418M, seven cars, the same set he had been driving since dawn.¹⁰ His conductor was a forty-two-year-old veteran of twenty-two years’ service who knew his face from a previous posting but not his name.¹¹

The early legs went normally. The rapid service picked up a minute or so of delay in the rush-hour crush, as it always did. Nothing on the recorders suggests anything wrong with the man in the cab until the empty run to Takarazuka.

That run had more than ten minutes of slack built into it, and he used the slack strangely. Between Kawanishi-Ikeda and Nakayamadera the train dawdled at times below 10 km/h on a stretch with no level crossings, as if he were killing time where it would inconvenience no one.¹² Then, approaching Takarazuka, he accelerated hard, at full power or close to it, towards a platform he was required to enter at 40 km/h across a diverging switch. The train hit the switch at about 65. A track beacon guarding the red signal ahead sounded the safety system’s bell in the cab. He had five seconds to acknowledge it, and the seconds ran out. The automatic train stop applied the emergency brake and the train halted straddling a level crossing, its nose just onto the end of the platform.¹³

The rules for what happens next are explicit. Before resetting the system, a driver must contact the dispatcher and receive instructions. He skipped that step, reset the system, called the conductor over the intercom to ask for a departure signal, and pulled forward. The system stopped him a second time, on a separate safeguard, just as the train reached its proper stopping mark. He arrived at Takarazuka at 8:56:14, forty-four seconds late, having triggered two emergency brake applications in ninety seconds and reported neither.¹⁴

Then he sat in the cab. The conductor, walking the length of the platform to swap ends, took about two minutes twenty to reach him, which is normal. Drivers usually take about a minute to gather their things and leave. This one stayed nearly three, and when the conductor arrived he was still in the seat. He stood up, though the conductor could not say whether it was because he had been noticed. As they passed at the cab door the conductor asked, ‘Did the P stop you?’, mistaking which safety system was in use. The driver gave him a look the conductor later described as ‘I don’t need to tell you,’ or simply sullen, and walked past without a word. At 8:57, while he sat in the cab, a text message had arrived on his personal phone. Whether he read it, the investigators could not determine.¹⁵

He took the rapid service out of Takarazuka at 9:04, fifteen seconds down. The stops at Nakayamadera and Kawanishi-Ikeda came and went, the delay creeping to thirty-five seconds as passengers pressed aboard. Through Kita-Itami without stopping, at about 120. Then he shut off power and let the train coast at 122 km/h for forty-two seconds.¹⁶

Itami was a stop. The train carried a device that knew this. Six hundred and forty-three metres out, at 113 km/h, a recorded woman’s voice said in the cab: ‘Stopping. Stopping.’ He kept coasting. Four hundred and sixty-eight metres out, at 112, a man’s voice and a mechanical alarm cut in together, ‘Stop, stop’, and only then did his hand go to the brake, and hard. The train entered the platform at 83 km/h. It passed the stopping mark at 46. The conductor, leaning from the rear cab with his hand on the emergency switch, saw the speed and pulled it. The train stopped seventy-two metres past where it should have.¹⁷

It reversed, briskly by several passengers’ account, and the doors opened three metres short of the mark. It left Itami at 9:16:10, one minute twenty seconds late.¹⁸ An overrun of that length was, under the company’s own classification, a reportable incident. In the company’s culture, the rest of this account will show, it was also something else: the trigger for a form of discipline drivers dreaded more than any timetable.

III.

The company called it nikkin kyōiku, ‘day-shift education.’ A driver who overran a platform, missed a roll call or triggered a safety system was taken off trains, reclassified from crew to day-duty, and set to writing. Incident reports, letters of apology, ‘why-why sheets’ of root-cause analysis, essays on assigned themes. There was no rule governing how long it lasted or what it contained. That was left to the depot head, who could keep a man at a desk for as long as he judged necessary.¹⁹ Across the Osaka branch in the two years before the accident it was imposed 321 times. The median stay was three days. The longest was forty-four.²⁰

The driver had been through it three times. Four days in 2002, as a conductor, when his train ran through a station it should have stopped at. One day in 2003 for dozing on his feet while supervising a trainee. And thirteen days in June 2004, three weeks into his career as a driver, after he overran the platform at Shimokoma on the Katamachi Line by a hundred metres.²¹ The Shimokoma incident has the same shape as Itami. He was distracted by something else, the company’s record says. The second warning voice sounded. He braked hard, then pulled the emergency, and the tail of the train finished past the end of the platform. He had to call the dispatcher before reversing. The train arrived eight minutes late.²²

For that he was formally reprimanded, lost 50,000 yen from his bonus, and spent thirteen days writing. His essay themes are preserved in the report’s appendix, and the committee’s summary of them is dry: mostly matters of attitude and spirit, little about how a driver distributes his attention, nothing at all about braking.²³ A colleague who was in the room for part of it described the process. The driver had come back to the depot half-smiling and had not gone straight to apologise to the instructors, which enraged his superiors. He was deeply low. He wrote his assigned reports and the supervising staff refused to accept them, so he erased half and rewrote, and rewrote again, until the instructors began suggesting the phrases they wanted. ‘Even if you didn’t believe it yourself,’ the colleague said, ‘you wrote it, because otherwise you thought they wouldn’t let you drive.’ The desk where he worked was in full view of the drivers coming through for roll call, who were handed a notice about his overrun and asked by the supervisor, ‘What would you have done?’²⁴

His friends heard a version of the same thing. ‘Day-shift education is just copying out something like the company creed,’ he told one. ‘I don’t see the point.’ The pay cut hurt. ‘I really hate it.’ To a woman he had known since high school he said that having to ask permission to go to the toilet was what he hated. She thought it was the lowest she had ever seen him.²⁵ He also told her about a Kyobashi driver who had reversed a train in error a month earlier and been moved out of the cab into a vehicle-management job.²⁶ After Shimokoma, he said, another mistake and they would take him off driving.²⁷

Other drivers gave the committee their own accounts. One, taken off for a month, had been made to stand on the Kyobashi platform for an hour and a half each morning and three hours each afternoon, greeting every arriving driver: ‘Good work. Please take care.’ Leaving his seat for the toilet required a supervisor’s permission. What he remembered most was being watched.²⁸ Another, disciplined for passing a red shunting signal, was told on his return that he was ‘unfit for duty’ and might never drive again. He asked to leave the cab and was reassigned five days before the accident.²⁹ The head of the Kyobashi depot defended the system in terms that had little to do with railways. Families no longer scolded their children, he said, and teachers had become salarymen, so the job of instilling discipline had fallen to companies.³⁰

What the committee found more troubling than the humiliation was what the fear of it did to a man at the controls. A driver who had once run this same curve at 80 to 85 km/h, years earlier, explained that he had been composing an excuse for a mistake earlier in the run, a false report he had given the dispatcher without quite meaning to, and thinking about how to sustain it. He never reported the overspeed. ‘Under-reporting how far you overran,’ he said, ‘is an everyday thing in this company.’³¹ Another Kyobashi driver, stopped by the safety system on the Osaka Loop, weighed whether to report it and decided against, because reporting meant nikkin kyōiku.³² A conductor’s testimony confirms the arithmetic from the other side. If an overrun was around five metres, it was understood, the punishment was light.³³

IV.

The company had also built the clock he was running against.

In August 1988, a year into privatisation, a management meeting on Osaka-area services set out a policy. Among the measures listed for raising speeds was a phrase the committee reproduces without comment: ‘the total abolition of margin time.’³⁴ In the national-railway era, timetables had been calculated on the assumption that drivers would run 2 km/h under every limit, which built a small cushion into the running time itself. That practice survived at JR West until around the end of the 1980s.³⁵ In March 1991 the line speed between Amagasaki and Shin-Sanda rose from 100 to 120 km/h.³⁶ In December 1996, to feed the new Tozai Line into Amagasaki, the gentle 600-metre curve north of the station was rebuilt as a 304-metre one.³⁷

Then the running times came down. The standard time for a seven-car rapid service between Takarazuka and Amagasaki was cut by twenty seconds in March 2002, twenty more in March 2003, and ten in October 2004. Fifty seconds in all, over a fifteen-minute run.³⁸ The staff who calculated running times did not initiate these cuts. They were asked for them by the timetable planners, who were carrying out commercial policy: a 2003 board paper promised to bring the morning rapid service from Takarazuka to Osaka down from twenty-six minutes to twenty-three.³⁹

Dwell times went the same way. In December 2003 the planner in charge shortened the stop at Itami from twenty seconds to fifteen. He had gone to watch trains at Itami and timed the stops at seventeen or eighteen seconds without any late-running passengers. Fifteen would be achievable, he reasoned, if passengers could be encouraged to queue in an orderly fashion. He asked the relevant departments to arrange this. It never happened.⁴⁰ At Kawanishi-Ikeda, the timetable allowed twenty seconds and the trains were actually taking twenty-five.⁴¹ At Takarazuka, the planner had set ninety seconds between the departure of the limited express from platform 3 and the rapid service from platform 2, based on a single measurement he made in autumn 2002 and could not otherwise document. The company’s own measurements, obtained by the committee, told a different story. The signalling and level crossing needed about ninety-three seconds to clear, before anyone closed a door.⁴²

The result was a train that was scheduled to be late. Of sixty-five weekdays in the months before the accident for which records survived, 5418M reached Amagasaki more than a minute behind time on more than half. The median delay was eighty seconds.⁴³ When the committee ran the same route through the timetabling method used by a major private railway in the Kansai region, the standard running time came out ten seconds longer than JR West’s, before allowing for the dwell times that were too short.⁴⁴

The company, meanwhile, required drivers to report any delay of thirty seconds. A delay of a minute attributable to a crew member’s error was a reportable incident, and reportable incidents fed nikkin kyōiku.⁴⁵ The committee put the contradiction plainly. An operator that runs trains at 120 km/h without any device to catch a driver’s overspeed error, it wrote, should build corresponding margin into its timetable. This one had abolished it.⁴⁶ In a survey of fifty-one Kyobashi drivers after the accident, more than half said that the greatest psychological burden came from a schedule too tight to keep even with clear signals. Three in five said the delay that weighed on them most was between one and three minutes. Beyond three, several explained, you gave up trying to recover.⁴⁷

V.

The device that would have caught the error was already fitted to the train. Nothing in the track beneath it could yet speak to it.

Japanese railways run two broad kinds of automatic train protection. The older kind, called ATS-SW on this line, works from fixed points. As a train approaches a red signal, a beacon in the track sounds a bell in the cab. The driver must brake and press an acknowledgement button within five seconds or the emergency brake applies. Near the signal itself, a second beacon stops the train outright if it is still moving. It can check speed at a curve only where a beacon has been installed for that purpose, and none had been installed here.⁴⁸ The newer kind, ATS-P, is continuous. Beacons tell the onboard computer where the next restriction is and what it requires, and the computer calculates a braking curve and applies the brakes itself if the train’s speed rises above it. Fitted at a curve, it will slow a train that is approaching too fast whether or not the driver has noticed.⁴⁹

JR West had decided to install ATS-P on its busy Osaka lines in 1989 and had done so, line by line, through the 1990s. From 1998 its medium-term plans carried a line for the Fukuchiyama section: 200 million yen in fiscal 2003, 600 million in 2004, completion by March 2005.⁵⁰ The formal investment decision slipped to September 2003. The work was then handed from the head-office construction department to the Osaka branch, on the reasoning that it was a simple new installation the branch could handle. The paperwork delegating it to the signal depot took until March 2004. The design contract was let in April. In October the signal depot head concluded that the schedule could not be met and pushed commissioning to June 2005. He told the branch electrical manager, who told the branch head in December and two head-office department heads in January. Nobody objected. ‘I didn’t say keep to the deadline,’ the head of safety promotion told the committee, ‘because rushing the work might cause trouble.’⁵¹ On the morning of 25 April 2005, the beacons on the Fukuchiyama Line were still speaking only the older language.

Even a completed installation might not have covered this curve. JR West fitted the curve-speed function only where the radius was under 450 metres, and it regarded the function as an afterthought. ‘I understood that overspeed on a curve could cause a derailment,’ the head of safety promotion said, ‘but I didn’t recognise it as a concrete hazard. The curve-speed function was more a precaution.’⁵² The executive director in charge of the railway division said toppling on a curve was something he had never considered, and that he had been unaware of the past cases.⁵³ There were past cases. Twice, in 1988 and 1996, JR Freight trains had derailed on a 300-metre curve on the Hakodate Line in Hokkaido at roughly 100 and 117 km/h. The transport ministry had briefed JR West on both. The 1996 case appeared in the appendix to the company’s own safety committee papers in March 1997.⁵⁴

There was a smaller sign, too. In October 2004 the operations department drafted a revision of the printed list of prohibitions that every driver carried. The proposed new item read: ‘Do not operate a train in excess of the prescribed speed.’ The reason given was to prevent derailment by overspeed. After discussion the item was dropped. It was, a manager explained, too obvious to need stating, and the list was getting long.⁵⁵ The revised list went up on the Kyobashi depot noticeboard on 30 March 2005. A briefing on it was scheduled at the depot for 10:55 on the morning of 25 April. The driver was down to attend.⁵⁶

VI.

On the lines where ATS-P had already been installed, it had been generating data for years, and after the accident the committee pulled the memory from the onboard units of other trains in the 207-series fleet. On the underground Tozai Line, which the driver had run through that morning, there were six curves fitted with the speed-check function. Across roughly 4,900 passages, the system had applied maximum brake ninety times. Once in every fifty trains. With 162 trains a day, that meant the automatic brake was slamming on about three times a day, every day, on one stretch of line.⁵⁷

Almost all of it was unnecessary. The speed limit at each curve was made up of a base figure plus a small ‘designated speed’ allowance for modern stock, five kilometres an hour. Whoever had programmed the beacons had entered zero. On JR West’s network, sixty-eight of ninety-four curve beacons carried wrong data of one kind or another. The engineer who supervised the later installation on the Fukuchiyama Line, where twenty-two of thirty beacons carried the same error, said he had not known the designated speed was a field to be entered at all. The company said it learned of the errors only after the accident.⁵⁸ The committee’s conclusion was that JR West had been ‘in a position to easily know’ of the fault from equipment already fitted to its trains, and should not have waited for crews to report what its own recorders were logging.⁵⁹

The pattern repeated at smaller scale. In May 2004 the speedometers in two cars of the set that would derail were found at inspection to be reading wrong, one five kilometres high and the other five low, because a switch had been set to feed one car’s wheel diameter into the other’s calculation. The cars went back into service anyway. Drivers complained three times. The switch was corrected in August.⁶⁰ The speedometer in the lead car, the one the driver was watching, under-read by about 2.7 km/h at 120 because of a rounding shortcut in the manufacturer’s software. It did not comply with the applicable standard. The programmer told the committee he had not known the standard existed.⁶¹ Four times in 2004, Kyobashi drivers had reported that the brake handle on certain cars could lodge between the maximum service position and emergency, applying neither. No modification followed.⁶²

The committee set these two halves side by side. A company that demanded reports from its crews, punished those who made them, and punished more harshly those who did not, was at the same time failing to act on anomalies its own machines were recording and its own staff were flagging. A reporting regime built this way, it wrote, ‘may conversely induce accidents.’ Its first recommendation to the minister was for non-punitive incident reporting across the industry.⁶³

VII.

The curve itself was ordinary. A right-hand bend of 304 metres radius, with sixty metres of transition before it, a cant of 97 millimetres to lean the track into the turn, and a lineside sign reading 70.⁶⁴ The question of what speed would tip a 207-series car over on it had a precise answer, and the company got it wrong on the day of the accident.

That afternoon JR West published a figure of 133 km/h, calculated for an empty car and ignoring the effect of the suspension.⁶⁵ The committee reworked it using a standard formula for rail-vehicle overturning, with ninety-three passengers aboard, the springs’ contribution to the effective height of the centre of mass, and the ordinary lateral sway of a train in motion. The answer was 106 km/h. At full standing capacity it was 104.⁶⁶ A separate computer simulation, built from the actual car and track, found that the lead car stayed on the rails at a steady 105 km/h and derailed at 110 and at 115.⁶⁷ The train entered the curve at about 116.

Two details from the simulation give the sequence its texture. Standing passengers thrown outward by the curve may have shifted their combined centre of mass by about twenty-one centimetres in the first 1.9 seconds, and the committee built that into its model.⁶⁸ And the car began to lean early enough that the skirt below its front coupling scraped the ballast some nine to twelve metres before it reached a lineside pole, throwing crushed stone onto the rail head. The white powder found there afterwards was ballast, ground fine under the wheels.⁶⁹

The drivers at Kyobashi had no idea where the edge was. Asked after the accident what speed they had believed would topple a train on that curve, half of fifty put it at 120 km/h or above. Many added that no driver, however late, would knowingly run a curve tens of kilometres over its limit.⁷⁰ The committee agreed, and gave five reasons for ruling out deliberate speeding. Among them: a man consciously taking a 70 curve at 110 would have gone to emergency brake the instant he felt the need, not eased through the service notches one by one.⁷¹

VIII.

At 9:16:10 the rapid service left Itami and accelerated hard. At Kita-Itami it had been at 120. Now it went past that, to 124 or 125, while the speedometer in front of the driver read 121 or 122. At 9:18:09 he touched the brake for two seconds to bring it back under the line limit, then shut off power and let the train coast. It passed through Tsukaguchi at 9:18:22 at about 122 km/h.⁷²

Behind him, in the rear cab, the conductor had made the next-stop announcement: Amagasaki, change for Kobe and Osaka. The intercom buzzed immediately. The driver said something the conductor rendered as ‘Can you knock it down for me?’ The conductor understood him to mean the overrun distance. He thought for a moment and said, ‘You went a long way.’ At that point a passenger knocked on the glass between the saloon and the cab and demanded to know why there had been no apology for the delay. The conductor put the handset down, opened the door, promised an announcement, and made one.⁷³

Then he picked up the radio. At about 9:17:57 he called the dispatcher and reported that the train had overrun the rear limit at Itami ‘by approximately eight metres,’ had reversed with the driver’s agreement, and was running a minute and a half late. He had believed the distance was thirty or forty metres. He chose eight, he told the committee, because five metres was known to carry a light punishment but seemed too small to be credible. He did it, he said, because he knew the driver’s face and thought it would help him. He had not considered that the lie could be found out.⁷⁴ The dispatcher read the figure back, ‘rear limit, eight metres over,’ and asked the driver to respond. Somewhere between that readback and five seconds after it, the train passed the point where braking for the curve should have begun.⁷⁵

The dispatcher called the driver twice. There was no answer.⁷⁶

At 9:18:50, having coasted for thirty-nine seconds after his last touch of the brake, the driver braked. The recorder shows the handle moving through the positions in sequence: the lightest notch, then the middle ones, then the seventh, held for just over two seconds. As the lead car passed pole 41, at about 105 km/h and already leaning, he moved to the maximum service notch. He never reached emergency. At 9:18:54 the car went over.⁷⁷ By the committee’s calculation, the brake application was about sixteen seconds late.⁷⁸

What was he doing in those sixteen seconds? The report is careful here. The driver, it says, may have believed the conductor had answered him coldly and hung up. He was probably listening to the exchange between the conductor and the dispatcher with particular attention. He was likely composing an explanation for his overrun that would be consistent with ‘eight metres.’ He may have been trying to write it down: a special red pencil issued for noting radio instructions was found loose on the cab floor, and his right glove was off, which nearly a quarter of his colleagues said they did when jotting something on the timetable cover.⁷⁹ Nothing in his health, his roster, his sleep or his blood offered a competing explanation.⁸⁰

The committee’s formal finding of cause is three sentences long, and the graded vocabulary matters. The accident ‘is estimated’ to have resulted from the driver’s delayed braking, which took the train into the curve at approximately 116 km/h. The delay ‘is considered’ to have come from his attention leaving the driving, because, believing the intercom call in which he had asked for a false report had been cut off, he was attending to the conductor’s exchange with the dispatcher and thinking up excuses out of concern that he would be sent to nikkin kyōiku. And the company’s method of managing its drivers, which imposed nikkin kyōiku or discipline, taken by some drivers as a penalty, on those who caused incidents, and harsher versions on those who failed to report them or reported them falsely, ‘may possibly have been involved’ in both the false report he sought and the lapse that followed.⁸¹

IX.

