The takeaway is not that they were unlucky or a weak team. The team actually went on to create many amazing products in other companies, but the issue was a lack of constraints and forcing functions. They said yes to everything instead of being forced to say no and focus.

I realized I now have the same problem with AI tools. You get this feeling of near-infinite output potential because everything is “just a prompt away”. Everything feels cheap now. It’s like all my computer-related work is on sale, and so now I’m overbuying. I’m like the guy who loads up on way too much during Black Friday because he can’t pass up the deal, even if he doesn’t actually need the stuff he’s buying. I constantly end up with dozens of small feature branches stacking up in DevX of things that sounded easy, but got stuck in the mud of manual Jon work that AI couldn’t offload.

I’ve started calling this gap the 20% tax. AI is great at starting things and tackling that first 80% that used to feel so hard. The last 20% is all the stuff AI doesn’t really do yet like taste, deciding what to build, integration, real-world validation, final reviews and verifications you still end up doing, project overhead, etc. The constraint used to be can I build this? Now it’s a stack of 20% questions: should I build this? Is this the right version? The right architecture?

AI collapses the marginal cost of starting to essentially zero, but the marginal cost of finishing is the same or even grows because now I’m juggling it across even more projects or tasks. Ironically this just means I end up working more. It’s like the Jevons Paradox where these outputs all feel dramatically cheaper (even cheaper than they are), so I spin up more work and more stuff and end up working more than I used to – not less. In some ways this all feels like the same problem I talked about in Promptcrastination, but instead of avoiding certain tasks in a project it’s spinning up more new projects or tasks. It’s all the same root problem of the capability setting the agenda instead of my priorities setting it.

Let’s be honest, though, some of the extra 20%s are totally worth it. AI making the work cheaper does genuinely expand the frontier of what’s worth doing and this means higher leverage for doers who know how to use these tools to their maximum potential. The goal isn’t necessarily to start fewer things, it’s to make sure you’re still choosing the tasks and weighing the full cost of them. Not artificial scarcity, but deliberate constraint.

I think the way I’m going to address this is to start testing my own constraints. Put a cap on the number of works-in-progress and make sure I try to really evaluate the last 20% before I start anything. Then capture that last 20% and compare it to my prediction to try to get better at estimating this, especially since it’s likely a moving target as models and tools improve. 

General Magic had seemingly every resource and no constraints, and ultimately it became a case study in the importance of constraints. AI gives us the same gift, but also the same challenge. The discipline we must all develop in the AI era isn’t better prompting, it’s fully pricing the 20% tax before you buy.

It feels like the same pattern we all went down for code review. Initially, you reviewed everything. Then you’d skip the simple stuff like unit tests and boiler plate. After a while you switched to just the more important sections. And as the tooling got better and we have more and more agent driven code review tools to help surface issues, I find myself only reviewing the critical sections or the high level outputs instead.

Now that the models have improved and agents can do way more for much longer runs, it only makes sense to give them more permissions. Who wants to sit around hitting enter to hand check every tool call? Why block the next 20 minutes of progress working in the background because it got stuck asking if it could run a fancy bash command to search for something in a local file.

And as agents can do more, we naturally want to give them access to more. When it can reach my database, my logs, my metrics, my emails, chats, transcripts, and docs it has more context and can actually do more for me.

But then you have that moment where it all goes wrong. Or nearly does. I’ve had a few now. Once, my agent was digging into why a prod service was throwing so many errors, went a little paperclip-optimizer on me, and decided the cleanest fix was to just remove the service. No more service, no more errors. Another time one blew away a chunk of a local dev database mid-migration. None of mine have been truly catastrophic yet, but every near miss makes me rethink how much access and permission I keep handing these things.

I didn’t want to babysit every tool call. I just wanted to keep running it and still catch the scary stuff. So I started building some tooling around it. Initially I just wanted better visibility for auditing. I started aggregating session logs across my different agent harnesses (Claude Code, Codex, Pi, OpenCode, etc). I added hooks to better log tool calls (especially potentially risky ones). Then I added an egress proxy and routed all network calls through it so I could log specific outbound LLM calls, MCP calls, third-party API calls and see exactly what my agents were doing – even when I wasn’t watching. And then I could have a clean agent post process and surface anything weird or scary.

Obviously that only catches things after the fact. The next logical step was control. So I started adding some smarts to the proxy so it could reason about each outbound request in context. Not just match a domain or IP against an allow/deny list, but make a real decision about whether to let it through. Improving my chances of catching the weird edge case before it happens instead of reading about it in the logs later.

How do you handle this? Do you find yourself wanting to give your agents more and more access and control too? How do you review and secure your setup? Or do you even care to?

But here’s the thing – writing on my Android phone sucks. The keyboard is small, the dictation is clunky and transcripty, and the whole thing makes me want to go running back to my laptop with a full keyboard any time I need to write more than a quick response.

That’s why I built Nibit. Nibit is a second keyboard for Android packed with snippets, transforms, smart dictation, and a bunch more. You can just use simple triggers like ",addr" and have it expand into your full text snippet. You can quickly type ",." and trigger the built-in smart dictation that uses AI assistance to clean up your ramblings, handle self correction (“Let’s meet Wednesday…no, wait, Friday” => “Let’s meet Friday”), and automatically remove all your fillers like um’s and uh’s.