The first emergency call was logged at 9:22. Fire crews were on scene at 9:26. Before they arrived, workers from the factories and businesses along the line had already begun pulling people from the wreckage, and they kept at it through the morning. More than thirty companies took part.⁸² What the railway’s own control centre did in that hour is recorded too. A power-district employee reached the site at 9:40 and at 9:59 radioed that members of the public were climbing on the carriage roofs beneath live overhead wires. The order to cut power went out at 10:03, forty-five minutes after the derailment, and the high-voltage distribution line was still live until 10:13, because the power controller had first cleared it with the train dispatchers so as not to strand trains between stations. The committee found in this a lack of regard for human life and told the company to put it in a manual: after a derailment, cut the power, then ask.⁸³

The last survivor was brought out at 7:06 the following morning, twenty-two hours after the crash.⁸⁴ The dead were almost all in the first three cars: forty-two in the first, fifty-seven in the second, three in the third.⁸⁵ In the second car, where the body shell had been crushed into a diamond and the interior space almost eliminated, eleven of the fifty-seven dead had suffocated under the weight of other passengers. Across the whole train, nineteen died that way.⁸⁶ The committee’s survival analysis found that people who kept hold of a strap or rail until the car stopped were injured less severely than those whose grip broke early, and recommended that handholds be designed for gripping hard and that the ends of bench seats be shaped to hold a body rather than pierce it.⁸⁷

The changes that followed were the ones the report had implied. ATS-P went live on the Fukuchiyama Line on 19 June 2005, when the section reopened. The curve limit fell to 60 and the line limit to 95, and forty-five seconds went back into the timetable.⁸⁸ By the end of fiscal 2006, JR West had fitted curve-speed protection at 1,370 sites, up from 105, and the ministry had made such protection compulsory across the industry, along with onboard data recorders.⁸⁹ Drivers were told, as a rule, to make radio contact only once stopped at the next station. Protective radios were rewired to switch to battery automatically. A safety research institute was founded. The Railway Business Act was amended to require every operator to name a safety manager and file a safety management plan.⁹⁰ The committee did not ask the company to abolish nikkin kyōiku. It asked only that it be remade into something that taught drivers to drive.⁹¹

The report closes, as such documents do, with recommendations, appendices and a glossary. It says in its preface that it was written to find causes and prevent recurrence, not to apportion blame, and it holds to that. It never names the driver. It records that he died in the cab of internal bleeding, that his left glove was on and his right was off, and that his shoulder bag held a banana and a bottle of tea.⁹² It records his declaration at the previous day’s sign-on, at 12:16 on the Sunday afternoon: ‘Reporting for holiday roster 75. Mental and physical condition good.’ At 6:08 the next morning the supervisor noted only that he seemed his usual self.⁹³

And it records one more thing. Six days before, on the Tuesday, he had sat down for his routine interview with the depot. Asked what work he hoped to be doing in a year or two, he said he would like to become an instructor. The reason he gave was written into the file. There would be more and more juniors coming through, he said, and he wanted to use his own experience so that none of them made the same mistakes he had.⁹⁴

Notes

RA2007-3-1 §2.2.10.1, passenger 1A. ↩
§2.2.10.2, passenger 1B. ↩
§2.2.10.3, passenger 1C. ↩
§2.2.13, eyewitness. ↩
§1.1, §2.4.2.1–2.4.2.2. ↩
§2.3.1, Table 1 (Hyogo Prefectural Police figures). ↩
The comparison with earlier accidents is not drawn from the report. The previous deadliest was the Tsurumi collision of 9 November 1963, 161 dead. ↩
§2.2.7, §4. ↩
§2.5.1, §2.5.12.1. ↩
§2.5.4.1. ↩
§2.1.1. ↩
§2.6, §2.2.8(1). ↩
§2.2.4, §2.14.9. ↩
§2.2.4, §2.10.1.8. ↩
§2.2.4, §2.13.7.2. ↩
§2.2.5, §2.2.8(5), §2.5.4.2, §2.21.8. ↩
§2.2.6. ↩
§2.2.6, §2.13.3.3. ↩
§2.2.6. ↩
§2.7.4.1. ↩
§2.7.4.3, Table 10. ↩
§2.5.7.4, §2.5.11.2, §2.5.11.5, §2.5.11.6. ↩
§2.5.11.2. ↩
§2.5.7.4, §2.5.10, §2.7.5.3(1). ↩
§2.5.7.4, driver F. ↩
§2.5.12.2, §2.5.12.3. ↩
§2.5.12.2, §2.20.1.12. ↩
§2.5.12.2. ↩
§2.7.4.9, driver I. ↩
§2.7.4.10, driver J. ↩
§2.7.4.7. ↩
§2.20.1.9. ↩
§2.20.1.5, driver K. ↩
§2.2.8(6). ↩
§2.14.2. ↩
§2.14.5.3. ↩
§2.10.1.2. ↩
§2.8.1. ↩
§2.14.3.1, Table 32. ↩
§2.14.3.3–2.14.3.4, §3.1.1. ↩
§2.14.4. ↩
§3.1.2.2. ↩
§2.14.7, §3.1.2.1. ↩
§2.14.6.3, §3.1.2.3. ↩
§2.14.5.4, Table 34. ↩
§2.10.10, §2.19.1, §3.1.2.4. ↩
§3.1.2.4. ↩
§2.21.10.1–2.21.10.2. ↩
§2.13.2, Table 28. ↩
§2.13.2. ↩
§2.13.8.3, §2.13.8.5. ↩
§2.13.8.5–2.13.8.6, §3.10.2. ↩
§2.13.8.4. ↩
§2.19.3.1. ↩
§2.20.2.1. ↩
§2.7.3.4, Table 9. ↩
§2.7.3.4. ↩
§2.20.1.1, Table 37, §3.13.1. ↩
§2.13.8.9. ↩
§3.13.1. ↩
§2.9.4.4. ↩
§2.9.4.1, §3.3.2. ↩
§2.5.11.3, §3.13.2. ↩
§3.13.1, §3.14.2, §5(1). ↩
§2.8.1, §2.8.2.1, §2.10.1.9. ↩
§2.21.6, Table 43 and note 2. ↩
§2.21.6, Table 43. ↩
§2.21.5.2, Table 42. ↩
§2.21.4.2, §2.21.5.1, §3.9.1 (the 21.3 cm weighted figure, hedged 可能性が考えられる). ↩
§2.21.5.2(3), §3.9.2.1. ↩
§2.21.10.10. ↩
§3.8.6.2. ↩
§2.2.7, §2.9.4.5, §3.8.6.1. ↩
§2.2.8(6), §2.2.10.15, §3.7.1. ↩
§2.2.8(6), §2.8.7.2(3). ↩
§2.2.7, §2.8.7.2(3). ↩
§2.8.7.2(3), §2.2.9. ↩
§2.2.7. ↩
§3.8.6.2. ↩
§3.8.6.2, §2.5.2.1, §2.21.10.9. ↩
§3.4, §3.5, §3.6.2. ↩
§4. ↩
§2.17.4.1–2.17.4.2. ↩
§2.17.2.2, §3.11.4, §6.1(3). ↩
§2.17.4.3(1). ↩
§2.3.2, §2.3.3.3. ↩
§2.3.3.3(2), §2.3.3.2(2), §3.11.1.1. ↩
§2.3.4.4, §3.11.1.3, §3.11.2.2. ↩
§7.1(7). ↩
§7.1(2), §7.2(3), §7.2(6). ↩
§7.1(3)–(4), §7.1(8), §7.2(7). ↩
§6.1(1). ↩
§2.5.2.1–2.5.2.2. ↩
§2.5.4.1 (24 April 12:16 出勤点呼; 25 April 6:08 発点呼). ↩
§2.5.4.1 (19 April 2005 interview record). ↩

Source: Aircraft and Railway Accidents Investigation Commission (航空・鉄道事故調査委員会), Railway Accident Investigation Report RA2007-3-1, West Japan Railway Company Fukuchiyama Line, Tsukaguchi–Amagasaki train derailment, adopted 22 June 2007, published 28 June 2007. All section references are to the Japanese original. Translations are the author’s.

The Station That Knew: On Governance That Cannot Read Signals
Reading the Room Before It’s a Finding
Organisational Gaps: the governance series index

Photo credits: https://commons.wikimedia.org/wiki/File:Fukuchiyama_joko20051.jpg (捕澤 - AI upscaled) and https://commons.wikimedia.org/wiki/File:Amagasaki_derailment_damage_to_the_first_car.jpg (Japanese Government).

The Value That Wouldn't Travel: Burberry, Pepsi and the Asset That Stays Behind

noreply@blogger.com (Unknown) — Sat, 26 Sep 2026 05:52:41 +0000

In 2015, one of Japan's most successful clothing companies discovered that the source of its success had never belonged to it. For roughly half a century, Sanyo Shokai had held the Japanese licence for Burberry. It had done more than import a British name. It had built two sub-brands, Burberry Blue Label and Black Label, that existed nowhere else, tuned precisely to Japanese sizing, taste, quality expectations and retail rhythm. At its height the arrangement generated revenue of around ¥111 billion (~US$1.03 billion) a year. Then Burberry jettisoned Sanyo Shokai, and took Japan under its own control.

A permanent downsizing

Sanyo Shokai never came back. Company revenue fell to roughly ¥68 billion within two years, a drop of nearly forty per cent, and a decade later it sits near ¥58 billion: a little over half of what the business earned when it still held the licence. Japanese apparel had a hard decade, and part of that decline was the market's. But a broad sector slump does not, on its own, halve a company.

Sanyo Shokai revenue, FY2011–FY2026.

The obvious but shallow reading is that Sanyo Shokai was punished for depending on someone else's brand. Value and transferability are not the same property: a thing can be immensely valuable to a business and still, when the relationship ends, walk out of the door with the other party. Economists have a name for this: an appropriable quasi-rent, value a business generates but which the other side of the relationship can capture when the terms are next set. The more useful question is why the value proved so impossible to keep in this particular category, because there are others in which the same divorce allows the value to be retained.

The mirror case

Consider a mirror case from a different industry and a different country. For sixty years, from 1952, a Thai company called Sermsuk bottled and distributed Pepsi across Thailand. By the early 2010s Pepsi held roughly 48 per cent of the Thai carbonated-soft-drink market, ahead of Coca-Cola. Then the relationship ended. PepsiCo had attempted to take control of its bottler and failed. The partnership dissolved, and in November 2012 Sermsuk launched a cola of its own, called est, into the distribution network it already owned.

The result was the inverse of Japan. The operator was not hollowed out, but the brand owner was. Pepsi's share collapsed to a low near 15 per cent, and Coca-Cola took the lead Pepsi had vacated. Pepsi did not disappear, and over the following decade it rebuilt, through a new joint venture and a distribution network reconstructed largely from scratch, clawing back to something in the region of 35 to 40 per cent. But it never regained the top spot, which Coca-Cola took and then held. Pepsi couldn't quickly replicate six decades of manufacturing and logistics capability. Even when it tried, it could only be rebuilt slowly, at great cost, and even then only to second place.

The name goes home with its owner. The network stays with whoever built it.

Where the value sits

A luxury brand is almost entirely story. Its value lives in the name, the heritage, the exclusivity. None of that is manufactured on the operator's premises. When the licence ended, Sanyo Shokai retained every capability except the Burberry brand: the one thing that mattered.

Compare to Sermsuk. Cola is close to a commodity, and brand only counts among what is actually on the shelf. A hot and thirsty customer having a spicy green curry at a roadside stall that stocks est does not go looking for a Coca-Cola. Availability decides whether you are in the choice at all, while brand decides who wins once you are. The name goes home with its owner. The network stays with whoever built it.

Dependency is not a failure of strategy. It is usually the strategy, and a sound one, right up until the arithmetic of the exit is run. The useful discipline is to know, in advance, which side of the ledger the transferable value sits on. Sanyo Shokai was, by all accounts, well run, but it is now a permanently smaller company because it ran, superbly, an asset that was always going to leave. Sermsuk played its hand well. But it had a hand to play because it held the part of the value that could not be taken back. The lesson is about the quiet difference between what an organisation earns and what it owns, and how rarely the two are counted separately until the day they come apart.

Elim Elytra Review: A Split Keyboard You Can Actually Carry

noreply@blogger.com (Unknown) — Sun, 20 Sep 2026 12:33:41 +0000

The Elim Elytra is a 63-key wireless split keyboard in CNC aluminium that weighs less than some tablets. It provides a fully customisable, row staggered layout meaning that people coming in from normal office keyboards will need little retraining. It is an excellent travel split keyboard given its low weight, though its padded case gives back some of the weight advantage, and mine shipped with firmware that could not hold a Bluetooth connection and required updating.

I paid for this unit as a Kickstarter backer. Elim has had no input into this review and did not see it before publication.

First published 20 Sep 2026 as a preliminary review after a month of use. Finalised 25 Oct 2026 after eight weeks, with Elim's answers to my questions and a battery result.

PriceUS$249

Our rating●●●●● 4.5 / 5.0

Pros

True wireless split at 420g
Four weeks of daily use on a single charge
Familiar row-staggered layout means low adaptation time
Hot-swappable low-profile switches with a cushioned feel
Despite its small size and low weight, it feels remarkably well built and solid

Cons

Shipped on firmware that could not pair over Bluetooth
Padded case erases most of the weight advantage in a bag
No Right Shift
No battery percentage, only a low-battery blink
Firmware updates are drag-and-drop, with recovery undocumented on Elim's site
Expensive, especially with accessories

Elim Elytra at a glance
Layout	63-key row-staggered split, split spacebar, no Right Shift
Connectivity	Bluetooth (3 devices), USB-C wired (left half only)
Switches	Kailh CloudShell White linear (40±10gf, 2.8mm) or JZF Mist silent, hot-swappable
Keycaps	Low-profile PBT, printed legends
Weight	420g measured (Elim quotes both 420g and 440g), 584g with case
Thickness	11.8mm excluding keycaps
Battery	About four weeks of daily use on one charge, matching Elim's claim
Software	Vial (browser, via USB)
Firmware	Shipped on 1.0. Current 1.4 (1 Sep 2026). Based on RMK
Price	US$249. Case US$39, tenting kit US$79–129, palm rest US$39
Rating	4.5 / 5.0

Layout

Many enthusiast split keyboards ask you to relearn typing. Column-staggered layouts, thumb clusters and forty-key minimalism are the norm, and the ergonomic benefit arrives with a fortnight of hunting for the semicolon. The Elytra keeps the row-staggered layout every laptop uses, trims it to 63 keys, and splits it down the middle. Anyone who has typed in the past century will find their fingers already know where to go.

I found a few things which may put you off-side initially. The spacebar is rather short, and my thumbs spent the first few sessions hitting the very edge of the bar. The second is the lack of a Right Shift button. I'm a left shift user, so didn't notice, but for those who rely on the right shift this may be a problem. Elim has published a Vial workaround that puts Shift on a held Up Arrow, which is less convenient and worth knowing if you use the right shift.

Typing experience

I was at 85 WPM with 98% accuracy within an hour, against the 85 to 95 I manage on my daily drivers, the Keychron K11 Pro/Max and the Microsoft Sculpt. That is the fastest I've adapted to any board so far.

This set came with the CloudShell White linear switches. They are well cushioned, and typing feels satisfying yet quiet and clickless. They are also firmer than either of the other low-profile linears in my line up, which surprised me, because on paper they shouldn't be.

Three low-profile switches compared
	NocFree Rime	Elytra CloudShell White	K11 Pro Gateron LP Brown
Type	Linear	Linear	Tactile
Force	40±5gf	40±10gf	55±15gf
Pre-travel	1.2mm	1.3mm	1.7mm
Total travel	3.2mm	2.3mm (Elim) / 2.8mm (product page)	3.0mm

These are three switches on keyboards previously reviewed. Two are 40-gram linears, but they feel nothing alike. The NocFree & fires if you rest a finger on it. The Elytra needs a bit of convicting to press. The Gateron Brown is nominally the heaviest of the three but for some reason it feels like the Goldilocks of switches: it doesn't offer too little or too much resistance, and once it gets past the bump it activates. The CloudShell actuates at 1.3mm and is factory lubed, so neither the trigger point nor a dry housing explains it. What remains is travel. Elim quoted me 2.3mm of total travel, and this figure is the shortest of the three. A shorter stroke means your finger reaches the aluminium floor sooner, and on a low-profile board the floor is most of what registers as firm. The Brown feels lighter than its 55gf because once past the bump it falls away, where the CloudShell resists evenly to a hard stop. Two months on I still notice it, but it has never been a problem. The key note for anyone shopping low-profile boards is that gram weight tells you little about how it'll feel in reality – travel and what sits under the switch matter more.

The keycaps have a stepped profile, like a large Lego brick. This has not affected the typing experience, but I do wonder how it will hold up when it comes to cleaning. Flat keycaps wipe down in one pass. These have more angles for dirt to accumulate, and a wipe-down may need more time to get into the valleys. The legends are printed on PBT, which is the combination most prone to wear. To their credit, the printing is sharp, but only long-term use will tell me how well they will survive daily use.

There is no backlighting, that's perfectly fine. In a brightly lit office, rainbow LEDs are fun to have and generate conversation, but they are not needed, and their absence means nothing extra is draining the battery.

Ergonomics

Unlike a unibody Alice board, where the angle is decided for you, a true split means you need to experiment to figure out what the right positioning is for your purposes, and it certainly took a while: I spent a few hours nudging the halves in different angles and placements before it felt right. I ended up positioning it like a split Alice board, which works for me because it means my wedge-shaped Logitech Vertical MX mouse has space to move on the right. I did try positioning them where the hands sit at shoulder width with the wrists straight, but then the mouse felt too far. As with many things, this is really based on your preference, and I found typing in both positions to be easy.

They keyboard does have magnets allowing you to snap both halves together to resemble an Alice board, but I found it to be uncomfortable. The execution is rather cool though.

The two halves talk to each other wirelessly, with no cable between them. In daily use, they have not lost their connection to each other even once.

Elim sells a CNC aluminium tenting kit (6° and 10°, US$79 for the keyboard, US$129 with palm rest) and a wooden palm rest (US$39). I bought neither, and so won't be testing them. The board as shipped sits flat, with no tilt or tent.

Portability

Naked, the Elytra weighs 420g on my scale. The K11 Pro it replaces weighs 510g, and Keychron's current Alice-layout replacement, the Keychron K15 Max, is 700g (1.5 lbs). On raw weight the Elytra wins by a wide margin, and at 11.8mm (0.46 in) it is thin.

Packed, it is a different story. The Elytra needs its case. I tried the two halves in the Keychron K11 Pro's unpadded sleeve and the sleeve wants to fold – two separate slabs will knock together in a bag, and a floppy sleeve cannot stop them. Elim's case fixes that. It is internally cushioned, secures both halves well, and takes the packed weight to 584g and the packed thickness to well beyond a slab in a sleeve. The K11 Pro in its sleeve, at ~550g (1.2 lbs), remains the (marginally) lighter keyboard in my backpack.

Note that the case was included in my backer tier and is a US$39 extra at retail. The halves are held in place by rubber straps, and I would expect those to loosen over a few years of travel, at which point the case stops doing the one job it exists for. I will report if and when that changes.

Build quality

The board feels solid. My initial concern was that the hollow shell at the bottom would cause it to bend and buckle under typing, but this has not been the case at all. Despite the low weight it stays put on the desk. The promise of a quiet typing experience is fulfilled, and it is closer to a standard office keyboard rather than a mechanical clack.

Each half has a hollowed aluminium back that Elim says is modelled on beetle wing cases, which is where the name comes from. They also shipped a solid insulation pad in the box as an alternative. I have left the board as it came, because that is what a buyer receives and I'm not keen on taking the board apart.

Battery

Each half charges separately over USB-C. The battery indicator is a red blink when low and a green breathe when charging. There is no percentage, in Vial or anywhere else.

Elim's FAQ said 'to be confirmed' during the campaign and now claims up to four weeks per charge. Mine went eight weeks on one charge with no red light. That exceeds the claim. Elim confirmed that each half carries a 300mAh cell, which is small (a Keychron K3 Pro carries five times as much) so the result is good power management rather than a big battery. One caveat: firmware 1.4, released on 1 Sep, claims a three-to-four-fold improvement in left-half battery life, and my test ran on 1.4. Readers on earlier firmware should expect less.

Software

Remapping. Vial runs in Chrome or Edge at vial.rocks over USB (to the left half), and every key is remappable with layers, combos and macros. The software feels intuitive. I added Page Up, Page Down, Home and End, which the 63-key layout omits, and assigned the three Bluetooth profile keys, which the default layout also leaves out. It took a few minutes and nothing surprised me. A short test with an iPad paired and typed without issue.

Firmware. My unit arrived in May on firmware 1.0. On 1.0 it would not reliably pair over Bluetooth, and when it did, it connected and disconnected every second. It was unusable wirelessly. The wired fallback works, but only through the left half – the right half's USB-C port is charge-only, which nothing on the box tells you.

Elim's release notes follows the update history. Firmware 1.1 (21 May) was 'general stability'. 1.2 (24 Jun) was 'improve the Bluetooth compatibility', which is the manufacturer's own description of the fault I hit. 1.3 (14 Jul) fixed Bluetooth dropping on charge-only power. 1.4 (1 Sep) fixed lag after sleep and extended battery life. Four releases in fifteen weeks. Since updating to 1.4 I have paired the board with three machines and it works perfectly so far.

The firmware is built on RMK, an open-source keyboard firmware written in Rust whose principal author, Haobo Gu, is Elim's founder. Vial support comes from RMK implementing the Vial protocol, so the Elytra is in effect the reference board for the project.

Before you pair it: if your Elytra arrived on firmware 1.0 to 1.3 and Bluetooth is unstable, update to 1.4 first. The board will work wired through the left half while you do.

Firmware updating experience: The update process is for the intermediate user. There is no updater app. You clear the board from your computer's Bluetooth list, connect a half by USB, double-press the reset button on its underside, and a drive appears. You drag the correct .uf2 file onto it and the half reboots. Then you do the other half with its own file. macOS will complain that the disk was not ejected properly, which Elim says is normal. Each half has its own file. If you flash the wrong one, Elim told me the fix is to double-press reset again and flash the correct file. I have not tested it, but the bootloader is always reachable, which is a long way from the NocFree &, where recovering from a failed update meant opening the case and shorting the board.

Verdict

I have three travel splits. The SpiderIsland is the cheapest, fully wired and thick. The NocFree & is a 75% board that splits, weighs 1.03kg and comes with a case the size of a suitcase. The Elytra is the lightest and thinnest compared to its competitors. Accessorised with case, the Elytra runs to US$288 before shipping.

It has replaced my (discontinued) Keychron K11 Pro/Max keyboard as my choice of daily carry. For anyone who needs the halves apart and wants to carry them, the Elytra is now the board I recommend. It loses half a point for shipping on firmware that could not pair, and for a layout that quietly drops Right Shift. But the typing experience has been excellent, and it just works.

Frequently asked questions

Is the Elim Elytra hard to learn?

No. It keeps the standard row-staggered layout, so there is nothing to relearn. The only adjustments are the split spacebar, which took me about five minutes, and the missing Right Shift, which matters only if you use it.

Does the Elytra have a Right Shift key?

No. The 63-key layout omits it. Elim publishes a Vial workaround that puts Shift on a held Up Arrow, but a Right Shift typist should know this before buying.

Is the Elytra truly wireless between the two halves?

Yes. There is no cable between the halves. In seven weeks of daily use they have not lost each other once. For wired use you plug in the left half only, and the right half links to it wirelessly.

How long does the Elytra battery last?

About 4+ weeks of daily office use on a single charge in my testing, which matches Elim's claim. Each half carries a 300mAh cell. There is no percentage indicator, only a red blink when low.

Does the Elytra work with Mac and Windows?

Yes. It pairs with up to three devices over Bluetooth and supports Windows, macOS, Linux, iOS and Android. Remapping through Vial works on any of them in Chrome or Edge.

Can you remap the keys on the Elytra?

Yes, every key, through Vial at vial.rocks in a Chromium browser, connected by USB to the left half. Layers, combos and macros are supported.

Is the Elytra good for travel?

Yes, with a caveat. At 420g and 11.8mm it is the lightest split I have used, but the two halves need a padded case, which takes the packed weight to 584g. A single-piece board in a thin sleeve still packs flatter.

What firmware should the Elytra be on?

Version 1.4 or later. Units shipped on 1.0 had unstable Bluetooth. The update is a drag-and-drop .uf2 file onto each half in bootloader mode, and the board works wired through the left half while you do it. Flashing the wrong half is recoverable by re-entering bootloader mode and flashing the correct file.

Trust, but verify: AI has made writing cheap, but checking optional

noreply@blogger.com (Unknown) — Sun, 20 Sep 2026 08:50:34 +0000

A report that once took forty hours to write now takes fifteen. AI drafts it, and the drafts are good. Good enough to be dangerous. But what disappeared with those twenty-five hours was not just typing, but an even more critical control: checking.

When you write something yourself, you check it as you go. Not perfectly, and not always. But you cannot write a claim without at least noticing it, and the weak ones tend to fall apart in your hands before anyone else sees them. That checking was never a separate step. It came free with the writing. The machine does not do it, and the document it produces looks exactly like one that has been verified: polished, fluent, and plausible. That is verification theatre.

It has already cost someone real money. In October 2025 Deloitte refunded part of a AU$440k (US$314k) fee to the Australian government. A report it delivered cited academic papers that did not exist and misstated a Federal Court judgment. The firm's business is assurance. It had used a large language model, and nobody had caught what came back. If a firm that sells checking can publish unchecked work under its own name, the belief that better people would not is wishful.