You can also push things (text, links, files, screenshots) between your phone and laptop, fully E2E encrypted, without texting yourself links. This sounds small, but has been a huge time saver in my workflow. The files have a preset TTL, auto-delete, and I have a Raycast extension to paste or push them with a keyboard shortcut. If you’re also a Raycast user you can easily import your existing Snippets and Quick Links as well.

There are a bunch more features, but the punch line is a quick access way to streamline all your writing on your phone and make it faster, easier, and just a more pleasant experience. And the best part is it auto-switches back to your main keyboard (i.e. Gboard) so you don’t have to give anything up to use it.

You can check out the full feature set at nibit.app or get it directly from the Play Store.

If you do real work from your phone and have the same frustrations I did, give it a shot and tell me what you think.

Most of the people I talk to find that AI isn’t reducing the amount of work they’re doing, it’s actually increasing it. The leverage it gives you makes you take on more. The agents are available 24/7 and you can have as many as you can afford to keep feeding tokens.

On the flip side, though, I feel like my work is more flexible than ever in many ways too. I’m no longer locked into meetings all day, only productive at my desktop, and limited to when my coworkers happen to be around to be most productive. I can now keep things spinning at my desk, when I have a down moment waiting in line, or when I’m outside sitting at the park. The moments I used to spend doom scrolling, I now spend pushing agents forward.

I haven’t figured out the right balance yet and maybe there isn’t one. Maybe Agent Angst is just the new normal for anyone building with AI. But I do know that the moments I used to waste are now moments that move things forward, and that feels like a net win even if my brain hasn’t fully adjusted to it yet.

Have you felt Agent Angst yet? I’d love to hear how you’re managing it.

I’ve started calling this promptcrastination: using AI-friendly tasks as a sophisticated form of procrastination.

What promptcrastination looks like in the wild

Promptcrastination is often rooted in valuing output quantity over output quality, and with AI it’s really easy to output A LOT of quantity quickly with the illusion of quality.

Promptcrastination is when you:

• add another round of unit tests you don’t really need
• start building out a greenfield feature that sounds cool, but you don’t know has value
• generate 20 marketing angles instead of emailing 5 potential users
• build a new agent workflow instead of doing the uncomfortable work it was supposed to replace
• “research” competitors with AI summaries instead of talking to a real customer

Each of these can be useful. The trap is that they’re endlessly expandable and emotionally safer than the hard thing.

Why this happens (and why it’s rational)

AI amplifies what’s already true about human behavior:

• We avoid ambiguity.
• We avoid rejection.
• We avoid tasks where progress is hard to measure.
• We seek quick wins, especially when we’re tired.

AI gives you a slot machine for “progress”: prompts in, output out. It’s immediate feedback, low friction, low risk. The really diabolical thing, though, is that it looks like productivity and feels great in the moment.

So your brain does the sensible thing: it routes work toward where the gradient is easiest.

That’s core trap in a sentence: when capability shapes priorities instead of priorities shaping capability.

The cost isn’t time, it’s trajectory

The obvious cost is wasted hours.

The bigger cost is that it subtly changes what your “default work” becomes.

You start identifying as someone who is always refining instead of someone who ships and learns. You start optimizing for polish before you have signal. You get good at producing artifacts and bad at making serious bets.

You can end up with a very clean pile of output… and no meaningful progress. And it all feels great until you look back at your day, week, month and realize you’ve accomplished little of actual value.

How to fight it without rejecting AI

The goal isn’t “use less AI.” The goal is to stop letting AI decide what “productive” means.

A few tactics that work:

1. Prioritize your task list explicitly. Don’t have 5 things to do and do them in any order. Make the order explicit and enforce it. You’ll naturally want to gravitate to the “easy” tasks, but stick to the priority.
2. Celebrate results, not output. Keep your focus on the goal or outcome you’re trying to achieve and don’t get caught up in false productivity along the way. Write “identify 3 features that would get Acme Corp to switch” instead of “do market research”.
3. Measure contact with reality. If the work doesn’t touch users or potential users it’s worth being extra diligent to verify it’s actually a leverage point and not just more promptcrastination.

The takeaway

The solution isn’t to avoid using AI. It’s holding yourself accountable for whether the thing you’re prompting is the thing that actually matters.

Next time you catch yourself three hours deep in a beautifully polished something, ask: am I working, or am I promptcrastinating?

Now the economics have changed. What took 10 engineers and 18 months now happens with a small team and AI in weeks. I’ve seen it firsthand. Tools that crank out code, prototypes, even market tests faster than ever. So why keep grinding the the old way? That’s where Crouton Creations comes in. It’s my next project, a lab where I blend startup grit with AI’s raw power to build new products.

After years of scaling teams and navigating bigger companies, this is about getting back to what I love: solving real problems, fast. No fluff, just shippable stuff.

Why Now, Why This

AI has changed the game. I started tinkering with it for my own headaches, like automating tedious tasks or building quick apps to scratch an itch. Long story short, it worked so well I thought: why not turn this into a deliberate experiment?