The machine making mistakes is not the problem. Everyone makes mistakes. The problem is that a check which used to happen automatically now happens only if someone decides to do it, and people are bad at that. Automate a task and you do not free the human – the psychologist Lisanne Bainbridge showed this in 1983. You turn them into a supervisor, and supervising is a poor fit for human attention. Doing something keeps you focused. Watching something does not. Forty years of research on automation bias has confirmed it: people accept machine output even when it is wrong, and miss what the machine left out. Training helps with the first. It does not help with the second.

Even someone who wants to check has less to work with. When you wrote the thing yourself, you knew where every fact came from, because you went and got it. Now you are checking claims you never held in your head, and a fabrication that sounds right looks exactly like a fact that is. The reviewer has less to go on than the writer ever did, and it shows up in three ways:

A signature at the bottom cannot make up for nobody reading the middle.

Polish disarms the reader. Every organisation has one: the charming talker who understands very little, yet speaks with unassailable confidence, and fails upwards. Such individuals are proof that we already reward fluency over substance when the fluency is good enough. The machine has made that fluency free. A clean, confident document still gets less scrutiny, not more, because until recently it usually meant someone had done the work. That signal is gone. Old warning signs, such as the hedge, the seam where two arguments were stitched together, the number that does not sit right, went out with it because the machine is built to sound sure of itself.

Checking the machine with the machine. This one looks like diligence, which makes it worse. The reader asks the AI whether it is sure, and takes the answer as a check. The model may well catch its own mistake. It may not. Either way, it is checking its own work, and any auditor will tell you that is not a check at all. Independence is the whole point – the checker cannot be the thing being checked. An independent check has to land on something the machine did not produce: a source, a record, a fact that is true whether or not the model says so.

The reviewer becomes a postbox. Work arrives, goes through the machine, and is forwarded on, barely read. It looks like laziness. Mostly it is what a busy person does when handed a tool that produces plausible output on demand. Leaning harder on whoever signs at the end will not fix it. A signature at the bottom cannot make up for nobody reading the middle.

The usual answer to all this is a human in the loop. When researchers tested it, people relied on the machine's answer even when it was wrong. Explaining the answer did not help, and sometimes made things worse. The only thing that reliably made people think was friction: designs that made it harder to click through without engaging. People rated those designs the worst. Checking is dull work, and without friction, given the choice, they skip it. Putting a human beside the machine does not create a check. It decides who gets blamed when there wasn't one.

More review does not fix this. More review is more theatre. What does fix it is smaller and less comfortable: the check has to belong to someone. Either the person using the machine checks the output as they make it, or someone is named to do it afterwards – and belonging only bites if the checking is built to interrupt, not merely requested. The box below shows what that looks like. What cannot work is the assumption that a signature at the end covers it.

The machine did make the work faster. What nobody said is that it also changed the job. The person who used to write the report now reviews one they did not write, with none of the writer's knowledge of where it came from. The title on the door did not change. The paperwork did not change. Only the work did, and it got harder in the one place nobody was looking.

What actually made people check

The argument above is that checking will not happen by choice. A 2021 study tested what does make it happen. Researchers tried "cognitive forcing functions" – small design frictions that interrupt the reflex to accept a machine's answer. Three worked.

Decide first, see the answer second. The person commits to their own view before the machine's output appears, so they cannot simply anchor on it.

Make them ask. The answer is requested, not served – a deliberate act rather than a default.

Impose a pause. A short enforced wait before the recommendation appears breaks the habit of automatic acceptance.

All three cut reliance on the machine's answer more than showing explanations did. All three were rated worst by the people using them, who also thought their own work had got worse, even as their decisions measurably improved. The discomfort is not a side effect. It is the check working.

The Sydney Opera House: Looking Beyond the Classic Postcard

noreply@blogger.com (Unknown) — Sat, 12 Sep 2026 12:24:43 +0000

The Sydney Opera House is a beautiful building, but it is often seen from only the typical angles - from the side, when you're right in front of it, or when you're on a ferry passing by. This is a selection of some of my favourite, alternative views of the Opera House, photographed over a period of years when visiting the city.

Taken from Berry's Bay lookout. The Opera House peaks behind the Harbour Bridge.

105mm · f/8 · 60s · ISO 100

Just a normal street in Kirribilli, the slope and the apartment buildings perfectly frame the Opera House.

32mm · f/8 · 15s · ISO 200

Fireworks in Sydney Harbour are a regular feature, and often are accompanied with a spectacular backdrop.

24mm · f/7.1 · 1s · ISO 400

Using the burst zoom technique, we see double, turning what is normally a boring shot and angle into something a bit more interesting.

93mm · f/7.1 · 8s · ISO 200

Why the Logitech Wave Keys is not an Ergonomic Keyboard

noreply@blogger.com (Unknown) — Sun, 06 Sep 2026 14:29:09 +0000

Logitech knows how to make ergonomic keyboards. Their Ergo K860 is proof of that, offering a split layout with negative tenting, decent wrist rest, and a design to reduce strain, which makes the existence of the Logitech Wave Keys so confounding.

[message]

ℹ️The Best Portable Ergonomic Split Travel Keyboards: https://www.bulsuk.com/2025/01/best-split- portable-ergonomic-keyboard.html

ℹ️All Our Ergonomic Keyboard and Mice Reviews, Guides, and Expert Editorials: https://www.bulsuk.com/p/ all-our-ergonomic-keyboard-reviews.html

Logitech markets the Wave as being ergonomic, but the design simply doesn’t meet the criteria of what an ergonomic keyboard should be. A good ergonomic keyboard is supposed to place your hands and wrists in a natural position, preventing your wrists bending outward into ulnar deviation, lift upward into extension, and shoulders from collapsing toward the midline, which all can contribute to Repetitive Stress Injury (RSI).

In practice, this means keyboards with a split layout at the bare minimum, with tenting and negative tilt also being of benefit.

The Wave Keys doesn’t meaningfully address any of these issues. At its core, it’s still a conventional office keyboard with a cosmetic “wave” shape added on top. The curvature may look ergonomic, but it doesn’t create the separation or angles required to actually change your posture. Logitech calling the Wave ergonomic would be like Honda arguing that their Civic accelerates and handles like a Ferrari.

One of Logitech’s key marketing points is that the keyboard is instantly familiar to anyone who has used a keyboard in the past century. The implication is that real ergonomic keyboards are difficult to use, but that argument quickly falls apart. For starters, real ergonomic keyboards are supposed to correct your typing posture, which naturally requires some adaptation. Secondly, adaptation doesn’t mean abandoning the familiar QWERTY layout. Many split keyboards use familiar staggered layouts, and most users adjust within days, not months.

So who is the Wave actually for? Realistically, it’s more for those who have very mild to no ergonomic discomfort, and those who want to feel like they’re doing something to improve their posture. It’s a slight upgrade, but doesn’t help to correct posture.

But for others who have wrist pain, forearm strain, or carpal tunnel issues, the Wave is the wrong choice. It won’t correct posture, reduce deviation, or meaningfully change the problems that cause repetitive strain injuries.

Ultimately, a keyboard cannot claim to be ergonomic unless it provides at least a split design at the bare minimum. The Logitech Wave Keys does not meet that threshold. For some, the Wave may be a pleasant keyboard to type on, but it is not an ergonomic keyboard.

If you’re looking for some good ergonomic split keyboards, have a look at some of our reviews:

[message]

✈️The Best Portable Ergonomic Split Travel Keyboards: https://www.bulsuk.com/2025/01/best-split- portable-ergonomic-keyboard.html

ℹ️All Our Ergonomic Keyboard and Mice Reviews, Guides, and Expert Editorials: https://www.bulsuk.com/p/ all-our-ergonomic-keyboard-reviews.html

The Logitech Ergo K860: an example of a actual ergonomic keyboard:

The 7 Reasons to Work for a Japanese Company in 2026

noreply@blogger.com (Unknown) — Fri, 04 Sep 2026 04:46:51 +0000

Japanese companies in 2026 are not the ones people warned you about in 2006. The biggest wage increases in three decades, a legislative overhaul of working hours, and a structural shift toward merit-based hiring have quietly rewritten the proposition, while the things that were always good about Japanese employers (stability, training depth, genuine loyalty to their people) haven't gone anywhere. If you're weighing up a Japanese company against a Western alternative, here's what the balance sheet actually looks like now.

1. Job security that still means something

Lifetime employment in its purest form, where you join at 22, retire at 60, and never once update your CV, is rarer than it was. But the underlying philosophy hasn't evaporated. Japanese companies still default to protecting headcount during downturns. Where a US firm reaches for layoffs in the second quarter of a bad year, a Japanese firm cuts bonuses, freezes hiring, redeploys people internally, and waits. Toshiba shed 7,000 roles between 2015-2016 and it made national news precisely because it was unusual.

The practical effect: you are far less likely to be made redundant on 30 days' notice because a new CEO wants to "right-size" the organisation. That security compounds over years. It lets you take longer-term career bets, build genuine expertise, and plan your life without the background hum of "will I survive the next restructure."

2. They will train you properly

Not a two-day onboarding and a login to the online training portal. Japanese companies still invest in structured, months-long training programmes for new hires, and rotational development that continues well into your thirties. The logic is straightforward: if you're keeping someone for 20 years, the return on a six-month training investment is enormous.

In practice, this means you'll rotate across functions on a cadence that has no real equivalent at a Western firm. The jinji idō system, or periodic reassignment across departments, sometimes across cities, is a normal feature of career development at many Japanese companies, not an exception reserved for high-potentials. The logic is producing generalists: a manager who has worked production, logistics, and sales is assumed to make better decisions than one who hasn't. That breadth, accumulated early and across the whole value chain, is genuinely hard to replicate anywhere else.

3. Wages have finally started moving

For decades, Japanese pay was the trade-off you accepted in exchange for stability and benefits. That equation has shifted. The 2024 and 2025 Shunto wage negotiations delivered the largest base-pay increases since the early 1990s – 5.1% in 2024, with major firms like Toyota, Honda, and Nippon Steel going higher. After 30 years of stagnation, Japanese companies are competing on compensation again because they have to: the labour market is the tightest it's been in decades.

The significant caveat is the yen. A ¥7 million salary looked very different at 110 to the dollar than it does at 155. If you're earning in yen and benchmarking against USD or AUD salaries, the purchasing power gap is real. But if you're living in Japan and spending in yen, your domestic standard of living has genuinely improved.

4. Benefits that go beyond the payslip

The Japanese benefits model has always been generous in ways that don't show up in a total-comp spreadsheet. Housing allowances (often ¥30,000-80,000 per month), commuting costs covered in full, company-subsidised meals, comprehensive health coverage, and a pension contribution that's meaningful rather than token.

What's newer is the expansion into lifestyle benefits, such as subsidised childcare, fertility treatment support, and sabbatical programmes at firms like Recruit and Shiseido.

5. Work-style reform is real, if uneven

The 働き方改革 (hatarakikata kaikaku, or work-style reform) legislation that came into force in 2019 capped overtime at 45 hours per month and 360 hours per year, with criminal penalties for non-compliance. Five years in, the effect is measurable – average annual working hours have dropped, and "premium Friday" may have been a joke, but the legal backstop is not.

The honest picture: this varies enormously by company and industry. A Tier 1 manufacturer or a large trading house in 2026 looks genuinely different from 2015 – lights off at 8pm, remote work two days a week, managers actively pushing people to leave. A mid-size construction firm or a regional bank? Closer to the old model with better paperwork. The gap between progressive and traditional Japanese employers is now wider than the gap between "Japanese" and "Western" as categories, which means your due diligence on the specific company matters more than it used to.

6. Structure that rewards patience

The shift to ジョブ型 (job-gata, or job-type) hiring, in which you're recruited for a specific role rather than as a generalist who'll be rotated at the company's discretion, has made career paths more legible. At firms like Hitachi, Fujitsu, and MUFG, you can now see the path from role to role without needing to wait for an opaque internal committee to decide it's "your time."

But the deeper advantage of a Japanese career structure is harder to articulate: the system rewards people who build institutional knowledge over years. If you're the kind of person who wants to go deep rather than hop between firms every 18 months, a Japanese company will value that instinct rather than penalise it. The compounding effect of staying, whether in relationships, in understanding, or in trust, is something Japanese organisations still recognise and reward in ways that Western firms have largely forgotten.

7. A genuine sense of belonging

This is the one that's hardest to quantify and easiest to dismiss as soft. But after years inside the system, it's the thing people miss most when they leave. Japanese companies invest in making you feel part of something – not through ping-pong tables and branded hoodies, but through the slower, less photogenic work of shared meals, shared struggles, and the accumulated weight of having weathered things together.

The end-of-year bōnenkai where the division head pours your beer. The morning chōrei that starts every day with the same rhythm regardless of whether the quarter is good or bad. The quiet pride when a product ships and everyone in the chain, from the factory floor to the sales team, knows they were part of it. It's paternalistic, certainly. But for people who want to feel that their work lives are more than a series of transactions, it's hard to find elsewhere.

* * *

Working for a Japanese company in 2026 is a different proposition from a decade ago – better in some ways, still imperfect in others. The structural advantages (stability, training depth, belonging) remain intact while the historical weaknesses (pay, hours, rigidity) are being actively addressed, if unevenly. Whether it's right for you depends less on "Japanese vs. Western" as a binary and more on which specific company, in which specific industry, at which specific point in their reform journey. The best Japanese employers in 2026 are genuinely excellent places to build a career. The question is whether you've found one of them.

[message]

Read more: Not everything is awesome, take a look at the Challenges of Working at a Japanese Company before deciding if you want to make the jump.

The 7 Challenges of Working for a Japanese Company in 2026

noreply@blogger.com (Unknown) — Fri, 04 Sep 2026 04:43:21 +0000

The gap between the best and worst Japanese employers has never been wider. A decade ago, you could generalise: long hours, rigid hierarchy, low pay, face time culture. In 2026, that still describes a meaningful slice of corporate Japan, but it no longer describes all of it. Hitachi's merit-based pay system and Toyota's remote work policies exist in the same economy as regional banks where leaving before the boss is still career suicide. The challenge for anyone considering a Japanese company isn't "is Japanese corporate culture difficult?". Instead, it's "which Japan am I walking into?"

These are the things you need to ask about before you accept.

1. The yen problem (and what's left after tax)

Japanese wages have risen, with the strongest Shunto results in 30 years, but the yen has fallen faster. At ¥155 to the dollar in 2026, a ¥7 million salary that looked competitive at 110 now converts to roughly US$45,000. If you're benchmarking against offers in New York or Singapore, the gap is brutal on paper.

What catches people on their first Japanese payslip is the all-in deduction rate. Japan doesn't just take income tax – it layers national income tax, a flat 10% resident tax, health insurance, pension contributions, and employment insurance on top of each other. A mid-career professional on ¥7 million realistically loses 30-35% before they see a yen of it. And the rate scales aggressively from there. Clear ¥10 million and you're above 40%, a threshold that Singapore and Hong Kong never reach regardless of income. That's comparable to London, except you're starting from a lower gross number. The effective bite is large relative to what's being bitten.

And the domestic cost of living has caught up too. After three decades where prices barely moved, Japan experienced its first sustained inflation since the early 1990s. Food prices are up over 7% in 2026 compared to 2025, energy costs elevated, rents in Tokyo and Osaka rising for the first time in a generation. The cumulative effect: prices went up and haven't come back down, while wages are only now starting to catch up. The era of "Japan is cheap", where a modest salary bought a reasonable life because nothing ever got more expensive, is over.

This matters less if you've already built savings in a hard currency and are spending down in yen. But if you're earning in yen, saving in yen, and watching your grocery bill climb while your take-home inches up 2-3% annually after deductions, the arithmetic is tighter than it used to be.

2. Face time is dying unevenly

Work-style reform legislation (2019) capped overtime and introduced criminal penalties for non-compliance. COVID normalised remote work. The result: at progressive firms, such as Recruit, Sony, Mercari, the major trading houses, leaving at 6pm is unremarkable, and two days remote is standard.

At traditional firms, the legislation created better paperwork without changing the culture. People clock out at the mandated time and keep working. The boss is still last to leave, and being seen at your desk still signals commitment in ways that a well-crafted email from home does not. "Premium Friday" became a punchline precisely because the underlying expectation didn't shift.

The gap between these two worlds is now wider than the gap between "Japanese" and "Western" companies as categories. Your due diligence in the interview process, asking specifically about average departure times, remote work uptake, and overtime hours by department, matters more than any generalisation about Japan.

3. Nomikai still exists, but the pressure has shifted

After-work drinking sessions haven't disappeared. The end-of-year bōnenkai, the welcome party for new hires, the farewell for a departing colleague are still part of the rhythm. What's changed is the frequency and the social pressure. Pre-COVID, two to three sessions a week was normal in some departments. Post-COVID, once or twice a month is more common, and younger employees are more willing to say no without career consequences.

What hasn't changed: if you're senior, you're still expected to pay a disproportionate share. The sliding scale of junior staff pay least, senior staff pay most, remains intact. Corporate entertainment budgets are still rare. And the unspoken truth remains: relationships built over drinks still influence who gets staffed on the good projects, who gets the favourable rotation, and who gets mentioned in promotion discussions. The sessions are optional in theory, but the career consequences of permanent absence are real in practice.

4. Gender equality remains structural, not cosmetic

Japan ranked 118th out of 148 countries in the World Economic Forum's 2025 Global Gender Gap Report. That's below every other G7 nation, below most of Southeast Asia, and roughly where it's been for years. While there have been marginal improvements in score, there is no meaningful change in rank.

The picture inside companies: women's labour force participation has risen sharply (to ~74%, up from 63% a decade ago), but much of that increase is in non-regular employment – part-time, contract, limited-term. Career-track roles remain disproportionately male. Women in administrative and clerical support functions with limited promotion paths are still common, particularly at traditional manufacturers and financial institutions. The expectation that women leave or downshift upon having children has weakened in policy but persists in practice – maternity harassment (マタハラ) is a legally recognised concept precisely because it needed to be.

Board representation has improved under the Tokyo Stock Exchange's corporate governance code revisions, but from a very low base. Female executives at the senior managing director level and above remain rare outside of foreign-affiliated firms.

For men: paternity leave uptake has risen (from 6% in 2018 to ~30% in 2024), but the average duration taken is still measured in days, not months. Taking extended paternity leave remains a career risk at most traditional firms.

5. The glass ceiling has a nationality, and a language

Japanese companies have globalised their operations but not their leadership. The expat model where Japanese nationals are dispatched to run overseas subsidiaries, remains the default at most large firms. Country heads, regional directors, and senior functional leaders in overseas offices are overwhelmingly Japanese, often on three-to-five-year rotations before returning to Tokyo.

The language barrier is more structural than it appears. Rakuten mandated English as its corporate language in 2010. Mikitani's "Englishnization" made Harvard Business Review and generated global headlines. Fifteen years later, with nearly 30,000 employees operating predominantly Japan-facing businesses, the working language remains Japanese for the vast majority of staff. If even Rakuten, the posterchild that made English-first a core strategic bet, still operates in Japanese at the working level, that tells you something about how deep the barrier runs. Takeda, with a non-Japanese CEO, is perhaps the most credible exception, but this an exception rather than the rule.

The practical implication: if you're a non-Japanese employee in an overseas subsidiary, your ceiling may be visible. Country manager of the local operation is possible at some firms. Regional leadership or a seat at the global table is rare without fluent Japanese, years of relationship building with HQ, and a willingness to navigate a system where the decisions that matter are still made in a language you may not fully command.

6. Communication is what's not said

This isn't a new challenge, but it's the one that foreigners consistently underestimate. Japanese business communication operates on what's unsaid as much as what's spoken. Consensus is built in pre-meetings (根回し / nemawashi) before the formal meeting. Disagreement is expressed through silence, through hedging language ("it might be difficult..."), through what's conspicuously absent from a response. Direct confrontation, even when done politely, can damage relationships in ways that are invisible until a promotion cycle passes you over.

Reading the air (空気を読む / kūki wo yomu) is not a soft skill in Japan. It's a core competency. If you come from a culture where clarity means saying what you mean, you will need to recalibrate. The colleague who says "I'll think about it" may be saying no. The manager who doesn't respond to your proposal may be telling you to withdraw it. The meeting where everyone nods may have decided nothing, because the real decision happened in a conversation you weren't part of.

It's a different operating system which takes years to read fluently, and in the interim, you will miss signals that your Japanese colleagues see clearly.

7. Seniority is cracking, but hasn't broken

The shift to ジョブ型 (job-type) hiring at firms like Hitachi, Fujitsu, and MUFG represents a genuine structural change – roles defined by function and output rather than years of service. Merit-based pay is no longer radical: it's policy at dozens of large firms.

But policy and culture are different things. In practice, the informal weight of seniority, such as who speaks first in a meeting, whose opinion carries more weight, who is "ready" for promotion, still maps closely to age and tenure at many companies. A 32-year-old outperforming a 45-year-old may be recognised in their pay grade while the 45-year-old gets the title. The system has added a merit layer on top of the seniority layer without fully replacing it.

If you're ambitious and impatient, ask about promotion timelines in the interview. Not the policy. The actual average age of people at each level. If everyone in a director role is 48 and above, the system is telling you something that the recruitment brochure won't.

* * *

Working for a Japanese company in 2026 is a question of which Japanese company, not whether Japanese companies as a category are good or bad. The best have reformed meaningfully. The rest are reforming on paper while waiting to see if the culture catches up. Your job in the evaluation process is to figure out which one you're talking to, and whether the specific trade-offs of that specific firm are ones you can live with for the years you'll spend there.

[message]

Read more: Take a look at the other side of the coin, and the Reasons to Work for a Japanese Company.

Keyboard Encyclopedia: Key Things to Know

noreply@blogger.com (Unknown) — Thu, 03 Sep 2026 12:14:00 +0000

These are some keyboard terms and concepts that we use in our articles and reviews. We are gradually building this section and will continue to add more information.

Diagram and Key Definitions

As a baseline, the diagram below shows the anatomy of a keyboard, and groups the keys into sections.

Alphanumeric keys: Also called "alphas" for short, these are the core typing keys consisting of letters (A-Z), numbers (0-9), and punctuation.
Function Keys: Located at the top row (F1-F12), these keys perform specific tasks depending on the software.
Modifier Keys: These include Shift, Ctrl, Alt, and the Windows or Command key on Macs. They modify the behaviour of other keys when pressed together (e.g., Ctrl+C for copy).
Navigation Keys: Moves the cursor or scrolls through documents.
Number Pad: A dedicated grid of numbers and math symbols, usually on the right side.

* * *

Keyboard Layout Types

"Layout" gets used for two different things. This section is about the physical arrangement of the keys. It is also used for what the keys produce - QWERTY, Dvorak, Colemak - which is a software setting you can change on any keyboard without changing hardware.

Staggered keyboard

The vast majority of keyboards have keys which are arranged in offset rows, which originated from mechanical typewriters. Virtually everyone learns to type on these.

Ortholinear keyboard

These keyboards offer a grid-like layout with keys aligned in straight rows and columns. It definitely comes with a learning curve. Advocates claim that these are better ergonomically, but there is little to no peer-reviewed research to prove or disprove this yet.

* * *

Split Keyboard Variations

Split keyboards basically cut the keyboard in half, to allow for a more ergonomic posture when typing. There are two primary variations:

Unibody split keyboard (aka Alice keyboards):

Both halves of the keyboard are split at the centre, at around a 30-degree angle. Examples are the Incase/Microsoft Sculpt Ergonomic Keyboard, or the Keychron K11 Pro/Max. Some people call these "Alice" keyboards, although unibody split keyboards predate the term.

Dual body split keyboard

The two halves are completely separate, meaning you can place them at any angle, or distance. These usually need a cable connected to both halves to function, although the Nocfree & has removed the requirement, and made both halves completely wireless. Other examples include the SpiderIsland Split Keyboard.

* * *

Wrists and Posture

Most keyboards sold as "ergonomic" are describing their shape, not a proven health outcome. The research here is thin. What the words below do is describe what your hands are actually doing, which is enough to work out why one keyboard feels better than another after four hours.

Pronation: Rotating your forearm to turn your palm downward. A flat keyboard asks for full pronation and then holds you there all day. Shake your hands out and let them fall naturally and you will notice your palms face inward, not down. That gap is what tenting tries to close.
Ulnar deviation: Your wrists bending outward, away from your thumbs. It happens because a conventional keyboard is narrower than most people's shoulders, so the elbows come out and the hands angle in to reach straight rows of keys. This is the single thing split and Alice keyboards exist to fix.
Wrist extension: Your wrists bending upward, back toward your forearm. Thick keyboards cause it. So do the little flip-out feet on the underside, which tilt the back up and are almost always the wrong direction.
Negative tilt: The front of the keyboard sitting higher than the back, so the surface slopes gently away from you. It is the fix for wrist extension, and it is rarer than it should be. The Sculpt's built-in riser is the best known example.
Tenting: Raising the inner edges of the two halves so your hands sit at a shallow angle, like a low roof. This is the fix for pronation. Dual body boards let you set it yourself. Unibody boards decide it for you.
Row stagger and column stagger: Nearly every keyboard, Alice boards included, uses row stagger - the offset rows inherited from typewriters. Column stagger instead offsets each finger's column vertically, to match how far each finger actually reaches. It is a much harder adjustment, which is why it stays a niche.
Palm rest, not wrist rest: The pad should take the heel of your palm. If it is taking the underside of your wrist it is pressing on the soft part and holding you in extension, this is not good.

* * *

Keyboard sizes

Keyboard sizes refers to how many keys on the keyboard you keep - the fewer the keys, the smaller the keyboard. These are often referred to in percentages.

Layers

Smaller keyboards do not remove functions, they hide them. A layer is a second set of assignments for keys you already have, reached by holding a modifier (usually marked Fn) in the same way Shift gives you a second set of characters. On a 65% board the arrow keys are physical but the function row is not, so Fn plus the number row gives you F1 to F12.

The smaller the keyboard, the more it leans on layers, and the more you have to remember. That is the real trade in everything below, not the desk space.

Full size (100%)

Most of us would be familiar with the full sized keyboard, which is the baseline for all other keyboards on this list. It contains the alphanumeric keys, function row, arrow keys, a number pad, modifier keys (e.g., Shift, etc), and other keys such as page up, home, and so on.

Tenkeyless (TKL, 80%)

This variation simply removes the number pad.

75%

The 75% keyboard is similar to the tenkeyless (TKL) keyboard, but it rearranges the keys so they fit into a more compact space, while still omitting the number pad. An example of this is the Incase/Microsoft Sculpt Ergonomic Keyboard, and the SpiderIsland Split Ergonomic Keyboard.

65%

The 65% keeps the key basics, providing arrow keys, but omitting the function row. Keyboards like this include the Keychron K11 Pro/Max.

Compact (60%)

The 60% layout removes everything except the alphanumeric keys, modifier keys, and a single row of function/nav access via layers. There are no dedicated arrow keys, no function row, and no navigation cluster. It is among the most minimal viable keyboard layouts, and accessing secondary functions requires either memorising layer shortcuts or dedicating a key to layer-switching.

The appeal is desk space and portability: a 60% keyboard is significantly smaller and lighter than a TKL. For writers and programmers who have internalised their shortcuts, the reduction in hand travel can also be an efficiency gain. For anyone who frequently uses arrow keys, function keys, or the number row for navigation, this may prove to be difficult to adapt to.

40%

One of the smallest keyboards possible, the 40% keyboard contains only alphanumeric keys. Accessing any other functions relies on layers, or the use of multiple buttons. This layout is the most difficult to adjust to.

* * *

Keyboard Switches

In keyboards, a switch is the mechanism beneath each keycap that registers a keypress. It's the heart of the keyboard's tactile feel, responsiveness, and sound profile.

Keycap

The removable plastic cap you actually touch, sitting on top of the switch. Worth naming separately because "the key" usually means the whole assembly, and the two are separate parts. Legends wear off keycaps long before a switch stops working.

Membrane switches (rubber dome)

The mechanism in most keyboards ever sold, including nearly every keyboard that came free with a computer. A sheet of rubber domes sits over a printed circuit, and pressing a key collapses a dome to complete the circuit. It is quiet and cheap to make. It also feels mushy, because the dome resists and then gives way all at once, with no clear point at which the press registers. If a keyboard's specifications do not say what switches it uses, it is more than likely they are membrane switches.

Mechanical Switches

The parts of a mechanical switch. Click to enlarge.

When it comes to mechanical keyboards, broadly speaking there are three types of switches: linear, tactile, and clicky. They are generally identified by a colour:

Linear (or red switch): Offers a smooth press, with no tactile bump or click. Some people like this because it is quiet, but others don't because it offers little feedback when you press the key.
Tactile (or brown switch): Has a noticeable bump mid-press, with a mild click. This is a compromise between the linear and clicky switches, offering physical feedback without significant noise.
Clicky (or blue switch): Provides a tactile bump and loud click. People love these for the feedback it offers, but it is also louder than the other options which may make you public enemy number one in a quiet, but crowded space.

All switches share the basic construction shown here:

Actuation force and travel

Two numbers manufacturers publish and most people ignore. Actuation force is how hard you have to press, in grams. Around 45g is the common middle, below 40g is light, above 60g is heavy. Travel is how far the key moves before it bottoms out, usually 4mm on a full size mechanical switch and 2 to 3mm on a low profile one. Light and short suits fast typists and quiet offices. Heavy and long suits anyone who rests their fingers on the keys or bottoms out hard.

Scissor Switches

A scissor switch mechanism. Click to enlarge.

Other switches used in split keyboards include scissor switches, which are the type usually seen in laptops and low-profile keyboards. It offers a slight bump, and is relatively quiet. It works by having a plastic scissor-like mechanism stabilising the keycap, guiding its downwards movement when pressed. This then activates a rubber dome underneath that provides resistance and registers the press.

* * *

Low profile

Describes height, not mechanism. A low profile keyboard uses shorter switches and flatter keycaps, so travel is around 2 to 3mm rather than 4mm and the board itself is thinner. Laptop keyboards are the reference point. A thinner board keeps your wrists flatter, which is why low profile and ergonomic often appear in the same sentence.

Optical switches

Mechanically the same idea as a mechanical switch, except the press is registered by breaking a beam of light rather than closing two metal contacts. Nothing touches, so there is less to wear out and no contact bounce to filter. For typing they feel like the linear and tactile switches they imitate, and you would struggle to tell the difference blind.

Magnetic switches (Hall effect)

These track the position of a magnet in the stem continuously, rather than registering a single on-off point. Because the keyboard knows how far down the key is at any moment, the actuation point becomes adjustable in software, and one key can trigger two different things at two different depths. That is genuinely useful for gaming. For typing it mostly means you can tune how sensitive the keys feel, which is a nice-to-have rather than a reason to switch.

* * *

Connectivity

How the keyboard reaches your computer. This decides more about daily use than most people expect, and it is the most common thing to regret.

Wired (USB): One cable, no battery, no pairing, lowest latency. Still the most reliable option and the only one that cannot run flat mid-sentence.
Bluetooth: No receiver to lose, and most Bluetooth keyboards hold two or three devices at once with a switch key, which is the real argument for it if you move between a laptop and a tablet. The costs are a short wake-up delay after the board has been idle, and the occasional pairing issue.
2.4GHz dongle: A small USB receiver, usually faster and steadier than Bluetooth. The catch is the dongle itself. It occupies a port, and if it is factory-paired rather than replaceable, losing it can end the keyboard.

Check three things before buying: whether the dongle is USB-A or USB-C and whether that matches the ports you actually have, whether a wired fallback exists, and whether the dongle can be re-paired if it is lost. The Sculpt is the cautionary example - factory-paired USB-A, no Bluetooth, no wired option.

* * *

ANSI vs ISO keyboards

There are two primary keyboard layout types: the American National Standards Institute (ANSI), or the International Organization for Standardization (ISO) format. ANSI keyboards are more common in the United States and US influenced countries, while the ISO format is more common in Europe and European influenced countries. The key differences are highlighted in purple below, with the enter key being the most obvious difference.

ANSI Layout

ISO Layout

Updates

[tab]

2026

21 Jan 2026: Added mechanical and scissor switch diagrams
21 March 2026: Added more details on 60% keyboards
4 September 2026: Updated content

2025

9 Oct 2025 First published
14 Oct 2025: Added narrative on switches
16 Oct 2025: Added keyboard diagram, ISO vs ANSI

Image source: https://www.keyboard-layout-editor.com

Why Your Hong Kong Milk Tea Tastes Off (and Why Most Recipes Get It Wrong)

noreply@blogger.com (Unknown) — Sun, 16 Aug 2026 14:00:00 +0000

I spent years in Hong Kong drinking this tea, so I always knew exactly what a good cup should taste like. What I didn't know, until I tried to make it myself years later, was how to get there – just being able to order it at the cha chaan teng (茶餐廳) rather spoils you. So I did what everyone does: I followed the recipes online, some from large, careful, well-loved food sites, and the results were thin, or grey, or oddly perfumed, or bitter in the wrong way. My mouth knew they were wrong long before I could explain why. It took a while, and a lot of wasted evaporated milk, to work out that the problem wasn't me. It was the recipes.

This is not a takedown of anyone. Most of these recipes are written by good cooks doing their best with a drink that guards its methods closely. But the same errors turn up again and again, they get copied from one recipe to the next, and they are the reason a homemade cup so often tastes "almost, but not quite." So here, for the record, is what goes wrong – and why.

(If you just want the version that works, the full recipe is here. This piece is about why the others don't.)

Mistake 1: measuring the tea by volume

Almost every recipe online measures tea in tablespoons or cups. This is the single most common error, and it is the reason two people following the same recipe get completely different results.

Tea is not a uniform substance. A tablespoon of fluffy, whole-leaf tea and a tablespoon of dense, finely-broken tea can differ in actual weight by half. So when a recipe says "three tablespoons," it is not telling you how much tea to use – it is telling you how much space to fill, and the amount of tea in that space depends entirely on what is in your cupboard. Measure a strong brew this way and you are as likely to get something weak and grey as something drinkable.

The fix: weigh the tea in grams. Hong Kong milk tea is a strong, concentrated brew (roughly the strength of an espresso relative to a normal cup) and strength that specific can only be hit by weight. A kitchen scale costs a few dollars and is the single biggest upgrade to a homemade cup.

Mistake 2: using the wrong teas - Earl Grey or English Breakfast
(or worse, tea bags)

Recipes reach for whatever black tea is to hand: English Breakfast, a couple of ordinary tea bags, sometimes Earl Grey. Each of these produces a cup that is subtly, identifiably wrong.

Earl Grey is scented with bergamot, a citrus oil. In a milk tea, that perfume fights the malty, creamy body you are trying to build: the result tastes faintly of confusion. English Breakfast is closer, because it is usually an Assam-Ceylon blend – but it is formulated to be drunk on its own with a splash of milk, not to stand up to concentrated evaporated milk in a hard boil. It gets swamped. And tea bags of any kind hold a small amount of low-grade tea – you would need a fistful of them to reach the right strength, and even then the depth isn't there.

The fix: use unflavoured, loose-leaf black tea with real body – a Ceylon base for backbone, Assam for malt, and a little bo lei (Pu'er) for earthy depth. Hong Kong cafés use finely-broken "tea dust" for a fast, hard extraction, a good loose-leaf blend gets you the same result at home with a cleaner pour. The one thing it must not be is perfumed or lightly-blended.

Mistake 3: steeping instead of boiling

A lot of recipes treat Hong Kong milk tea like any other tea: steep gently, don't let it get bitter. This is backwards. The drink is meant to be brewed hard – a full rolling boil for several minutes, not a delicate steep – because it has to be strong and tannic enough to survive being cut with rich evaporated milk.

If you brew it gently to avoid bitterness, you get a cup that tastes fine on its own and then vanishes the moment the milk goes in: thin, pale, and forgettable. The bitterness you were trying to avoid is the backbone the milk is supposed to lean against.

The fix: boil it hard, and don't panic when the unmilked tea tastes harsh. That's correct. The milk is what balances it. If your finished cup is too bitter, the answer is more milk – not weaker tea.

Mistake 4: fresh milk instead of evaporated

Some recipes (including ones that should know better) call for regular fresh milk, or offer it as an easy substitute. It is not a substitute. Fresh milk, even full-cream, is mostly water, and it is far too thin to stand up to a brew this strong. The result tastes weak and washed-out, like tea someone spilled milk into rather than a real HK milk tea.

Hong Kong milk tea is built on evaporated milk – milk with around 60% of its water removed, leaving something concentrated, rich, and faintly caramelised. That concentration is the entire point: it is thick enough to give the tea its signature velvety body. Fresh milk cannot do this, and no amount of it will.

The fix: evaporated milk for body, and (at home) a little condensed milk for sweetness. In Hong Kong the default brand is Black & White, prized for its mouthfeel. This single swap is the most common reason a homemade cup tastes "off," and correcting it fixes more bad milk tea than anything else on this list.

Mistake 5: the "red tea" translation trap

This one isn't about technique, but rather about a word that sends people to the wrong shelf entirely. Hong Kong milk tea is made with black tea. But the Chinese name is 紅茶, which translates literally as "red tea," and recipes repeat that phrase without explaining it.

In English, "red tea" usually means rooibos – a completely unrelated South African herbal tea. Follow a recipe that says "red tea" at face value and you can end up brewing the wrong plant. The name refers to the reddish colour of the brewed liquid, not the leaf. In every practical sense, reach for strong black tea.

Mistake 6: treating "orange pekoe" as an answer

Search for what tea actually goes into Hong Kong milk tea and you will usually be told to use broken orange pekoe. It sounds authoritative. For a home cook standing in front of a shelf, it is very close to useless.

Orange pekoe is not a variety of tea. It is not a region or a flavour profile, and it has nothing to do with oranges. It is a size grade – a description of how big the leaf pieces are once they have been sifted. Orange Pekoe is whole leaf, Broken Orange Pekoe (BOP) is the same leaf broken smaller, Broken Orange Pekoe Fannings smaller again, and Dust is the finest of all. A packet marked BOP is telling you the dimensions of what is inside. It says nothing about where the tea came from or how it will taste.

So "use broken orange pekoe" answers the wrong question. It is the tea equivalent of asking which flour to buy and being told "milled." Two BOP teas from different estates can taste nothing like each other, and neither is guaranteed to bring the malt and backbone this drink needs.

What the grade does tell you is how quickly the tea will give itself up. Smaller pieces mean more surface area, which means a faster, harder extraction. That is why Hong Kong cafés work with dust-grade tea, which surrenders everything into the water in minutes, at volume, under pressure. The grade points at method, and stops there.

The fix: buy on origin and character. Look for unflavoured loose-leaf Ceylon for backbone, Assam for malt, and a little bo lei (Pu'er) for depth. If the packet happens to say BOP or fannings, treat that as a small bonus rather than a requirement. Dust-grade is what the cafés use and it is genuinely hard to source outside Hong Kong wholesale channels, so loose leaf is the practical answer at home – brew it hard enough and long enough and it will get you there.

Why the same mistakes keep appearing

These aren't random errors. They cluster and persist for reasons that have little to do with the tea itself.

Recipe writing online is, by necessity, a fast business. Recipes get adapted from other recipes; a measurement in tablespoons gets copied because the last ten versions used tablespoons, an easy fresh-milk substitution gets added because it makes the recipe more accessible to a reader who doesn't have a can of evaporated milk in the cupboard. Each individual choice is reasonable. Compounded across a whole category, they drift the drink steadily away from what it actually is. And because most of these recipes are written from outside the cha chaan teng tradition, there is no first-hand memory pulling them back to the original. The copy becomes the reference.

It is just what happens to a strong, specific, culturally-embedded drink when it is documented at scale by people optimising for readability and speed. The result is a kind of consensus recipe that is plausible, popular, but completely wrong.

Setting the record straight

So, for anyone who has made a disappointing cup and assumed they were the problem: you probably weren't. Weigh your tea in grams. Use a strong, unflavoured loose-leaf blend, not bags and not Earl Grey. Boil it hard. Use evaporated milk, not fresh. Ignore the words "red tea" – it's black tea. And buy on origin, not on grade – "orange pekoe" describes leaf size, not flavour.

Get those five things right and you are most of the way to a proper cup. The full recipe, with quantities and method, is here: the one I landed on after making every mistake on this list myself.

Best Ergonomic Desktop Split Keyboards 2026: Reviewed & Ranked

noreply@blogger.com (Unknown) — Sun, 16 Aug 2026 12:18:47 +0000

Last updated: August 2026.

Some keyboards are made to be carried. This guide is about the other kind, the ones that earn a permanent place on your desk.

If you sit at the same desk most days, portability stops mattering and other things take over. How comfortable is it over an eight-hour stretch? Does it hold your wrists at a neutral angle? Does the build last decades rather than years? A board that is a little heavy or a little bulky is no handicap at a desk, which opens the field to keyboards that would never make a travel guide, and to some of the best ergonomic typing you can buy.

Need one you can carry? If you need a split keyboard you can throw in a backpack and carry between offices, that is a different set of trade-offs. See our companion guide to the best portable and travel split keyboards.

This guide sticks to familiar, row-staggered layouts, the unibody splits (also called Alice) and the dual-body splits, boards you can sit down at and type on immediately. We leave out the columnar and ortholinear boards, the Moonlanders and Glove80s of the world. They are fine machines, but their straight-column layouts mean relearning how to type from scratch, which is a different journey for a different reader.

The Full Collection. Our expert ergonomic keyboard and mouse guides, reviews, and previews: https://www.bulsuk.com/p/all-our-ergonomic-keyboard-reviews.html

This site is a one-person effort, with keyboards bought and tested on our own dime. If a guide here saved you from a bad purchase, consider buying us a coffee.

☰ In this guide

Our top picks
Comparison table
On our radar
FAQs

The Best Ergonomic and Split Keyboards for Desktop in 2026

Top pick

Microsoft / Incase Sculpt Ergonomic Keyboard

More than a decade after its 2013 release, the Sculpt is still the best all-round desktop ergonomic keyboard you can buy, and it is not close. What sets it apart is the negative riser: where most keyboards sold as "ergonomic" prop the back up and bend your wrists upward, the Sculpt slopes gently downward, keeping your wrists neutral. Combine that with a gentle split curve, a firm cushioned palm rest, and scissor switches with 2.3mm of travel, and you get a keyboard your hands stop noticing. Its one real weakness is the factory-paired USB-A dongle, with no Bluetooth or wired option. Read our full Sculpt keyboard review.

In stock at some resellers, backordered at others – check before buying

Best dual-body split

NocFree &

If you want two genuinely separate halves, the NocFree & is the best dual-body split we have tested. It is a solid, well-built mechanical wireless split, with the freedom to position each half exactly where your shoulders want it, all in a relatively small 75% form factor. The typing feel, ergonomics and build are outstanding. What holds it back from a higher score is the ownership experience rather than the hardware: out-of-the-box QC issues and a risky firmware update process. Given its weight class it straddles the line between portable and desk-bound, but on a desk that weight is a non-issue. At $269–349 it is a premium choice, but for typing comfort it comes recommended. Read our full NocFree & review.

In stock

Compact / also travels

Keychron K15 Max

Most people reading this want a dedicated desk keyboard. But if you hot-desk, or want one board that works at your desk during the week and slips into a bag at the weekend, the K15 Max is worth a look. The reason we are comfortable naming it is simple: it is the same board as the Keychron K11, which I typed on daily for three years. It uses the same low-profile Gateron switches, keycaps, QMK firmware, and aluminium body, in a larger 75% Alice layout. That proven typing feel is what earns it a place here, and at a desk its 700g weight is a non-issue. It starts around $100 and, unlike the K11, you can actually order one. Read our review of the K11 to understand the K15's typing experience.

Orderable direct from Keychron

Runner up

SpiderIsland Tech Split

For old-school typists who miss the deep travel and thunk of a proper mechanical keyboard, the SpiderIsland delivers the desktop feel the Sculpt deliberately does not. It is a dual-body split, so both halves move freely, and despite a no-name pedigree the keys are comfortable and accurate. It is wired-only and a thick slab, which is no problem on a desk, and at $82 it is the affordable way into a mechanical split. Read our full Spiderisland split review.

Available

Budget – testing

BOW Ergonomic Keyboard

Every so often a generic board turns up that has no business being as good as it is. The BOW Ergonomic Keyboard, sold under its full name "B.O.W. Hangshi three-model ergonomic keyboard", is shaping up to be one of them. It is an inexpensive unibody split (the same family as the Sculpt), a 75% scissor-switch board with 81 keys, a gentle upward curve, and tri-mode connectivity (Bluetooth, 2.4GHz, and wired USB-C). In early use it works surprisingly well as a budget Sculpt alternative. Still in testing, so no rating yet, but at around $55 it fills a real gap.

Available

Quick Comparison: Desktop Ergonomic Keyboards

Scroll → to see all columns

Keyboard	Price (USD)	Rating	Switches	Connectivity	Best For
Microsoft / Incase Sculpt Top pick	$120	●●●●● 5.0	Scissor	USB-A dongle	Best all-round desktop ergonomic
NocFree & Best split	$269–349	●●●●○ 3.5	Mechanical	Wireless	Premium two-piece split
Keychron K15 Max	from $100	Not yet rated	Mechanical	BT + 2.4 GHz + USB-C	If your desk board must also travel
SpiderIsland Tech Split Runner up	$82	●●●●○ 4.0	Mechanical	Wired only	Deep-travel desktop feel
BOW Ergonomic Keyboard Reviewing	~$55	Reviewing	Scissor	BT + 2.4 GHz + USB-C	Budget Sculpt alternative

Prices approximate as of July 2026. USD throughout.

On our radar

These are other desktop ergonomic and split keyboards that come up often, or that we would like to get our hands on. We have not tested them ourselves, so none carries a rating, this is context rather than a recommendation. If you want the full head-to-head on Sculpt-style unibody boards specifically, our dedicated alternatives guide has the detailed comparison.