• Scratch my own itch first. Every idea starts with a problem I face. If I wouldn’t use it daily, it’s dead on arrival.
• Bootstrap by default. Profitability from day one keeps things lean. Control beats chaos every time.
• Small team, big speed. No meetings marathons. Just me and a couple sharp collaborators iterating in days.
• Embrace the fails. Some products will flop. That’s ok. Document it all, learn quick, pivot faster.

Ultimately, Crouton Creations is about playing to my strengths: building what works, without the pressure to scale at all costs. We’re shipping AI-powered tools – think productivity tools and utilities – that real people (starting with me) actually use.

This is my bet, but it’s not a risky one. I’ve seen what works when you strip away the theater and focus on real problems. I’ve built companies before. This time I’m just doing it faster and smarter, shipping AI-powered tools.

If you’re curious about what happens or want to help me figure out what bootstrapping a small, AI-native startup looks like, follow along. I’ll be sharing the wins, the mistakes, and the lessons learned along the way.

I’ve learned so much along the way and have met and worked with some truly great people. The time with Torbit and Walmart has meaningfully changed my life. It’s allowed me to learn about and build a bunch of really interesting products around proxies, load balancers, edge infrastructure, BPF, RUM, website optimization, edge security, and many other facets. Between the flexibility of working remote and traveling to various offices and markets, it’s let me really embrace my passion for travel. It’s allowed me to grow as a manager, mentor some great people, and build and manage a large distributed team which has been a really amazing experience. I’ve learned a ton about how large enterprises and eCommerce sites work from my time at Walmart as well. I had a relatively uncommon opportunity to continue to build what was very close to a startup within a large company and it was overall a great experience getting to see how that can work after a startup acquisition.

To everyone that was a part of that journey with me I want to say thank you. It has really been an adventure and I can’t image how my life would’ve turned out without this opportunity.

Now I’m excited to see what’s next. I haven’t yet decided on my next project or opportunity, but if you happen to have some ideas or roles that might be interesting feel free to reach out and let me know. I’m looking forward to starting the next big chapter and a whole new set of learning, experiences, and people to share it with.

At work, this was a common mantra for me in my 1-1’s with my teams this year. It felt like across the board we were often trying to find the most important things to focus on and focus more deeply on those things in order to make sure we got where we needed to be. This was a really challenging year at work and it required us to be much more diligent, focused, and disciplined about picking our goals carefully.

This also struck my wife and me in our travel planning this year. We managed to cross off some of our biggest bucket list items including Antartica and Egypt among several others, but we also traveled to far fewer countries and places than we have in recent years. We slowed down for a lot of the rest of our travel throughout the year in order to try to enjoy where we were more and prevent (or at least slow down) burning ourselves out as much.

On the personal development side I found it also applied. I found myself taking this year and focusing on one or two single goals at a time and planning them out a bit more deliberately across the months of the year instead of just trying to tackle all my goals simultaneously or continually adding new ones on as I go as I’ve been known to do in the past.

In the end, 2019 was a good year for me in terms of accomplishments and goals. I think focussing and being deliberate about them made it a much stronger year than the previous one. However it is also a year that has left me feeing drained and needing to regroup a bit. I’m in the midst of some life transitions that I think will continue on into 2020.

Here’s to hoping for a great 2020.

The final stats were:

• 145,000 miles flown
• 4 continents
• 22 countries
• 1 US Territory

It’s been a whirlwind of new experiences, food, and amazing sites and I’ve loved every minute of it.  Especially getting the chance to catch up with old friends, and make new ones, all around the world.

I’ve also found an odd joy in maximizing my travel rewards, status, and credit card spending to get the most of my new found travel expenditures.  I’ll probably elaborate on some of these in a future post.

So in short I’m thankful for a wonderful year that allowed me to travel the world with the coolest person I know – my wife.  Here’s to hoping 2016 is just as amazing.

All of these traps are easy to fall into, but are the wrong way to think about it. You are not your company. If you are, then you are either just starting out and can’t really assess this new creation yet, or you are probably doing it wrong. The idea of an entrepreneur is to ideally build a company that is bigger than the founder. In the early days you are critical, but over time your job is to build systems and hire people (or teams of people) to fill in these critical roles. In many ways the true measure of success is when you’ve managed to make yourself no longer a necessary piece.

In this way, the company’s success or failure is clearly separate from your own personal success or failure. The “I must be essential to my business” mantra starts to sound like more of an ego driven desire for importance than a cry of exhaustion. And the ability to sell a company that can operate and be valuable to someone else (especially in your absence) starts to sound like a success instead of “selling out” and “giving up your baby”.

When you are building your company, it’s going to be your life for a period of time. It’s long hours, constant fires, and completely all-consuming. It’s still important, though, to remember that the end goal is to build something bigger than you. To build something where you are not a critical piece of it’s successful daily operation. Something other people find value in with or without you. The sooner you can untangle the company and your own sense of worth and move to this idea of “bigger than me”, the easier it will be for you to be pragmatic about your company and ultimately be more successful; in particular when viewed over a career instead of just a single company.