Keychron Q11 and Q11 Ultra 8K

Keychron's own dual-body split, and the brand-name answer to the NocFree &. The original Q11 ($204.99) is a wired 75% aluminium split, orderable on backorder. The newer Q11 Ultra 8K ($239.99) adds fully wireless halves over 2.4GHz plus Bluetooth, though as of this writing it shows sold out with a notify-me only. Worth knowing that a full-metal brand-name dual-body split costs less than the NocFree &, which sets the bar its extra outlay has to clear. We have not tested either, so no rating, but they are firmly on our radar.

Logitech ERGO K860 – a popular unibody ergonomic board with an integrated numpad.
Kensington KB675 EQ – a TKL ergonomic option.
Matias Ergo Pro – a mechanical dual-body split aimed at Mac users.
Kinesis Freestyle / mWave – long-standing adjustable dual-body splits.

Deliberately not here

We leave out the columnar and ortholinear enthusiast boards, the ZSA Moonlander, the Glove80, and their kind. They can be wonderful for the people who commit to them, but they ask you to relearn typing on a straight-column grid, and they cost far more than anything on this list. Great machines, wrong guide.

Keyboards we are reviewing

BOW Ergonomic Keyboard – in hand, testing underway. (See above.)

Best ergonomic keyboard for spreadsheet and office work

For heavy number entry the separate numpad of the Sculpt is a genuine advantage, letting you centre the main keyboard and keep the mouse close. If you want the number row built in, the Keychron K15's 75% layout keeps digits a single reach away.

Best ergonomic keyboard for wrist pain and comfort

The Sculpt's negative riser is the single most important feature here, keeping the wrist neutral rather than extended. No keyboard is a medical device, and none of this is medical advice, but the science of why typing posture matters is covered in our dedicated piece: Wrist and Shoulder Pain from Typing.

Best split keyboard layout, explained

Unibody splits (also called Alice), dual-body splits, and columnar boards all mean different things, and the difference decides how much relearning you face. Our keyboard encyclopedia breaks them down.

Need a keyboard you can carry? If portability is your priority, see our companion guide to the best portable and travel split keyboards, covering foldables, the Keychron K15, and ultralight premium splits.

FAQs

[accordion]

What is the best ergonomic keyboard for a desk?

The Microsoft/Incase Sculpt Ergonomic Keyboard. More than a decade after release it is still the best all-round desktop ergonomic keyboard, thanks to its negative-riser design that keeps the wrists neutral, a comfortable split-dome layout, and a separate numpad. Its only real weakness is that it connects via a USB-A dongle, with no Bluetooth or wired option.

Is the Microsoft Sculpt still worth buying in 2026?

Yes. Although Microsoft stopped making it in 2023, Incase now manufactures it, and the design has aged remarkably well. It remains our top pick. The main caveat is availability: it is in stock at some resellers and backordered at others, so check current stock before you commit.

What is the best split keyboard for a desktop?

For two genuinely separate halves you can position shoulder-width apart, the NocFree & offers an excellent wireless dual-body split, and at a desk its weight is a non-issue. For a more affordable wired option, the SpiderIsland Tech Split delivers deep mechanical key travel. Keychron's own Q11 is a brand-name alternative worth watching.

What is a negative riser or negative tilt on a keyboard?

A negative riser slopes the keyboard gently downward, away from you, rather than propping the back up. This keeps the wrists neutral instead of bending them upward, which reduces the strain that causes typing pain. The Sculpt is the standout example, and it is a big part of why it remains our top desktop pick.

Should a desktop ergonomic keyboard be wired or wireless?

At a fixed desk it matters less than for travel. Wireless keeps the desk tidy, but a wired board never runs out of battery and has no dongle to lose. The Sculpt is dongle-only, the SpiderIsland is fully wired, and the Keychron K15 and NocFree & offer full wireless. Choose based on how much you value a clean desk versus never charging anything.

Do I need a numpad on an ergonomic keyboard?

Only if you do heavy number entry. The Sculpt includes a separate numpad, which is ideal because it lets you centre the main keyboard and keep the mouse close, better ergonomics than a built-in numpad that pushes the mouse further away. Most of the split keyboards here omit the numpad to stay compact.

Why are columnar or ortholinear keyboards not recommended here?

Columnar and ortholinear boards like the ZSA Moonlander and Glove80 arrange keys in straight vertical columns rather than the staggered rows everyone learns to type on. They can be excellent for people who commit to them, but they require relearning how to type, and they cost far more. This guide sticks to familiar layouts you can use immediately.

Why the Keychron K15 Max Is Our Pick To Replace the K11 Max

noreply@blogger.com (Unknown) — Mon, 10 Aug 2026 14:16:50 +0000

There is a problem at the affordable end of the split-keyboard market, and its name is stock.

For three years, the board we recommended above all others was the Keychron K11 Max, a 65% Alice-layout keyboard we carried, tested, and wrote about at length. On typing feel and build it is the best portable split keyboard we have used. The trouble is that you can no longer buy one. As of August 2026, Keychron's own product page shows the K11 Max as "sold out", offers only a "notify me" button, and gives no way to place an order. It has disappeared from AliExpress and Amazon as well. Keychron has announced nothing, but a keyboard that cannot be bought anywhere, from the maker or any reseller, is discontinued in all but name.

So what should you buy instead? Our answer is the Keychron K15 Max.

It is the K11, with more keys

The K15 Max is not a different keyboard that happens to look similar. It is the same platform. It uses the same low-profile Gateron mechanical switches, the same LSA keycaps, the same QMK firmware and Keychron Launcher software, and the same aluminium body. We spent three years learning how that platform types, how it holds up, and how its software behaves, and the K15 inherits all of it.

The difference is layout. The K11 is a 65% Alice board and the K15 is a 75% Alice board. In plain terms, the K15 adds a top function-and-number row and a navigation column. If you have ever wished the K11 had dedicated function keys without a layer, the K15 is that wish granted. Everything that made the K11 good to type on should carry over unchanged, because mechanically it is the K11.

The one thing shared engineering does not tell us: size

More keys mean a bigger board. The K15's 75% layout has a wider footprint than the K11's 65%. It is still a keyboard you can carry in a backpack, but it is a larger one, and if outright portability is your priority, the smaller K11 had the edge.

That said, if you are shopping now, at this budget, for a low-profile Alice board you can actually order, the K15 Max is the standout, and it is built on a platform we have trusted for years.

What it costs, and where to get it

The K15 Max starts at US$100 for the white-backlight version, direct from Keychron, rising to roughly US$130 to $150 for the hot-swappable RGB version. Crucially, and unlike the K11, it is orderable. Keychron's page shows an active "Add to Cart" on backorder, so you can place an order today rather than wait on a restock that may never come.

The bottom line

The K11 Max is gone. If you already own one, hold on to it, because it remains an excellent keyboard. For everyone buying today, the K15 Max is the same keyboard in a slightly larger 75% layout, it is actually in stock, and it sits comfortably under the price of anything genuinely better.

Back to the full guide: see all our tested picks in The Best Portable and Travel Split Keyboards.

NocFree & Review: Great Keyboard, Bumpy Ownership

noreply@blogger.com (Unknown) — Sun, 02 Aug 2026 14:41:52 +0000

The NocFree & is a fully split, wireless mechanical ergonomic keyboard. Unlike similar boards, both halves are completely independent of each other, letting you place each side wherever suits you. Build quality is solid, the typing experience is excellent, and while it is portable, it straddles that definition. It weighs about as much as a laptop.

PriceUS$269–349

Our rating●●●●○ 3.5 / 5.0

Pros

Outstanding, cushioned low-profile typing feel with hot-swappable switches
True standalone wireless split (no cable between the halves)
Substantial build weight stops any desk sliding
Quick, highly accessible support team via Discord
Clean browser-based key remapping and layer support

Cons

Heavy, and an oversized case negates true portability
Severe battery drain with backlighting on (1–2 days max)
Out-of-the-box QC issues (missing feet and screws) and risky firmware updates
Keycap printing looks soft and fuzzy
No battery indicator on the board itself
Expensive

NocFree & at a glance
Layout	75% row-staggered split
Connectivity	2.4GHz dongle, Bluetooth 5.0 (×3), USB-C wired
Keycaps & switches	Low-profile PBT (dye-sub), hot-swappable
Battery life	2–3 weeks (RGB off) / 1–2 days (RGB on)
Weight	1.03 kg / 2.27 lbs
Software	NocFree Link (Chromium WebHID)
Price	US$269–349

Typing experience

The NocFree & is a 75%, row-staggered keyboard, which means anyone who has learned to type in the past century will pick it up quickly without relearning the skill. It ships with a choice of linear or silent switches and is hot-swappable if you prefer something else. As a low-profile board, it will feel familiar to anyone who has used a laptop keyboard.

I ordered the linear switches with some trepidation. My previous experience with linears was the Spiderisland keyboard, which felt hollowed out while typing. The difference here is that the NocFree & is both low-profile and cushioned, which prevents that hollow feeling.

On sound, it stays quiet, muted and gentle. It produces a soft, cushioned tone rather than the sharp clack you get from many mechanical boards, which makes it easy to live with in a shared office or on calls.

Each keyboard takes a period of adjustment, and I got my typing speed back to normal quickly. On first use I managed 75 WPM, and over the following days this climbed to 85 WPM. For comparison, on my usual daily drivers, the Keychron K11 Max and the Microsoft Sculpt, I manage 85 to 95 WPM on a good day.

One thing I did find annoying was the placement of the Print Screen key. I kept hitting it as my finger reached for backspace. It was annoying enough that I remapped the key to "End" to stop taking unintended screenshots.

Design and ergonomics

After several months as my daily driver, I can confirm this is a comfortable keyboard thanks to its flexible design. I felt no fatigue even after a long day of extensive typing, and it takes up relatively little space on the desk.

The keys need very little force to actuate, and that ease of actuation took some adjustment. Occasionally, while resting my fingers on the keys, I pressed one by accident.

Unlike Alice (unibody) keyboards, the fully split design means you can place the halves as close together or as far apart as you like, at almost any angle. This is great for typists who want the keyboard at shoulder width, or who need a specific angle that Alice boards cannot provide.

Another benefit is that the keyboard is fully wireless. Getting set up at the desk is as easy as turning on two switches. Most split keyboards are either fully wired or have a cable bridging the two halves, so a truly cordless split is still a rare thing.

Portability

There is no way around it. The NocFree & is heavy, weighing 1.03 kg (2.27 lbs). That is as much as an ultraportable laptop such as the Lenovo X1 Carbon, and you will feel it in your backpack.

While it is technically portable, it really straddles the line between a portable split keyboard and one that should stay on a desk.

The case, for those who order it, is more akin to a suitcase. It is well designed and has a compartment for everything, but it is gigantic. It will fit into a backpack but leaves room for little else. It is impractical to use, and I am a little disappointed the team at NocFree did not design something smaller and lighter.

To carry it around, I ended up buying a couple of padded envelopes. It is not the most elegant solution, and since the envelopes were larger than the keyboard it wasted space, but it beat carrying the board in my backpack with no cushioning.

Build quality

Because of its weight, this keyboard is a tank. It feels substantial and does not slide on the desk. The only niggle is that the USB-A dongle fits tightly in USB-A ports and can be a bit awkward to pull out.

The included low-profile PBT keycaps feel solid, but the printed letters are not especially sharp. They almost look like they were printed on an older inkjet, which is surprising given they are dye-sublimated, and I found that a little disappointing.

Customisation

The NocFree & supports full customisation through NocFree Link, which lets you remap the keys and set up layers. NocFree Link uses WebHID for browser-based remapping. Like most WebHID tools, it currently requires a Chromium-based browser (Google Chrome, Edge or Brave) and will not work in Firefox or Safari.

NocFree also includes a few extra keys in the package, including Mac- and Windows-specific keycaps and some orange ones. My pack came with a special cat keycap, which I repurposed as my End key. They sell additional switches and keycaps separately if you want spares.

Connectivity

You can connect the keyboard wirelessly through the included USB-A dongle, over Bluetooth (which supports up to three connections), or wired. The USB-A dongle lives underneath the "&" symbol on the right side of the board, cleverly hidden behind a metal cover.

It works with both Mac and Windows, and includes keycaps for each in the box.

The keyboard has USB-C ports on both sides for charging. If you use a wired connection, you only need to connect the left half to your computer, and the right half connects wirelessly to it.

The keyboard wakes quickly. I just tap a key to make sure it has reconnected before carrying on.

In earlier firmware I occasionally experienced sticky keys, but subsequent updates sorted this out. I have relied solely on Bluetooth and had no connectivity issues. My tests with the wired connection and the USB-A dongle also showed no problems.

Battery life

With the backlight off, the keyboard manages 2 to 3 weeks of average office use. With it on, you will get 1 to 2 days at most. The battery life with backlighting on is so poor that I turned it off permanently in software. It is a little perplexing how much it drains. For comparison, the Keychron K11 Max manages almost 2 weeks with backlighting, and that is after more than 2 years of consistent use.

You can check the battery level through NocFree Link. As of the current firmware, there is no way to check the remaining charge on the unit itself.

Note on charging safety: NocFree explicitly warns against using high-wattage USB-PD quick chargers (such as MacBook or phone bricks), which do not negotiate properly with the board's internal power delivery. Charge it via a standard 5V computer USB port to avoid damaging the batteries.

Other accessories

NocFree does sell a separate tenting kit, numpad and charging stand, but since they add to the overall price, I did not buy them to test.

Firmware update experience

Updating the firmware is a smooth process, though it is geared towards the intermediate user. You download the firmware, which comes as three separate files: one for the left side, one for the right, and one for the dongle. You then activate firmware update mode and drag the correct files into the Finder or Windows Explorer drive that the keyboard mounts. Install the wrong firmware and you brick the keyboard.

While not the most user friendly method, the instructions for firmware updates are actually pretty good, and they also have videos taking you through it.

Challenges and other quality issues

NocFree is a small company, and I give it more leeway than I would a product from Logitech or Microsoft. That said, some of the issues I ran into are difficult to ignore.

Quality control issues: when the keyboard arrived, the screws holding the unit together and the rubber feet were completely missing. I contacted NocFree through Discord and, to their credit, they responded quickly and shipped the screws, tools and rubber feet so I could install them.

Firmware fiascos: about a month in, I updated the firmware NocFree posted and ended up bricking my keyboard. Support was quick to investigate and respond, and it took two rounds of troubleshooting before the board was restored, roughly two weeks end to end.

Flashing the firmware on a bricked board is also awkward. It involves removing the keyboard's back and shorting the circuit board with two screwdrivers, all while it is plugged into the computer. Non-technical users may struggle.

I do not anticipate further firmware problems, since these felt like early teething issues.

Conclusion

The NocFree & delivers on its core promise: an exceptionally comfortable, low-profile, truly wireless split typing experience that opens up your posture. It is also portable, though at the very edge of what counts as easy to carry.

The hardest part of scoring it is that the keyboard is better than the ownership experience. The typing feel, ergonomics and hardware build are outstanding, but out-of-the-box quality control hiccups and a failed firmware update stopped me fully trusting the board during the review period.

Where factory QC falls short, though, NocFree's human element shines. Their support operates with a speed, directness and genuine care you rarely get from megacorporations like Logitech or HP, who often seem allergic to customer contact altogether.

It is also expensive. You can get capable split and ergonomic boards from Keychron, or cheaper alternatives on AliExpress, for noticeably less. The premium here buys the truly wireless split design and the direct, responsive support, not a longer spec sheet.

If you are an ergonomic enthusiast who appreciates hardware innovation and values responsive, direct-from-maker support, it is easy to recommend. But if you want a seamless, plug-and-play travel companion, you may want to wait while NocFree irons out the early-batch teething pains.

The NocFree & is sold direct from NocFree.

Frequently asked questions

Can you fast-charge the NocFree &?

No. NocFree warns against high-wattage USB-PD chargers such as MacBook or phone bricks, because they do not negotiate correctly with the board's internal power delivery. Charge it from a standard 5V computer USB port to avoid damaging the batteries.

How is the battery life on the NocFree &?

With the backlight off it lasts 2 to 3 weeks of average office use. With RGB on you get 1 to 2 days at most, so I turned the lighting off permanently in software.

Does the NocFree & work with Mac and Windows?

Yes. It ships with both Mac and Windows keycaps, and connects over a 2.4GHz dongle, Bluetooth (up to three devices), or USB-C wired.

Is the NocFree & truly wireless between the two halves?

Yes. Both halves are fully independent with no bridging cable. If you connect it wired, you only plug the left half into your computer and the right half links to it wirelessly.

Can you remap the keys on the NocFree &?

Yes, through NocFree Link, a browser-based tool that uses WebHID. It handles remapping and layers, but only works in Chromium browsers like Chrome, Edge and Brave, not Firefox or Safari.

Is the NocFree & good for travel?

Yes, but it weighs 1.03 kg (2.27 lbs), about as much as an ultraportable laptop, and the optional case is suitcase-sized. It is technically portable but sits at the very edge of easy to carry.

Reading the Room Before It's a Finding: A Diagnostic for Governance That Reads Signals

noreply@blogger.com (Unknown) — Sun, 19 Jul 2026 14:04:38 +0000

In April 2005 a rapid commuter train near Osaka took a 70 km/h curve at 116, derailed into an apartment building. One hundred and seven people died. That failure, told in full in the original article “The Station that Knew: On Governance that Cannot Read Signals,” was not a failure of information. It was a failure to read it. JR West had formal risk functions, reporting and governance. What it lacked was the ability to read the weak signals its own people and systems were already producing. The pattern is not particular to railways: Wells Fargo's account-opening figures told the same story for years before the bank acted on them as one.

Most organisations' risk reporting talks about emerging risk. This generally covers horizon scanning and early indicators, and sits on top of a governance function focused on reading and managing the familiar categories: financial, operational and compliance. These are mature disciplines. The ability to read weak signals, patterns below the threshold of action but too persistent to be dismissed as noise, receives less attention.

This piece is a diagnostic. It asks three questions about your own function, to test whether that gap exists. Each one makes visible a choice you have probably already made without noticing.

1) When someone sees a weak signal, does it have somewhere to go that is not the machinery built for findings?

In the years before the Amagasaki derailment, drivers on the line kept noticing things: timetables that left no margin, near-misses they quietly absorbed. The information existed. It simply had nowhere to go that would treat it as anything other than a driver's failing. In your organisation, where does that kind of report go?

For many, it is ingested into the same machinery which handles confirmed problems. Weak signals, often several at once, none conclusive alone, are asked for evidence they do not have, measured against a rubric which assumes the problem has already happened, and closed when they cannot meet those standards. The function was simply not designed to hold a hunch, or a yet-unvalidated pattern, long enough to determine whether it is real.

The harder challenge is not building this capability but protecting it: keeping it from being rolled into the existing machinery, where it will be held to the wrong standard and quietly blunted. It is a discipline before it is a department. But it still needs an owner, and in a regulated firm that owner is an accountable individual who already exists; it does not require a new one.

This is not only a matter of better indicators. A good key risk indicator suite catches a great deal – but an indicator only tells you about a risk someone already decided was worth watching. Weak signals are the other kind. JR West's method for grasping incidents was to wait for crew to report them, rather than to mine the data its own recorders already held. There was more in those records than anyone was reading. No one had been given the job. A metric without an owner is not a control. It is a record kept for the inquiry.

The test: ask to see the route a pattern-level concern actually took last quarter – where it was raised, where it went, where it sits now. If the only route runs through the process built for confirmed breaches, a weak signal has nowhere to go but the machinery built to close it.

2) What happens to the person who raises a concern that turns out to be wrong?

Consider the position of someone who notices something early. The pattern is real to them but not yet provable. If they raise it and they are right, there may be quiet credit once the thing has confirmed itself. If they raise it and they are wrong, the cost is immediate and personal: the time they took from colleagues, the meeting they called, the senior attention they spent on something that came to nothing. The reward is back-loaded and uncertain. The cost is front-loaded and sure. A rational person doing that arithmetic waits until the pattern is strong enough to survive challenge – by which point it is no longer an early signal but a confirmed finding, and the window has closed.

The function does not lose its early warning to bad people or bad culture. It loses it to ordinary people responding sensibly to the incentives around them.

Aviation faced a sharper version of this problem and solved it. Since 1976, US aviation staff report near-misses not to the regulator that can ground or fine them, but to NASA, which cannot. NASA strips out the identifying detail, finds the patterns across thousands of such reports, and passes the hazards to those who can act on them. Because the receiver has no stake in the reporter's fate, and because filing itself earns limited protection from the regulator, the reporting is honest, and the near-misses surface while they can still be acted upon.

A corporation cannot build a NASA. There is no neutral, stake-free receiver inside a company. But it can create distance: whoever receives a weak signal should have no hand in judging the person who raised it.

None of this is whistleblowing, and the distinction matters. Most jurisdictions already mandate a protected channel for disclosable conduct: whistleblower regimes, mandated by statute across the UK, US, EU and Australia, and by listing rules and corporate codes in parts of Asia. They exist for a specific and serious category: wrongdoing, breaches, danger to the public. A weak signal rarely starts as any of those. The statutory regime was not built for weak signals, and routing early patterns through it either inflates them into allegations they are not, or sees them thrown out without further investigation. The channel this piece describes sits alongside the whistleblower regime, not inside it, and does the work the statute was never meant to do. None of this diverts a genuine disclosure from the protected channel. That route must remain.

The design question is narrower than it looks. Three things have to hold. The act of raising a concern must be insulated, as far as anything can be, from the individual's performance review. The judgement to raise it must be recognised on its own terms, not only when it turns out to be right. And the record must not score people for being wrong: investigated and closed is honest; false positive presumes a correct answer existed at the time, which for a weak signal it did not.

The obvious objection is that this licenses crying wolf. It does not, because what it protects is the substance of an honest concern that did not pan out, and nothing more. Fabrication, bad faith, raising a flag to damage a rival: these are different matters, handled as they always were. The line between bad faith and honest-but-wrong will always be governed by professional judgement, usually made in hindsight, often by the same people the protection is meant to hold at arm's length. No structure removes that judgement entirely. What a good one does is raise the bar for treating a wrong concern as misconduct, and put the decision somewhere other than the desk of the manager whose work was questioned. The protection is for honest judgement. It was never meant to cover everyone.

The test: find someone who raised a concern in the last two years that came to nothing, and ask whether they would do it again. Their answer is the state of your architecture, whatever the policy says.

3) Can your function see a pattern forming between reporting cycles?

Governance runs on a calendar. This is appropriate for deliberate decisions, which should be made with evidence and at a measured pace. It is poorly suited to a pattern that forms in the six weeks between two meetings and is eight weeks old by the time anyone whose job it is to act on it is in the room.

Reading signals in real time does not mean replacing the calendar. It means having something alongside it: a standing point at which the current pattern landscape is reviewed at higher frequency than the board cycle, an agreed authority to pull a developing concern forward to the risk committee chair ahead of the next scheduled slot when it crosses a threshold, and a sharper setting during periods of known strain, such as a restructure, an integration, or a leadership change – when the signals matter most and the organisation is least inclined to look.

This costs something, and the cost is worth naming. It is senior attention, the scarcest thing a governance function spends, and the discipline to keep the exercise from slumping into one more status meeting that reviews what is already known. For an organisation in a high-consequence sector, that cost is smaller than the failure it guards against. For others it is a real trade-off, honestly made. The point is not that every function should build it, but that a leader should know whether theirs has.

The test: ask when the function last brought a concern to the board off-cycle, on the strength of a pattern rather than a finding. If the answer is never, the function runs on the calendar alone, which may be the right design, but should be a decision rather than a default.

What the three questions have in common

None of these is, in the first instance, a question about culture. Not because culture does not matter, but because culture is downstream. It forms in response to the structure around it. That is the argument the original piece makes at length. The enforced system, not the stated one, is what shapes how people actually behave. Structure comes first. It does not remove the human judgement at the centre of all this, but it shapes and constrains it – and it is the part a leader can actually decide: how the function is built, what it is allowed to hold, who is protected, and when it is permitted to look. You cannot mandate a culture into being. You can build the structure it forms inside. Most of the time, that structure is the thing no one chose to examine.

Answering all three questions honestly means going and looking: pulling the actual escalation cases, asking the people who raised them, finding out what happened to those people afterwards. The going-and-looking is the diagnostic. The answers are usually less comfortable than the policy would suggest.

Governance is not, in the end, what is written in the framework. It is what happens to the person who notices something before the system does.

[message]

Read more: Take a look at our ongoing series on Governance.

Will a Watch Winder Charge a Seiko Kinetic? (No It Won't - Here's Why)

noreply@blogger.com (Unknown) — Sun, 19 Jul 2026 13:08:49 +0000

If you own a Seiko Kinetic and an automatic watch winder, the assumption is reasonable enough: the winder keeps your mechanical watches wound, so it should keep the Kinetic charged too. Both are powered by wrist motion, both have an oscillating weight spinning away inside, and a winder exists precisely to supply that motion when your wrist can't.

It doesn't work. Or rather, it works so poorly that it isn't worth doing. Put a dead Kinetic on a standard winder and, days later, you'll often find it no better off than when you started. The reason is not a fault in the winder or the watch – it's a difference in physics. The pieces of the answer are scattered across watch forums and technical manuals; the point of this piece is to pull them together in one place.

Two watches that look the same and aren't

A conventional automatic watch stores energy mechanically. The oscillating weight – the rotor – turns on a pivot as you move, and through a train of reverser and reducing gears it winds a mainspring, a coiled ribbon of spring steel. A sliding bridle inside the barrel lets the spring wind to full tension and then quietly slips to prevent overwinding. The mainspring holds that energy as physical tension and releases it slowly to drive the watch, typically for around two days on a full wind.

The key point for our purposes is that mechanical winding is cumulative. The energy stored builds with how far the spring is wound, and it does not much care how fast that happened. A rotor turning slowly still winds the spring, just gradually. Through high-ratio reduction gearing, even gentle, unhurried motion adds tension. This is exactly why a winder works for an automatic: slow, steady, programmed rotation – a few hundred to a couple of thousand turns a day – is perfectly sufficient to keep the mainspring topped up.

A Seiko Kinetic looks almost identical from the back. Same oscillating weight, same swinging motion. But it stores nothing mechanically. The rotor instead drives a tiny electrical generator through a step-up gear train, spinning a small magnet-and-coil generator at very high speed – figures of up to 100,000 rpm are cited in secondary sources, though Seiko's own technical pages decline to commit to a per-calibre number. That generator induces an electric current, which is rectified and stored: in a capacitor on early movements, or in a rechargeable cell on later ones. Seiko refers to this storage component, with admirable evasiveness, as the "Kinetic Electricity Storage Unit," which neatly avoids committing to either word.

The mechanism that makes the Kinetic clever is the same mechanism that makes it incompatible with a winder. And the reason is one line of physics.

Why a generator doesn't care how far you turn it – only how fast

A generator produces voltage in proportion to how quickly the magnetic field through its coil is changing. Turn the rotor slowly and the voltage is low; turn it fast and the voltage is high. This is Faraday's law, and it has a brutal consequence for the winder question: a Kinetic's output depends almost entirely on rotor speed, not on accumulated turns.

Worse, the generated voltage has to clear a threshold before the charging circuit will accept any of it at all. Below a certain rotor speed, the watch banks precisely nothing – not a little, nothing. There is no slow trickle the way there is with a mainspring. The energy either arrives fast enough to count, or it doesn't arrive.

Now consider what a standard winder actually does. It rotates the entire watch smoothly and continuously around a fixed axis. The weighted rotor, pulled by gravity, hangs at the bottom of its arc while the case turns slowly around it. Relative to the generator, the rotor is turning – but at the winder's pace, perhaps a fraction of a revolution per minute. That is far below the speed the generator needs to induce a usable voltage. The rotor is technically moving. The charging circuit sees nothing.

This may also explain why Kinetics are commonly reported to go flat on desk-bound office workers. It isn't about hours worn. A still wrist held at a keyboard for eight hours delivers steady, low motion – plenty for an automatic, useless for a generator. The Kinetic doesn't want hours. It wants events.

What the Kinetic actually wants: acceleration, not rotation

The motion that charges a Kinetic is not smooth rotation but abrupt change – acceleration, deceleration, and reversal of direction. When the case suddenly changes direction, the rotor's own inertia makes it lag behind and then overswing, and it is precisely that lag, that relative motion against the case, which spins the generator fast enough to produce useful current. A wrist supplies this constantly and chaotically: every step you take, every reach for a coffee cup, is a small jolt in a new direction. It is messy, irregular motion, and the generator loves it.

A winder supplies the opposite. Its entire design goal is smoothness – a slow, even, repeatable turn that a mechanical movement finds ideal and a generator finds nearly worthless. The better-engineered and smoother the winder, the worse it is for a Kinetic. There is a certain irony in spending more on a winder to get a result that's actively counterproductive.

To be fair to the winder, the picture isn't quite a flat zero. Most winders reverse direction periodically to suit bidirectional automatics, and each reversal is a brief moment of exactly the kind of acceleration a Kinetic likes. Some owners who leave a Kinetic on an aggressive bidirectional winder for days do report a trace of charge accumulating. But a trace, eventually, if you leave it for days, is not charging in any useful sense. It is an accident of the winder occasionally doing the wrong thing slightly less wrongly.

So what does work?

Anything that delivers repeated changes of direction rather than smooth rotation. Wearing it, obviously. A brisk session of hand-shaking will wake a stopped Kinetic – Seiko's own manuals describe charging in terms of swinging the watch, not rotating it.

And a pendulum. A watch hung from a swinging arm changes direction at the end of every swing, which is exactly the acceleration event the generator needs, repeated indefinitely and for free. It's the cheapest honest solution to the problem, and I've written separately about how to charge a Seiko Kinetic with an $18 pendulum, including how long it takes from dead and the one thing you must get right when mounting the watch.

Seiko itself sidestepped the motion problem entirely. Its official Kinetic charger, the now-discontinued YT02A, used electromagnetic induction – an alternating field that induced current directly in the watch's generator coil, no spinning required; rather like an induction cooktop. It is the same principle behind the toothbrush-charger trick, though unreliably in the toothbrush case because the charger's coil isn't tuned to the Kinetic's. These are a genuinely different route to the same end: where a pendulum charges the watch mechanically by moving the rotor, an inductive charger charges it electrically by energising the coil. Both work. A rotary winder does neither.

Scroll to compare →

Method	Charges a Kinetic?	Why
Wearing it	Yes	Constant irregular motion with frequent changes of direction – exactly what the generator needs.
Pendulum	Yes	Reverses direction at the end of every swing, delivering the acceleration that spins the generator. Cheap and effective.
Hand-shaking	Yes, briefly	Rapid direction changes will wake a stopped watch, though it is tedious for a full charge.
Seiko's own charger (YT02A)	Yes	Charges inductively, energising the coil directly rather than moving the rotor. Long discontinued and now scarce.
Standard watch winder	Barely	Smooth continuous rotation turns the rotor far too slowly to induce a useful voltage.
Toothbrush charger	Sometimes	Inductive, like the YT02A, but only when the charger's coil happens to match the watch's. Unreliable and hit-or-miss.

But the headline is simpler than any of that. If you've been leaving your Kinetic on a watch winder and wondering why it keeps dying, now you know. The winder isn't broken, and neither is the watch. They were just never speaking the same language.

FAQs

Will a watch winder charge a Seiko Kinetic?

Not effectively. A standard winder rotates the watch smoothly and continuously, which is ideal for a mechanical automatic but nearly useless for a Kinetic. A Kinetic charges its generator through speed and changes of direction, not steady rotation, and on a smooth winder the rotor turns far too slowly to induce a useful voltage. A bidirectional winder may bank a trace of charge during its direction reversals, but not enough to rely on.

Why does a winder work for an automatic but not a Kinetic?

An automatic stores energy mechanically by winding a mainspring, and winding is cumulative - even slow rotation adds tension over time. A Kinetic stores energy electrically, and its generator's output depends on how fast the rotor spins, not how far. Slow, smooth winder rotation produces little or no current because the generated voltage stays below the charging circuit's threshold.

What kind of motion does a Seiko Kinetic need to charge?

Acceleration and repeated changes of direction, not smooth rotation. When the watch suddenly changes direction, the rotor's inertia makes it lag and overswing, spinning the generator fast enough to produce useful current. A wrist provides this constantly; a smooth winder does not. A swinging pendulum does, which is why it works where a winder fails.

How does Seiko's own Kinetic charger work?

Seiko's official charger, the discontinued YT02A, uses electromagnetic induction rather than motion. It generates an alternating magnetic field that induces current directly in the watch's generator coil, charging the storage cell without spinning the rotor at all, on the same principle as an induction cooktop. This is a different route from mechanical charging, and the same principle behind the electric-toothbrush-charger trick.

How fast does a Seiko Kinetic generator spin?

Very fast. Figures of up to 100,000 rpm are cited in secondary sources, achieved through a step-up gear train from the slowly turning rotor, though Seiko's own published technical pages do not commit to a precise number. The important point is that the generator depends on high rotational speed, which is why slow winder rotation produces so little.

Does a Seiko Kinetic use a battery or a capacitor?

It depends on the movement. Early Kinetic movements store charge in a capacitor; later movements use a rechargeable cell. Seiko changed the storage component across the range over the years, so the exact type depends on the specific calibre. Seiko refers to it generically as the Kinetic Electricity Storage Unit. In all cases it is recharged by the wearer's motion rather than being a disposable battery.

On sources. The mechanism and history described here were reconciled against primary documentation rather than forum recollection. Seiko's own Kinetic technical pages (Seiko Watch Corporation, Kinetic) and Seiko Epson's corporate record of the original AGS movement (Seiko Epson, "Seiko AGS Quartz Watch") confirm the architecture, the 1988 launch, and the AGS-to-Kinetic history. Seiko's 2007 press release "Kinetic: 20 Years of Success" confirms the caliber history and the four-year reserve of Kinetic Auto Relay. The mainspring, reverser-gear and sliding-bridle mechanism of a conventional automatic is documented in the Wikipedia entry on the automatic watch; the capacitor-or-rechargeable-cell distinction in the entry on automatic quartz. The "up to 100,000 rpm" generator figure is widely cited across watch literature but is not stated in Seiko's own published technical pages, which decline to give a per-caliber figure; it is presented here as a widely-reported approximation, not a Seiko specification.

How to Make Hong Kong Milk Tea (Authentic Recipe)

noreply@blogger.com (Unknown) — Sun, 05 Jul 2026 04:09:20 +0000

Prep: 2 min · Cook: 9 min · Total: about 11 min · Makes: about 5 mugs (1.2 litres of tea) · Keeps: 2 days refrigerated

Most Hong Kong milk tea recipes online have one of two problems. Either they measure the tea by volume – cups and tablespoons – which is unreliable for tea, where a few grams genuinely change the cup, or they're tangled up in a translation error that sends people looking for the wrong ingredient.

Here's the one that trips everyone up:

Hong Kong milk tea is made with black tea. The Chinese name, 紅茶 (Cantonese: hùhng chàh, Mandarin: hóngchá), translates literally as "red tea" – and a lot of recipes repeat that word. But "red tea" in English usually means rooibos, which is something else entirely. What 紅茶 actually means is what English calls black tea. So when a recipe says red tea, reach for a strong black tea – specifically a Ceylon-based blend.

I grew up in Hong Kong drinking this tea, and the version below is the one I make at home. There's no single "correct" recipe – every cha chaan teng guards its own blend. This is mine. It works, and it reminds me of home. It's in grams for the tea (where it matters), with US measures in brackets, and spoons for the milk. It's also far simpler than the long recipes make it look – boil, strain, mix.

→ Jump straight to the recipe

The Recipe

Hong Kong Milk Tea

Prep 2 min Cook 9 min Total ~11 min Makes ~5 mugs (1.2 L) Keeps 2 days

Ingredients

The tea blend (makes ~5 mugs / 1.2 L)

1,200 ml (5 cups) water
15 g (0.5 oz) Ceylon black tea
10 g (0.35 oz) Assam
4 g (0.15 oz) bo lei (普洱, Pu'er)

Per mug (150 ml tea)

2 tsp condensed milk (sweetness)
50 ml (1¾ fl oz) evaporated milk (body – up to 75 ml / 2½ fl oz if you like it milkier)
No separate sugar – the condensed milk does that job.

Method

Boil the water. Bring all 1,200 ml (5 cups) to a boil.
Add the tea and boil for 7 minutes. A full rolling boil – not a gentle steep. It'll taste harsh on its own. That's correct.
Rest 2 minutes. Turn off the heat. It settles and keeps extracting off the boil.
Strain through a cloth sock. A single strain is fine. Pull it back and forth a few times if you want it smoother.
Assemble each mug. Fill ~halfway with tea (150 ml), add 50 ml evaporated (up to 75 ml for milkier) and 2 tsp condensed milk, stir. Leave a little headroom.

Cold or iced? See the cold & iced versions below.

What you'll need (the detail)

Why three teas? Ceylon gives the bright backbone, Assam adds body and malt, and a little bo lei rounds it out with an earthy depth. Most recipes reach for English Breakfast tea bags – that's not what gives it the cha chaan teng character. Most also skip the bo lei, which is a small thing that makes it taste properly Hong Kong rather than just "strong black tea." (Cha chaan tengs typically use finely broken "tea dust" for a fast, hard extraction. Loose leaf works just as well at home, with a cleaner pour.) And here's the thing most recipes get wrong: measure your tea by weight, not volume. A tablespoon of fluffy whole leaf and a tablespoon of dense fannings can differ in weight by half – which is exactly why so many homemade cups come out randomly strong or weak. Grams fix that.

A note on cheap bo lei (普洱). If you're using a very low-grade bo lei, give it a quick rinse first: pour boiling water over just the bo lei, swirl for a few seconds, and tip that water out before you brew. It washes off the musty note that cheap Pu'er can have. Good bo lei doesn't need this – skip it if yours smells clean.

Per mug (a mug holds roughly 150 ml of tea):

2 teaspoons condensed milk (sweetness) – a little more if you're serving it over ice
about 50–75 ml (1¾–2½ fl oz) evaporated milk (body) – start at the lower end and add more to taste
No separate sugar – the condensed milk does that job.

A note on the milk. This is the part recipes muddle most. The two milks do different jobs: evaporated milk gives the creamy body, condensed milk gives the sweetness. You don't need both and sugar – the condensed milk is the sugar. In Hong Kong the default is Black & White (黑白淡奶), the Dutch evaporated milk brand the cafés swear by. It's not richer than other evaporated milk – it's just the cha chaan teng standard, prized for its mouthfeel and faint caramel note.

Cha chaan tengs usually only use evaporated milk in their preparation, allowing the customer to add sugar as they please. This home recipe uses both for practical reasons: at home, with loose leaf rather than the café's fine tea dust, and without a commercial urn holding the brew at temperature, the cup comes out slightly less bodied than the café version. Condensed milk's caramelised milk solids restore that body. It also dissolves cleanly into hot milked tea, where granulated sugar tends to crystallise at the bottom of the mug. If you'd rather stay closer to the café original, use evaporated milk only and add sugar to taste (about 2 teaspoons per mug, stirred into the tea before the milk goes in).

And a warning: don't reach for the fresh milk in your fridge door. Some recipes suggest regular milk, and I've tried it – even full-cream is too thin and watery to stand up to a brew this strong, and the result tastes weak and washed out. Evaporated milk is concentrated and richer, which is the whole point: it has the body to match the tea.

Equipment: A cloth "sock" strainer – the fine-mesh cloth bag used for tea across Asia, from Hong Kong to Bangkok. It's a few dollars at any Asian supermarket. (This is the "silk stocking" the tea is named after.)

Method

Boil the water. Bring all 1,200 ml (5 cups) to a boil.
Add the tea and boil for 7 minutes. Yes, a full rolling boil – this isn't gentle steeping. It pulls out a lot of tannin, and the tea will taste harsh and bitter on its own. That's correct. The bitterness is the point. The milk balances it. If you've made HK milk tea before and it came out "too bitter," you weren't doing it wrong – you just hadn't added enough milk.
Turn off the heat and let it rest for 2 minutes. It keeps extracting gently off the boil and settles before straining.
Strain through the cloth sock. A single strain works perfectly – it's what I do at home. If you want to go further, you can "pull" it: pour the tea back and forth between the sock and a pot several times. This aerates it and smooths the tannin. It's not necessary, but it does make it better.
Assemble each mug. Fill the mug about halfway with the hot tea (roughly 150 ml), then add about 50–75 ml (1¾–2½ fl oz) evaporated milk and 2 teaspoons condensed milk, and stir. The milk makes up the rest of the mug – you'll have a little headroom at the top, not a brimful cup. Adjust to taste: start with the lower amount of milk, then add more until it's the colour you like. (I make this by eye, so treat the numbers as a reliable starting point rather than gospel.)
Serve with your favourite Hong Kong dish or pastry. A sausage bun or a pineapple bun from Maria's Bakery or Maxim's is great for tea.

Diabetic or sugar-free version. Skip the condensed milk and sweeten with 1–2 sachets of Equal (or your sweetener of choice) per glass. Because condensed milk also adds a little body, add a touch more evaporated milk to keep it creamy.

Why your homemade Hong Kong milk tea tastes wrong

If you've tried this before and it didn't taste right, it's almost always one of these:

You used fresh milk. The most common mistake. Fresh milk – even full-cream – is too thin for a brew this strong. Use evaporated milk for body and condensed milk for sweetness.
You didn't use enough tea. This is a strong brew by design. A couple of teabags in a mug will never get there.
You brewed it too gently. It needs a hard boil, not a delicate steep. The tannin you pull out is what the milk balances against.
You thought the bitterness meant you'd failed. The unmilked tea should taste harsh. That's correct. Fix balance with more milk, not weaker tea.
You went looking for "red tea." It's black tea – see the FAQ below.

A little background

Hong Kong milk tea is the everyday drink of the cha chaan teng – the fast, cheap "tea restaurants" that have fed Hong Kong since the 1950s, serving Western-influenced comfort food at local speed. The tea is the backbone of the place: strong, smooth and a little bracing, built to be drunk all day and to cut through a plate of macaroni or a pineapple bun. It runs on two things – serious caffeine and condensed milk. Its nickname, "silk stocking tea" (絲襪奶茶), comes from the fine cloth sock it's strained through, which over a day of use stains the colour of a stocking.

Cold and iced Hong Kong milk tea

Cold (straight from the fridge) is how I most often drink it in summer – and it's where the make-ahead concentrate really earns its keep. The proportions are exactly the same as the hot cup, per glass:

about 150 ml of the cold tea concentrate
50–75 ml evaporated milk
2 teaspoons condensed milk

Pour the milk over the cold concentrate, stir, and drink it as is – no ice needed. Because the concentrate is already cold and strong from its night in the fridge, it doesn't get watered down, so the balance stays identical to the hot version. This is genuinely my everyday summer cup.

Over ice, one thing changes: use 2–3 teaspoons of condensed milk instead of two. As the ice melts it waters the glass down, and cold blunts sweetness, so it needs a touch more to hold its balance. Everything else stays the same – 150 ml tea, 50–75 ml evaporated milk – poured over a tall glass of ice.

Make it ahead – and it actually improves

Here's a trick the recipes don't tell you: the brewed tea keeps, and gets better. I make the full batch, serve some fresh on the day, then strain the rest and keep it in the fridge. The next day the flavour has deepened – and it's perfect for the cold version above.

Storage: the strained, unmilked concentrate keeps in the fridge for about two days, and the flavour deepens overnight. Mix with milk fresh each time rather than storing it pre-mixed.

One warning: this is strong stuff. A mug made from a brew this concentrated carries roughly the caffeine of two to three cups of ordinary tea. It will keep you up – which is rather the point in a cha chaan teng, but worth knowing before a second mug at dinner.

Enjoyed this? We have much more on Hong Kong's history. From KPS Video Express to Maria's Bakery, we cover topics which would otherwise be forgotten.

FAQ

Hong Kong milk tea recipes call for "red tea" – what is that?
It's just black tea. The confusion is a translation artefact. The Chinese name is 紅茶 (Cantonese: hùhng chàh, Mandarin: hóngchá), which literally means "red tea" – named for the reddish colour of the brewed liquid, not the leaf. In English, though, 紅茶 is simply black tea. Confusingly, "red tea" in English usually means rooibos, which is unrelated and not used here. So when a recipe says "red tea," reach for a strong black tea – a Ceylon-based blend.

What makes Hong Kong milk tea taste different?
Three things: a strong blend of black teas (here Ceylon, Assam and a little bo lei / Pu'er) brewed hard and boiled rather than gently steeped, evaporated milk for a rich body that thin fresh milk can't match, and condensed milk for sweetness. The strength and the evaporated milk are what give it the distinctive cha chaan teng character.

How do you make Hong Kong-style milk tea at home?
Boil 1,200 ml water with 15 g Ceylon, 10 g Assam and 4 g bo lei (普洱, Pu'er) for 7 minutes, rest 2 minutes, then strain through a cloth sock. Per mug, fill about halfway with the tea, add roughly 50–75 ml evaporated milk for body and 2 teaspoons condensed milk for sweetness (a little more if serving over ice). No separate sugar.

What tea is used in Hong Kong milk tea?
A strong black tea, usually a Ceylon-based blend. This recipe uses Ceylon, Assam and a little bo lei (普洱, Pu'er). Most blog recipes reach for English Breakfast tea bags, which won't give the same depth.

Can I use Earl Grey, English Breakfast, or another flavoured tea?
No – and this is where a lot of recipes go wrong. Earl Grey is scented with bergamot, which is a citrus flavour that fights the malty, milky profile you're after. It tastes distinctly "off" once the milk goes in. English Breakfast is closer, since it's usually an Assam-Ceylon blend, but it's built to be drunk on its own with a splash of milk, not to stand up to evaporated milk in a hard-boiled brew – the result is thin and one-dimensional. Hong Kong milk tea wants unflavoured black tea with real strength and body: a Ceylon base, Assam for malt, and a little bo lei for depth. Anything perfumed or lightly-blended pulls it away from the cha chaan teng taste.

Can I use tea bags for Hong Kong milk tea?
Loose-leaf is what makes it taste right – tea bags are only a fallback. They hold a small amount of lower-grade tea, so you'll need a lot of them and it still won't have the depth of a proper loose blend. If bags are all you have, use strong black tea bags (Ceylon or English Breakfast), at least one per 150 ml, and boil hard. But loose Ceylon-Assam is what makes it taste like the real thing.

Why do some Hong Kong milk tea recipes ask for "tea dust" or "orange pekoe"?
Both are just grades of black tea, not different teas. "Tea dust" (or "fannings") means finely broken tea – small particles that brew fast and strong, which is exactly why Hong Kong cafés use it. "Orange pekoe" is a whole-leaf grade. Despite the name it has nothing to do with oranges or flavour (it most likely refers to the Dutch House of Orange from the early tea trade). Either works at home. Café-style dust just extracts harder and faster. A quality loose-leaf black tea like the Ceylon-Assam blend here gives you the same result with a cleaner pour.

Why does my Pu'er or milk tea smell musty?
Cheap, low-grade bo lei (普洱, Pu'er) can carry a musty note. The fix is a quick rinse: pour boiling water over just the bo lei, swirl for a few seconds, and tip that water out before you brew. It washes off the musty note that cheap Pu'er can have. Good bo lei doesn't need this – skip it if yours smells clean.

How much milk do you put in Hong Kong milk tea?
Per mug of roughly 150 ml tea: about 50–75 ml evaporated milk for body and 2 teaspoons condensed milk for sweetness. Cold from the fridge it's the same proportions. Over ice, bump the condensed milk to 2–3 teaspoons to make up for the melting ice. Recipes vary a lot here, so treat it as a starting point and adjust. The key is evaporated milk for body, condensed for sweetness.

Should I use sugar or condensed milk in Hong Kong milk tea?
In a sit-in cha chaan teng, the standard preparation uses evaporated milk with white sugar left on the table for the customer to add. For the home brewer using loose leaves rather than the café's fine tea dust, combining evaporated milk for body and condensed milk for sweetness gives the most consistent, velvety cup.

Why is my Hong Kong milk tea weak or watery?
Almost always because you didn't use enough tea or didn't brew it hard enough. This is a strong, concentrated brew – boiled hard, not gently steeped – and it needs to stand up to rich evaporated milk. If it tastes thin, use more tea and boil it harder, and make sure you're using evaporated milk (not fresh milk, which is too watery).

Is Hong Kong milk tea the same as yuanyang?
No. Yuanyang (鴛鴦) is a Hong Kong drink made by mixing milk tea with coffee. Hong Kong milk tea is just the tea-and-milk base on its own. Yuanyang uses this milk tea as one of its two halves.

Where did Hong Kong milk tea come from?
It's a legacy of British colonial rule. Hongkongers wanted the milky tea the British drank, but fresh milk was a scarce luxury – so they used canned evaporated milk instead, and brewed the tea far stronger to suit local taste. That's how the cha chaan teng version was born in the 1940s and 50s.

Is Hong Kong milk tea the same as Thai milk tea?
No. Both use strong black tea (often Assam) and sweetened condensed milk, but Thai milk tea – the familiar bright-orange version, usually Cha Tra Mue – is sweeter, has an added flavouring, and gets its colour from food colouring (FD&C Yellow No. 6). Hong Kong milk tea has no flavouring or colouring: it's just tea and milk, stronger and brown.

Is Hong Kong milk tea the same as Malaysian or Singaporean teh tarik?
They're close cousins. Teh tarik ("pulled tea") uses strong black tea with sweetened condensed milk, and the dramatic pouring back and forth – the "pull" – is the same technique that gives Hong Kong's "silk stocking tea" its smoothness. The main difference is that teh tarik is typically sweeter and frothier from all the pulling, while Hong Kong milk tea is stronger and less sweet. Same family, different dialect.

Why is my Hong Kong milk tea bitter?
Because it's meant to be brewed strong and boiled hard – the tea is bitter on its own by design. The milk balances it. If it's too bitter, add more condensed and evaporated milk rather than weakening the tea.

What milk is used in Hong Kong milk tea?
Evaporated milk for body and condensed milk for sweetness – traditionally Black & White (黑白淡奶), the Dutch evaporated milk brand Hong Kong cafés swear by. No separate sugar.

Can I use regular milk instead of evaporated milk?
No – or rather, you can, but it won't taste right. Even full-cream fresh milk is too thin to balance such a strong brew, and the result is weak and watery. Evaporated milk is concentrated and richer, which is exactly what Hong Kong milk tea needs. This is the single most common reason a homemade version tastes "off."

Can I make Hong Kong milk tea without sugar?
Yes. Skip the condensed milk and sweeten with a sweetener like Equal (1–2 sachets per glass), then add a little extra evaporated milk to keep the body. This makes a diabetic-friendly version without weakening the tea.

What do you strain Hong Kong milk tea with?
A cloth "sock" strainer – a fine cloth bag, the same kind used across Asia, from Hong Kong to Singapore to Bangkok. It's what gives "silk stocking tea" its name.

Can you make Hong Kong milk tea ahead of time?
Yes. Brew the concentrate, strain it, and keep it in the fridge for a day or two. The flavour deepens, and it's excellent for the cold version.

The 5-Whys: A Guide to Root-Cause Analysis

noreply@blogger.com (Unknown) — Thu, 25 Jun 2026 12:36:01 +0000

The 5-Whys is Toyota's method for finding the root cause of a problem. The technique is exactly what it sounds like: state the problem, ask why it happened, ask why about the answer, and repeat until you reach a cause that is structural rather than symptomatic. Five iterations is the convention. Sometimes three is enough, occasionally seven is needed, but five is where most problems yield.

The method exists because most people, when confronted with a problem, fix the first thing they see. The caterer was late, so switch caterers. The server crashed, so restart the server. The report was wrong, so correct the report. Each of these fixes addresses the symptom rather than the underlying condition that produced it, which is why the same problems keep surfacing in slightly different forms. Fixing problems at the surface is like cutting weeds – they grow back. Fixing them at the root cause is like pulling them out entirely.

The 5-Whys provides the discipline to reach the root before acting.

How It Works

The method is straightforward, which is part of its power and part of its danger.

Step 1: State the problem clearly. Be specific. "Our delivery was late" is better than "logistics isn't working."

Step 2: Ask: why did this happen? Note the answer.

Step 3: Ask: why did *that* happen? Note the answer.

Step 4: Repeat until you have asked "why" five times, or until the answer you reach is a structural cause that, if fixed, would prevent recurrence.

Step 5: Formulate countermeasures to address the root cause. The emphasis is on prevention, not correction.

A worked example makes the method concrete. Imagine a team that has just organised an event where the catering arrived two hours late:

Why	Question	Answer
1	Why was the food two hours late?	The purchase order was issued only three days before the event.
2	Why was the PO issued so late?	The approval signatures took two weeks to gather.
3	Why did the signatures take so long?	The primary approver was on leave, with no documented backup.
4	Why was there no backup approver?	Nobody had set one up – the issue had never surfaced before.
5	Why hadn't it surfaced?	There was no checklist for time‑critical procurement tasks, so the leave‑cover gap was invisible.

The root cause is not the caterer, not the approver, and not the leave policy. It's the absence of a checklist for time-critical tasks. Switching caterers – the instinctive fix – would have changed nothing. Installing the checklist prevents recurrence permanently.

When To Ask More (or Fewer) Than Five

Five is a convention, not a rule. In practice:

Fewer than five usually means the analysis hasn't gone deep enough. If your root cause is "the supplier was bad" at iteration two, you've almost certainly stopped at a symptom.

More than five sometimes happens on complex problems, but going past seven or eight tends to produce diminishing returns and analyst fatigue without proportional benefit.

Branching is common. A single "why" can produce two or three answers, each of which becomes its own line of inquiry. The table method and the fishbone (Ishikawa) diagram are two ways to visualise branching analyses and keep them manageable.

Who Should Run the Analysis

This is the single most important implementation detail and the one most often ignored: the 5-Whys should be run by the people closest to the problem, not by management at a distance.

In Toyota, the analysis is almost always conducted by the subject-matter expert on the ground – the person with firsthand knowledge of what happened. Analysis conducted at a distance, by someone reading a report rather than having lived the situation, will produce educated guesses. Countermeasures based on inaccurate root causes will fail, and the problem will return.

The instinct in many organisations is to conduct root-cause analysis in a meeting room, with senior people reviewing incident reports. This is structurally backwards. The people in the meeting room are furthest from the actual problem, while the people on the ground, those who know exactly what happened and why, are absent from the analysis. Toyota's approach inverts this: trust your people to provide the root-cause analysis, and use management's role to remove barriers to implementation rather than to diagnose.

Known Weaknesses

The 5-Whys is qualitative, not quantitative, and its conclusions depend on the perspective and experience of the analyst. This has several implications:

Bias: Two analysts examining the same problem can reach different root causes. Someone predisposed to blame suppliers will find supplier issues. Someone predisposed to blame process will find process issues. Both may be partially right.

Subjectivity: There is no objective "correct" answer to any given "why". The method relies on judgment. High-stakes problems, where political incentives exist to skew the conclusion, are particularly vulnerable.

The remedy is group work: Toyota's standard approach is to have multiple subject-matter experts produce their own analyses independently, then reconvene to present findings and discuss. The group prioritises fixable root causes collectively, reducing the impact of any one person's bias. A deeper discussion of these limitations – and how to compensate for them – is covered in 5-Whys Weaknesses.

Common Mistakes

Stopping at blame: "The supplier failed" is not a root cause. It's an attribution. The question is always: why did the system allow the supplier to fail without a fallback?

Skipping straight to solutions: Teams under pressure often abandon the analysis at iteration two or three because they've already thought of a fix. The fix may work. It may also be a surface fix that addresses a symptom. The discipline is to complete the analysis before acting.

Running it top-down: When management conducts the analysis without the people who experienced the problem, the results are abstractions. The answers to each "why" become what management *assumes* happened rather than what *actually* happened.

Never acting on the output: The most common failure of all. The analysis is completed, the root cause is documented, and the countermeasure is never implemented because the organisation has already moved on to the next crisis.

The Toyota Production System Series

This article is part of a 40+ article series on the Toyota Production System, drawn from years working inside Toyota. For the distilled version of the whole system, see The Kaizen Roadmap, a free briefing covering Kaizen, PDCA, Horenso, the 5-Whys, Mieruka, Genchi Genbutsu, and 5S, with case studies and a diagnostic.

How to Charge a Seiko Kinetic Watch Without Wearing It - Cheaply and Reliably

noreply@blogger.com (Unknown) — Sat, 13 Jun 2026 11:28:15 +0000

Seiko's instructions on charging a Seiko Kinetic watch are simple: wear it.

But when you're cycling through a collection, or an Apple Watch has displaced the traditional wristwatch entirely, that advice is rather difficult to follow. The Kinetic sits in a drawer, the charge drains, and eventually it stops.

The forum answer to this is almost universally: use an electric toothbrush charger. I tried it. It didn't work. It made the watch case uncomfortably hot, and the charge never took. It turns out this method only works with specific charger models (particular older Braun/Oral-B units are the ones most commonly cited on forums), which are increasingly hard to find and priced up on the secondary market when they surface. For many people who try this, these chargers do nothing except warm the watch.

After that, the alternatives get expensive fast. Seiko's own Kinetic chargers were never sold to the public and appear occasionally on eBay for hundreds of dollars. Modified Kinetic-specific winders start around US$100. A standard watch winder does almost nothing, and the reason why is the key to charging a Kinetic without wearing it.

The solution that worked for me was a pendulum. At US$18, it charged my 2011 dress watch, a 5M62 movement that had been sitting for months, in around 12-15 hours. No modifications, no specialty equipment, no hundred-dollar winder.

Before landing on the pendulum, I went through the usual alternatives. None were great.

Seiko Kinetic charger alternatives (and why most don't work)

Seiko's own Kinetic chargers were never released to the public, though they surface occasionally on the secondary market for hundreds of dollars. The YT02A is the model that appears most often. At that price, you're paying more for the charger than most Kinetics are worth.

Electric toothbrush chargers are the method every forum thread eventually lands on, but they're unreliable. Repurposing a wireless toothbrush charger to inductively charge a Kinetic only works if the charger's coil alignment and output happen to match the Kinetic's generator. For most people, most chargers simply doesn't work, and you can spend a week testing positions before giving up.

Standard watch winders do almost nothing for a Kinetic. They're built for automatic mechanical watches, where smooth continuous rotation winds a mainspring. A Kinetic needs something fundamentally different, and the reason why is the key to charging one without wearing it.

Modified Kinetic-specific winders start at around US$100 and are engineered differently from standard winders. They work, but the price is hard to justify for occasional use, especially when a pendulum does the same job for a fraction of the cost.

Why a pendulum charges a Seiko Kinetic

The Seiko Kinetic generates electricity through electromagnetic induction. A weighted rotor spins past a set of coils, and that motion produces current, which is stored to power the watch – in a capacitor on early Kinetic movements, or a rechargeable cell on later ones.

The key word is relative. What matters is the rotor moving independently of the case, not the whole watch moving through space. This is why a standard winder doesn't work: smooth, continuous rotation carries the rotor with the case rather than against it. The rotor just sits there, turning in sync, generating nothing.

What the movement actually needs is oscillation: the same erratic, back-and-forth motion your wrist produces while walking. That acceleration and change of direction is what causes the rotor to spin freely relative to the case, generating current.

A pendulum replicates this cheaply and effectively. The watch changes direction with each swing, the rotor lags behind due to inertia, and current flows. The mechanism doesn't know or care whether it's on a wrist or a pendulum. It only responds to the physics.

Seiko's own manuals describe charging as dependent on wrist motion with direction changes. A pendulum provides exactly that. Kinetic-specific winders work on the same principle, which is why they're engineered differently from standard automatic winders, and priced accordingly. (For the full physics of why an ordinary winder barely charges a Kinetic at all, see why a watch winder won't charge a Seiko Kinetic.)

How to charge a Seiko Kinetic with a pendulum

The process is straightforward:

Get a pendulum. I used a pendulum watch winder available for around US$18 on AliExpress. The key requirement is a swing arm long enough to produce meaningful direction changes, and a way to secure the watch at the end. Examples of the one I bought: Store 1 and Store 2
Secure the watch. Place or strap the watch at the top of the pendulum. It won't charge if secured at the bottom. My kit came with a low tech solution: rubber bands. The crown should face outward. The watch needs to be firmly attached - you don't want it flying off mid-swing.
Let it swing. Set the pendulum in motion and leave it. For a completely dead Kinetic, expect 12-15 hours to reach a full charge. A partially depleted watch will take less.
Check the charge indicator. If your model has a charge indicator (press the crown on most Kinetics), verify the seconds hand jumps in two-second intervals (low) or one-second intervals (charged). On the 5M62, a full charge provides approximately 6 months of power reserve.

Will a pendulum damage a Seiko Kinetic?

The rotor pivot and mainspring-equivalent components in a Kinetic aren't designed for continuous mechanical stress. A few sessions a year when the watch has been sitting is fine - that's well within the kind of use the movement was built for. Running it on a pendulum daily or indefinitely is a different proposition, and probably not one Seiko engineered for.

Use it to wake the watch up. Don't use it as a permanent charging solution.

A dead Kinetic isn't broken

A Kinetic that's stopped isn't a Kinetic that's failed. The storage cell has simply drained below the threshold needed to drive the movement. The charge system is intact. It just needs the right kind of motion to start generating current again.

At $18, a pendulum is a more dignified solution than the alternatives: cheaper than a winder, more reliable than a toothbrush charger, and considerably more practical than Seiko's own advice to simply wear it more.

Available on AliExpress for around US$18, or Amazon at a premium.

*I used this method on a 5M62 movement. If you've tried it on other Kinetic movements (5M42, 5M43, 5M65, 5J21, 5J32, etc.), I'd be interested to hear whether it worked.

FAQs

How do you quickly charge a Seiko Kinetic?

The fastest method without Seiko's own charger is a pendulum. Place the watch on a swinging pendulum and leave it for 12-15 hours for a full charge from dead. The back-and-forth motion replicates the wrist movement the Kinetic's rotor needs to generate current. For a partial top-up, a few hours is usually enough.

Will a watch winder charge a Seiko Kinetic?

A standard watch winder does very little for a Kinetic. Winders are designed for mechanical automatics, where smooth continuous rotation winds a mainspring. A Kinetic generates electricity through changes in direction, not continuous motion. The rotor needs to move independently of the case, and a smooth-spinning winder carries the rotor in sync with the case, generating almost no current. Kinetic-specific winders exist but are engineered differently and start at around US$100.

How to charge a Seiko Kinetic with a toothbrush charger?

Some electric toothbrush chargers (particularly certain Braun/Oral-B models like the Type 3757) can inductively charge a Kinetic by placing the watch case back on the charger. However, compatibility depends on the charger's coil frequency and alignment matching the Kinetic's generator. Many chargers don't work at all. A pendulum is a more reliable and universal method.

How long does a Seiko Kinetic charge last?

It depends on the movement. A fully charged 5M62 holds approximately 6 months of power reserve. Older movements like the 5M42 and 5M43 hold less. If the watch is sitting unworn, the charge will gradually drain. Once depleted, the watch stops but the movement isn't damaged, it just needs to be recharged.

Will a pendulum damage a Seiko Kinetic watch?

Occasional use is fine. The rotor and its pivot mechanism are designed for the kind of intermittent acceleration a wrist produces. A few charging sessions a year when the watch has been sitting idle is well within normal use. Running a Kinetic on a pendulum continuously, day after day, would put sustained mechanical stress on components not designed for it and is not recommended.

Why doesn't the electric toothbrush charger method work for most Seiko Kinetics?

The toothbrush charger method relies on the charger's inductive coil aligning with the Kinetic's internal generator coil at a compatible frequency. Only specific charger models — most commonly certain older Braun/Oral-B units — happen to match. Most modern toothbrush chargers use different coil geometries or frequencies and produce nothing useful, or worse, generate heat in the watch case without transferring charge. If your watch is warming up on the charger without gaining any power reserve, the coils aren't matched and no amount of waiting will fix it.

Why did my Seiko Kinetic get hot on the toothbrush charger?

Heat without charge means the charger's coil is inducing eddy currents in the watch case metal instead of driving the Kinetic's generator. This happens when the coil frequency and geometry don't match what the Kinetic is designed to receive. The energy dissipates as heat in the case rather than flowing into the charge system. It's not damaging in a brief attempt, but it's a sign the method isn't going to work with that charger — and prolonged sessions probably aren't good for the watch. Stop and try a different approach.

Is the "toothbrush charger" advice for Seiko Kinetics reliable?

Not really. The advice has been repeated on forums for years, but it works only for a narrow set of specific charger models — most of them older Braun/Oral-B units — and even then it depends on physical placement of the watch. Many people trying it find the watch heats up but doesn't charge, or find that their charger doesn't induce a current at all. The right chargers are increasingly hard to find and priced accordingly on the secondary market. A pendulum works universally because it uses the watch's own generator mechanism directly — same principle as wrist motion.

The Station That Knew: On Governance That Cannot Read Signals

noreply@blogger.com (Unknown) — Sun, 07 Jun 2026 13:25:31 +0000

Monday morning. Kishibe Station, a local train station on the JR line close to Osaka. 25 April 2005. The cycle in was eventless. Cold air, clear skies, the ojisan at the bicycle parking was chipper as usual. The ticket gate clipped the ticket at speed. The stationmaster’s face, as I collected my ticket, was different. A scowl, quickly suppressed, the opposite of his usual greeting.

On the platform, an announcement apologised for the delay into Osaka. The platform display showed eight minutes late. Then fifteen. Then twenty-two.

When the train finally arrived, the rollsign on the front read Osaka. Trains from this station did not terminate at Osaka. Ever.

Standing on that platform, I had sensed a signal. I did not yet understand what the signal was. I wouldn't until that evening. Earlier that morning at 9.19 am, 107 people died near Amagasaki Station, thirteen kilometres from where I stood. A rapid commuter train had entered a 70 km/h curve at 116 km/h, and derailed into an apartment building, wrapping cars around the structure. It remains one of the worst post-war rail disasters in Japan.

The cascading failures that were ignored

The investigation took two years. The driver, Ryujiro Takami, running 80 seconds late after overshooting Itami Station earlier in the run, had been distracted in the final minutes by an internal call to his conductor. In Japan, an 80-second deficit is the difference between catching a connecting train and being late to work. It also meant a razor margin for error: running closer to speed limits, taking station stops shorter than designed.

Takami was pleading with the conductor to under-report the size of the overrun to transport command to avoid nikkin kyōiku (日勤教育). This was the company’s programme of ritual humiliation for drivers who made errors. Drivers subject to it copied the rulebook by hand, wrote reflection essays that were rejected and rewritten, performed manual labour around the depot, and were placed in visible positions where colleagues could see them. Programmes ran for days. Drivers feared nikkin kyōiku more than rule bending, because the psychological and career damage was clear and visible. The consequences of rule-bending were probabilistic. Nikkin kyōiku was not.

The official report described it as a form of mental-attitude education (精神論的な教育, seishinron-teki na kyōiku), a phrase that carries pre-war militarist resonance. Drivers who erred and reported accurately faced nikkin kyōiku. Drivers who under-reported and were caught faced worse. The bet, for most, was that they would not be caught. Takami was desperate to avoid it. His attention was on the call. He failed to brake.

The two governance systems

Takami was 23. He had been subjected to nikkin kyōiku before. The decisions he made in those final minutes were locally rational, given the system around him. Locally rational does not mean exonerated. It means the object of examination is the system that produced the rationality, not the driver who acted inside it.

Every organisation runs two governance systems: the one written in policy, and the one enforced by incentive and consequence. Under pressure, when they diverge, the enforced one wins.

On paper, JR West had a coherent governance system. In practice, the disciplinary regime ran the operation. The proclaimed policy was "safety first", but the enforced policy was the timetable, backed by nikkin kyōiku. Safety said to slow down. Practice rewarded speeding up.

Signals existed but were ignored. Investigators found timetable pressures built up over the years, across the network. Near misses happened. They were logged. Nothing was done. After the crash, drivers came forward with incidents they had never disclosed – including one driver who had previously taken the exact same curve at 85 km/h, against a 70 km/h limit. He had never reported it. The reason he gave was the certainty of nikkin kyōiku. The commission's own conclusion was sharper: the company's method of grasping incidents – requiring drivers to report them, then punishing the reporters – risked, in its words, inducing the very accidents the system was meant to prevent.

Drivers kept crossing the limits of what was acceptable. What was unacceptable a year earlier became normal a year later – what Diane Vaughan called the normalisation of deviance. Reading the signals required a mechanism JR West did not have, and that few commercial operators do.

A different industry, the same architecture

During the same period, in a different industry on the other side of the world, the Mid-Staffordshire NHS Foundation Trust was running a failure of the same kind. The Department of Health had set aggressive bed-management targets. Over years, the trust's management enforced them with career-ending consequences for senior staff who missed them. The proclaimed priority was patient care. The enforced priority was throughput.

The targets had teeth. Failure to meet them meant loss of Foundation Trust status, board reorganisation, and dismissal of senior staff. The trust was being measured on throughput and rewarded on throughput. The mechanism was similar to JR West's: a metric that mattered, enforced through career consequence, applied without regard to the operational reality it was distorting.

The 2013 Francis Inquiry documented standards of basic care that had deteriorated: patients dehydrated, calls for help unanswered, people left in their own waste. Some died in conditions that were visible to the clinical staff around them.

Staff who raised concerns were not just ignored. Helene Donnelly, a nurse in Accident & Emergency (A&E), reported falsified records and was warned to “watch her back” in the car park. Doctors who raised concerns about staffing were treated as the problem. The trust had formal governance – board reporting, clinical audit, regulatory oversight. Somehow, those governing did not see what frontline staff did.

The mechanisms differ from JR West. Theirs was a designed disciplinary instrument, while Mid-Staffs’ was target pressure operating through career incentives. The instruments differ, but the blindness is the same.

What governance cannot see

What both organisations lacked was the ability to read weak signals – patterns which are below the threshold of action, but too persistent to be dismissed as noise. Investigation is a different function. Reading weak signals is different again. The information arrives ambiguous, partial, often contradictory, through both official and unofficial channels. Hypotheses only emerge when the information is interrogated into a pattern. These do not always pay off, but when they do, they matter disproportionately. The output is not a finding, but a question worth pursuing.

This is not a discipline that audit, as it is typically constituted, is designed to perform. The skill is different. The evidentiary standard is different. The temporal rhythm is different. Audit reports on what has happened, while weak-signal reading attends to what is happening. The reward structure is different: audit succeeds when findings are closed; weak-signal reading succeeds when patterns are detected before they require closure.

Karl Weick and Kathleen Sutcliffe’s research on high-reliability organisations describes how nuclear power plants, aircraft carriers, and air traffic control operate in environments where failure would be catastrophic, and yet maintain exceptionally low failure rates. They call this property mindful organising: the practice of treating small anomalies as information rather than noise. Five principles characterise it: preoccupation with failure, reluctance to simplify, sensitivity to operations, commitment to resilience, and deference to expertise.

What was missing

JR West exhibited the inverse of each. Near misses were treated as driver failings requiring re-education, not as system signals requiring investigation. The single explanation for lateness was bad drivers, and the single intervention was punitive retraining. Senior management had little direct contact with driver decision-making under time pressure. The system was designed to prevent lateness, not to recover from it. Drivers' operational judgement was systematically overridden by timetable pressure.

The signals were legible. The function to read them did not exist. Not because no one was paying attention, but because the organisation had no role whose mandate was to notice patterns, no authority to escalate them, no measure of success that rewarded detection over closure, and no protection for the people who escalated and were wrong. Equipment on JR West's own trains had been logging emergency-brake activations, roughly three times a day, for years. The commission found the company “could easily have known.” The data had been captured. No one had been tasked with reading it. One hundred and seven people died at a curve the system had been signalling for years.

Most governance functions lack what JR West lacked. Most have not had the failure that reveals the absence. When it comes – and at some point it will – the question is whether the function has been built, and whether the people who escalate face career consequences when their concerns prove unfounded.

The second question is the harder one. It requires distinguishing structurally, between honest pattern-reporting that turns out to be wrong, and reckless behaviour that warrants accountability. The first must be protected. The second must not. How to design that distinction is the subject of a companion piece: Reading the Room Before It's a Finding: A Diagnostic for Governance That Reads Signals.

[message]

Read more: Take a look at our ongoing series on Governance.
Related: The 2005 Amagasaki Rail Disaster: What the Official Report Reveals About Japan’s Deadliest Train Crash

What is Kaizen?

noreply@blogger.com (Unknown) — Sun, 31 May 2026 14:27:35 +0000

Kaizen (改善) is the Japanese word for continuous improvement, and it's the foundation of the Toyota Production System. The basic idea is simple: small changes, applied consistently over years, compound into something far larger than any single project could ever deliver. The hard part isn't understanding the concept – it's staying with it long enough for the compounding to actually happen, which is where most companies quietly give up.

I worked inside Toyota for several years, and what struck me was that Kaizen wasn't a slogan or an initiative anyone talked about as a thing – it was just how the place ran. People on the factory floor and in the offices spent part of every day looking at their own work, identifying small improvements, making the change, checking whether it worked, and standardising the result if it did. There was no programme office, no consultant, and no kick-off event. It was simply part of the job.

A practical example helps here. On Toyota's assembly lines, the workers themselves design their own tool carts – each cart holds every tool needed for the specific job that worker is doing, nothing more, arranged in the order they'll reach for them. The worker designing the cart knows the job better than any engineer or manager possibly could, because they're the one doing it 200 times a day. Each cart cuts a few seconds off every car. Toyota builds around 10 million cars a year, and the arithmetic is generous.

That, in practice, is Kaizen. Not a revolution. A single cart.

Why Kaizen Looks Slow Until It Doesn't

The hardest thing to communicate about Kaizen, especially to leaders raised on transformation projects and quarterly KPIs, is that nothing dramatic happens for a long time. When Toyota started exporting cars in the 1950s, "Made in Japan" was not yet a mark of quality. It took roughly forty years for that to fully reverse, and by 2008 Toyota had become the largest car manufacturer in the world.

Forty years is not a strategy most boards will sign off on. It is, however, what actually happened, and the climb was not the result of a few brilliant initiatives. It was the cumulative effect of millions of tiny improvements, each one unremarkable on its own.

The mathematics is straightforward. A 1% weekly improvement compounds to 67% after a year, and roughly twelve‑fold after five years. The numbers themselves aren't the point – the point is that small changes compound the same way savings do, and an organisation that makes small improvements relentlessly will eventually find itself at a level its competitors cannot reach without a much larger and more disruptive intervention.

This is why Kaizen is the pursuit of perfection. Perfection is, by definition, unattainable, but that is the point. Each improvement is one step on a stairway that never ends. The stairway is what keeps people climbing.

The Four Tools

Kaizen is not a single tool. It is a system of four tools operating together inside a culture that supports them. Removing any one of the tools, or attempting to use them outside that culture, is where most adoption attempts fail.

The four tools are:

PDCA (Plan-Do-Check-Act) is the project management discipline at the centre. Plan a change, run it, check what actually happened, act on what you learned. Then repeat. PDCA exists to stop people from operating in what Toyota calls the Do-Do-Do-Do cycle, applying fixes without first examining whether they address the actual problem.

Horenso (報連相) is the communication discipline. It's a compound of three Japanese abbreviations: report, update, and consult. The original concept was developed for hierarchical factory environments, but the version used today operates in three directions – up to supervisors, across to teammates, outward to stakeholders. It's what keeps the other tools connected.

The 5-Whys is how Toyota finds root causes. The method is exactly what it sounds like: state the problem, ask why it happened, ask why about the answer, repeat five times. Five iterations is usually enough to reach a structural cause rather than a surface symptom. Solutions applied at the surface tend to fail: the problem returns in a slightly different form. Solutions applied at the root tend to stick.

Mieruka (見える化) is visual control. Its purpose is to make the state of work visible at a glance. In a Toyota factory this means whiteboards showing daily progress, colour-coded floor markings, magnetic Gantt charts, labels on every shelf. The principle is that information which has to be looked up gets ignored, while information which is unavoidable gets used.

Two supporting principles sit alongside the four tools: Genchi Genbutsu (go and see for yourself), and 5S (workplace organisation). Both feature heavily in factories. Both require significant adaptation when used in a service or knowledge-work context.

Why Most Kaizen Programmes Fail

The tools themselves are not difficult to teach. The cultural conditions under which they actually work are harder to build, which is why so many adoption attempts produce paperwork rather than improvement.

The pattern I see most often is the suggestion box. A company decides to "do Kaizen", installs a box on the wall, runs no training on root-cause analysis or the four tools, and waits. Suggestions trickle in. Management reviews them slowly, ridicules a few of them, and eventually stops reviewing them at all. The box becomes a piece of office furniture. The programme is declared a failure six months later, and Kaizen joins the long list of management fads that didn't work for that particular organisation.

What actually went wrong is rarely the suggestion box itself. The deeper problems are structural. There are six failure modes I see repeatedly, and each one is the mirror image of a cultural condition under which Kaizen succeeds. Briefly:

The organisation treats Kaizen as a short-term project rather than a multi-decade commitment.
Leaders over-emphasise tying improvements to KPIs, ignoring the incremental nature of compounding gains.
The organisational culture punishes change rather than rewarding it.
Senior management endorses Kaizen but does not visibly practise it.
Staff are expected to use the tools without proper training in any of them.
Improvement ideas are required to flow top-down rather than bottom-up.

Any one of these is enough to stall a programme. Most failed Kaizen efforts have three or four of them simultaneously.

How To Actually Start

If you want to introduce Kaizen into an organisation that doesn't yet support it, the path is not a launch event. It is a small experiment, run by a small team, on a real problem, using the four tools properly.

A useful starting point is to pick a recurring annoyance and run a single PDCA cycle on it – something that frustrates the team weekly but isn't life-threatening. Identify the problem. Plan a small change. Make the change. Check whether it worked. Run a 5-Whys analysis on whatever didn't work as expected. Standardise the result. Then move to the next problem.

The first cycle will be awkward. The second will be slightly better. By the fifth, the team will be running them without thinking about it, and at that point the practice has started to embed. This is how Kaizen actually takes root – not through a launch, but through the quiet accumulation of cycles that worked.

If, in forty years, the organisation looks back and sees a stairway, the climb will have been worthwhile. If it looks back and sees a suggestion box gathering dust on a wall, the diagnostic isn't that Kaizen failed. It's that nothing was ever really tried.

The Toyota Production System Series

The Celebration Test: Three Questions for the Applauding Organisation

noreply@blogger.com (Unknown) — Mon, 18 May 2026 12:41:32 +0000

You know these stories. The IT team who worked through the weekend to correct a sudden system failure. The consultant who delivered 80-hour weeks to deliver a last minute deliverable. The applause is genuine. The appreciation, sincere. The reward, token but symbolic.

These are not bad instincts. Recognising effort is a function of healthy culture. The problem begins when recognition becomes the final act: when gratitude closes the conversation rather than opening it. The team that worked overnight becomes a culture story of raising the bar. The individual who put in 80 hours becomes a story of grit and heroic effort. Aspiration stories. Role models.

But celebration obscures the underlying structural question. Why was the weekend necessary? What decision was (or wasn’t) made? And most importantly: what if the team was unable to absorb the load? Because there was no disaster, organisations assume that the system worked. Celebration closes the inquiry, before it begins.

These three questions share a common function: they reopen an inquiry that celebration has closed. They do not replace the structural work of defining decision rights. But they help leaders to detect if they have conflated a rescued outcome with the presence of a functioning system.

The Decision Nobody Made: Silence as Governance

How decision rights migrate downward when no one above claims them. Three cases: The Bangkok Floods, The Fukushima Disaster, Alcoa.

Question 1: The dependency test: If the person/team was not available, what would have happened?

An IT team works throughout the weekend to restore a failed migration. They are successful. Management thanks them, and Monday resumes. Now remove the team’s existence. Would the recovery have proceeded without them? If you cannot point to a documented alternative pathway – a playbook, a cross-trained team, a vendor SLA – it is an exposure. It indicates a single load bearing element in a system with no redundancy. Celebration is cheering on the organisation’s single point of failure.

This is not a question about whether incidents occur. They will. It is a question about whether the recovery was a function of the system, or a function of specific people compensating for the absence of one.

Redundancy is not a commentary on the team's competence. It is a commentary on what the organisation has built around them. A system with no fallback is not resilient. It is lucky. Celebration of luck is not governance.

Question 2: The decision audit: What decision was actually made, and by whom?

Global leadership greenlights a new country entry. Because the commercial window is narrow, the compliance review is deferred. The local team builds the operation. Two years later, an audit discovers the compliance review was never revisited. The team is not compliant. At a town hall, global celebrates the local team’s entrepreneurial spirit.

Who decided to defer the compliance review? If the answer is "no-one", then the decision was reached through silence. This is what Bain's RAPID framework calls the absent "D" – a decision that has no owner, yet was made nonetheless. The celebration was the organisation ratifying a decision it never consciously took.

Most decision audits are retrospective, conducted after failure, in hindsight. The value of asking "who decided?" during celebration is that it is prospective. The outcome was fine. The question is whether the architecture would survive a different outcome.

The question is not whether someone had standing authority to defer the review. They may have. The question is whether the deferral was a conscious exercise of that authority, or a drift that no one logged.

Question 3: The defensibility test: Could the path to this outcome be stated aloud?

A procurement team delivers a critical system migration under budget and ahead of schedule. Leadership celebrates the result. The speed was achieved by sole-sourcing a vendor without competitive evaluation, not as a documented exception, but because the timeline did not permit one and no one escalated the constraint. The procurement policy was not overridden. It was simply not followed.

Two years later, a foreign regulator publishes an enforcement action against the vendor for government corruption. The organisation's name appears in the supplier list. The organisation is now exposed, not because it participated, but because it cannot demonstrate the due diligence that would have surfaced the risk. The process that would have caught it was the process that was never followed. The absence that enabled the speed is now the absence that enables the liability.

Could that procurement path be defended in front of a board? A regulator? The question is not whether the migration succeeded. It did. The question is whether the path to it can be stated aloud. An organisation that celebrates an outcome it could not publicly defend has confused speed – and good intentions – with governance. The front page of the New York Times is a cold place to distinguish between them.

What the questions share

These three questions diagnose what celebration conceals at a single moment. They do not diagnose what the culture rewarded, what the team paid, or what the organisation learned. Those questions matter, and belong elsewhere. These are the ones to ask while the applause is still in the room.

They are not exhaustive. They are chosen because each maps to a distinct governance domain: system, decision, process. Each is most easily concealed by the warmth of recognition.

An organisation that cannot distinguish between a system that worked, and a person who compensated for the lack of decision infrastructure, is an organisation that will need that person again. And one that cannot explain how it arrived there is waiting for someone else to ask. That person may not be there next time. The question is whether the organisation will notice: before or after they are gone.

[message]

Read more: Take a look at our ongoing series on Governance.

📥 Download summaries of this diagnostic (Free): A pocketable summary card · An A4 handout

The Celebration Test

Three diagnostic questions for the moment of applause

	Q1: Dependency	Q2: Decision Audit	Q3: Defensibility
Dimension	System design	Decision rights	Process integrity
Question	If the person or team was unavailable, what would have happened?	What decision was actually made, and by whom?	Could the path to this outcome be stated aloud?
Concealment	A system held together by individuals	A decision reached through silence	A process that was not followed
Exposure	The organisation is lucky, not resilient	The organisation ratifies decisions it never consciously took	The organisation cannot defend the path that produced the outcome

Source: bulsuk.com

The Decision Nobody Made: Silence as Governance

noreply@blogger.com (Unknown) — Sun, 10 May 2026 12:42:00 +0000

In late 2011, the flooding that had worked its way south through Thailand since July reached Bangkok. At its peak, one-third of the city sat under water. Transport networks collapsed. Residents relocated to wherever was dry.

Ann, a consultant, had taken refuge in a conference room with the rest of her family. Her home had been submerged, and its contents destroyed. Having received little guidance from the firm, she continued to travel to the client daily – wading through floodwater, navigating broken public transport, and boarding an overcrowded military truck, the only vehicle tall enough to make it through a metre of water.

Senior management sent out a firm-wide email celebrating her commitment, noting that, despite her personal losses, she had continued to travel to the client, billing and delivering work. The underlying message was clear: others should do the same.

The tone was not lost on the rest of the firm. The celebratory and directive nature of the email contrasted with daily news of electrocutions and drownings.

The Silence That Travels Downward

Ann's decision, on the back of that truck, was not whether the firm should continue operating. That question had long since been settled: by silence. This is what governance failure actually looks like at ground level – not a boardroom catastrophe, but a junior consultant doing risk calculus she was never meant to do, with information she didn't have, in floodwater.

Silence is not the absence of a decision. It is a decision made by default. What happens is that decision rights migrate downwards to the lowest point in the hierarchy. The CEO, by not declaring operations suspended, had effectively delegated the question to the partnership. The partners, reading the CEO’s silence as permission, passed it to their managers. The managers passed it to the field. By the time it reached Ann, the original question of “should the firm continue to operate in a national emergency” had changed into “what do I do to keep my job?”

By that time, it was no longer a governance question. At each layer, the calculus was the same: making the call carried personal risk. Not making it did not.

In a well-governed organisation, decision rights exist to ensure that consequential questions have owners: people with the authority, the information, and the organisational standing to make them. Someone decides to suspend operations. Someone above them ratifies it. The organisation shifts into emergency mode. It is not a cheap decision, or a comfortable one. But it is a decision, made by someone accountable for it, rather than a silence that travels downward until it finds the person least equipped to carry it.

One team leader, in the same firm, during the same flood, made a different calculation. He told his team to stay home and re-organise client work around the assumption that the firm was, for the moment, not operating normally. His team did not end up on army trucks. The organisation did not notice the difference.

Fukushima

The same governance failure, at a different scale, cost Japan three reactor meltdowns, the forced evacuation of 150,000 people, and the only Level 7 nuclear event outside Chernobyl.

The engineering problem was relatively simple. The governance problem was not. The reactors had lost cooling and needed to be vented and flooded with seawater – a technical call which would also destroy the plant commercially. TEPCO’s headquarters, working from incomplete information at a distance, neither ordered the action nor delegated the authority to order it. Plant manager Masao Yoshida had the ground-level picture but not the organisational mandate. Headquarters approved instructions that contradicted decisions already made at the plant. The subsequent investigation was unable to identify, for several critical decisions, who had been authorised to make them.

When Yoshida was eventually ordered to stop the seawater injection, he ignored the instruction and continued pumping. The decision probably saved Tokyo, and subsequent investigations vindicated it. What it did not vindicate was the governance that forced him into it. A system that requires a plant manager to disobey a direct order to produce the right outcome is not a system. Good outcomes delivered in that manner are not governance. They are luck, dressed up.

Repossessing the Question

If Thailand and Fukushima show what happens when decision rights migrate downward, Alcoa shows the opposite: what it looks like when a leader deliberately pulls decision rights upward and refuses to let silence govern the system.

When Paul O’Neill took over aluminium producer Alcoa in 1987, injuries and fatalities were routine. They were technically compliant with regulations, and had a ‘low’ lost workday injury rate of 1.86 per 100 employees, compared to a US industry average of 5.0 per 100 employees. Plants were decentralised, with their own cultures and management.

O’Neill believed injuries were not an acceptable cost of doing business. Six months into his tenure, a new employee was killed when he was crushed helping to fix a jam inside their machines. O’Neill’s response was pointed: “We killed this man. It’s my failure of leadership. I caused his death.”

He implemented a new mandate: he was to be informed of all incidents, no matter how small or seemingly inconsequential. As the information flowed to the CEO directly, it removed the ability of each layer to compress, delay, or reinterpret events. The relevant unit executive was also required to provide a plan on how to fix the incident. No longer was it an ‘acceptable loss’, it was now a problem that needed solving at speed.

Even when a plant achieved a period without a lost-time injury, O’Neill asked them to also show near misses, and what they had fixed as a result. He refused to let the absence of a bad outcome act as evidence that the system was working.

This was tested in Mexico where 150 employees were injured after a carbon monoxide leak at a plant. The local executive did not report the incident, and O’Neill was unaware until a Benedictine nun from the area raised the issue at a shareholders meeting. He ordered an investigation, and discovered evidence of a cover up. The executive was immediately fired.

Each of these moves was an effort to repossess decision rights. The authority to define the system's health was pulled upward, out of the hands of the people whose performance depended on the system looking healthy.

Alcoa's lost-workday injury rate fell by approximately 90 per cent over O'Neill's tenure. He built an organisation that can distinguish between ‘the system worked’ and ‘we were lucky’. The same discipline that kept injuries from being concealed upward kept quality problems, supply failures, and strategic weaknesses from being concealed upward. Safety was the first domain. The governance architecture expanded to other functions.

Success by Omission

Back in Bangkok, the firm celebrated success because nothing had gone wrong. Ann made it back to the office. She billed. She delivered. By every measure of success, the system had worked.

But it had not. It simply transferred the cost of its own failures, and of its unmade decisions, onto the person least equipped to bear it – and then patted her on the back for doing so.

The most dangerous governance system is the one that appears to be working, mistaking omission for success. Fukushima produced a commission, a Level 7 classification, and a national reckoning. Ann's firm produced a congratulatory email. One of these organisations learned something. The other may still do it again.

[message]

Read more: Take a look at our ongoing series on Governance.

The Celebration Test

Three diagnostic questions to ask the next time an organisation celebrates individual effort.

Karn Bulsuk: Full Speed Ahead

Goodnight, Kyoto | Photography

Best Split Keyboards 2026: Portable Ergonomic Picks Reviewed & Compared

The Best Portable and Travel Split Keyboards in 2026

Quick Comparison: All Keyboards

Which split keyboard is right for you?

Weight is your only criterion

You want the best typing feel you can still carry

You need your hands genuinely far apart

You are on a tight budget

You use a Mac

You write for hours at a stretch

On the horizon

The Compose Keyboard – $300+

Keyboards we are reviewing

Reviews

Keychron K15 Max

SpiderIsland Tech Split Macro Keyboard

Pros

Cons

ProtoArc XK03

Pros

Cons

Best Budget Split Keyboards Under $50

JOMAA Ergonomic Split Foldable Keyboard

Upgrade picks

NocFree &

Pros

Cons

Elim Elytra – $249

Pros

Cons

Retired benchmark

Keychron K11 Pro/Max – retired benchmark

Pros

Cons

Keyboards we considered, but did not pass screening

Keychron B11 Pro

NocFree Lite Wireless Split Keyboard

R-Go Split Keyboard

ZSA's Voyager

Goldtouch Go!2 Bluetooth Wireless Mobile Keyboard

Keyboards we reviewed, but can't recommend

Generic Foldable Split Ergonomic Keyboard

Looking for a desktop split ergonomic keyboard?

How We Test Split Keyboards

Must-haves:

Nice-to-haves:

Optional:

FAQs

Read more

Updates

How to Make Hong Kong Lemon Tea (凍檸茶 Dong Ling Cha): Authentic Recipe

Ingredients

Method

A note on hot lemon tea

FAQ

The 2005 Amagasaki Rail Disaster: What the Official Report Reveals About Japan’s Worst Train Crash Since 1963

I.

II.

III.

IV.

V.

VI.

VII.

VIII.

IX.

Notes

The Value That Wouldn't Travel: Burberry, Pepsi and the Asset That Stays Behind

A permanent downsizing

The mirror case

Where the value sits

Elim Elytra Review: A Split Keyboard You Can Actually Carry

Pros

Cons

Layout

Typing experience

Ergonomics

Portability

Build quality

Mistake 2: using the wrong teas - Earl Grey or English Breakfast
(or worse, tea bags)