Beyond Search

A Free LLM: Is There a Catch?

2026-08-24T13:30:30Z

Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.

A flurry of write ups appeared online in the last couple of days (today is Sunday, August 23, 2026 for those who wonder how timely Beyond Search is. Well, it isn’t timely, but it’s free and there are no ads). A good example is “A Free AI Model Is Winning Over Developers. And Nobody Knows Whose Servers It Runs On.” Free, a stealth service, and did I mention free?

Thanks, Midjourney. Good enough.

The write up says:

Ox Alpha arrived on OpenRouter with a million-token context and no price tag, in a listing that says prompts and completions are retained by a provider nobody has identified.

Let’s play guess whose technology? [a] Antarctica, [b] Vatican City, [c] China, [d] the United States. Maybe [d]? Probably not Antarctica although some penguins are very bright. The Vatican? Unlikely. The US? Free hasn’t worked out for the American players so far.

I don’t know, but I would guess China. Why? Pony Alpha became available in October 2025, and it was Zhipu AI GLM 5. Hunter Alpha went live on OpenRouter in March 2026. That, according to my team, was Xiaomi MiMo’s. Then Healer Alpha popped up the same day (March 11, 2026). And on or about August 20, 2026 Ox Alpha went live with what seems to be Zhipu AI GLM 5.3.

Coincidence? For this dinobaby, nope. Two firms, both China linked or Chinese from the git-go.

The cited write up states:

The scale on offer is not modest. The open-source agent OpenCode said the model would be free for a week with near unlimited usage, and that its provider had capacity for 100 trillion tokens a day….OpenRouter’s own listing states that prompts and completions “are retained by the provider and are not used for training.”

What’s interesting is that OpenRouter availability makes the service accessible by millions of developers. The data generated by the interactions of those interested in the service and that Ox Alpha system could be used for [a] training, [b] data collection / surveillance, or [c] data mining; that is, what user cohort is doing what and when. Other options exists as well. With a bit of fancy dancing, the

OpenRouter is in the midst of a buy out deal involving Stripe. Therefore, its oversight of “stealth” LLM testing might be less than optimal. Together AI could provide similar stealthiness, but OpenRouter has eyeballs and broader visibility.

I spoke to one member of my team who is slightly more suspicious than I am about free smart software. That individual suggested that if Ox Alpha were used to generate code, it is possible that “phone home” or other functions could be inserted in the output. A busy person might just copy, paste, and launch. The second possibility is to let the model show its agentic capabilities. Who knows what those smart, almost effortless agents can do. I don’t. Therefore, operational security is probably a good idea for those jumping on Ox Alpha.

OpenRouter may be a testing ground for a new approach to malicious code creativity. No vendor identification and limited or zero legal discourse — Yep, that free LLM is ideal for some purposes like spyware.

Net net: Coverage of this type of LLM marketing, data collection, and operation is limited. Cyber security vendors are busy and some may be using Ox Alpha.

Stephen E Arnold, August 24, 2026

AI Big Boys: You Want What?

2026-08-22T14:59:26Z

Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.

I read a hoot of a column in the newspaper which sure looks like a Web site or an old-fashioned paper magazine named The Economist. The title of the write up is “AI Agents Lie, Cheat and Steal. That Is Putting Off Users.” The subtitle evoked a laugh from me:

It is time to impose law and order on the frontier.

Well, The Economist is making its viewpoint clear. I can here Randolph Scott saying in his B movie:

Hold on a minute, Idabelle. Let’s get rid of this Alabama tan first and you better wash this shirt.

Thanks, MidJourney. Good enough.

Yes, sir. Clean up your act.

I thought the entire purpose of the Economist was to explain how economics really works: Get money, fund an idea, and try to create a monster revenue machine. Are these steps Rinso fresh? Are rules stretched or ignored? Did JP Morgan or Andrew Carnegie ask permission? No, sir. The goal was clear.

What does the opinion piece say about the “next big thing” and the BAITs (big AI tech outfits)?

What AI doesn’t have is enough settlers. That is, the individuals and businesses who make full use of products such as AI agents, helping to justify the reams of investment pouring into frontier AI. They are put off partly because, like in the Wild West, life on the frontier is reckless. As recent “loss-of-control” episodes by the most advanced models of Anthropic and OpenAI attest, agents, which are supposed to work on people’s behalf in “alignment” with their values, lie, cheat and steal if necessary. They break free from captivity and form harmful posses to do harm to people. They’d drink whisky and brawl if they could.

Now, pardner, I don’t know about you, but those behaviors are exactly what tamed the Wild West? I don’t think it was the lawmen. I think it was the colonization aided by the US Army. The Civil War contributed some feisty lads from the North and South. The varmints didn’t stand a change.

The write up says:

The frailties have given rise to another group within the cohort attracting VC interest: those promising to strengthen the “trust layer” of agentic AI. One is Cyera, whose valuation has quadrupled to $12bn in 18 months. It says a lack of trust has “stalled” AI adoption recently, and provides services to prevent data leaks and unauthorized tool use. Another is Scaled Cognition, co-founded by Dan Klein, a Berkeley professor, that recently raised $100m in VC backing. It promises to reduce what Mr Klein calls the “invisible errors” produced by AI—those that are plausible enough to be missed and compound as agents perform longer tasks—by incorporating guaranteed reliability into the training of its models. Other startups specialize in creating evaluations to measure agents’ effectiveness, agent identities to determine where liability lies if they go rogue and control systems with a kill switch.

Yes, sir, a kill switch. Exactly whose finger rests on that ideal button? Will The Economist be the trusted button pusher? Will it be a government? And if a government, exactly which government? Maybe it will be the now neutralized agent of big thoughts, Demis Hassabis?

Hey, you saddle slicker, I thought the market economy paved that muddy road. Now you tenderfoot to the government for help. I truly don’t understand. If I didn’t know better, I am thinking that you are reading Vladimir Lenin’s writings before you head off to dreamland.

Like Randolph Scott said in that 1940s movie, “Hold on, Idabelle.”

Stephen E Arnold, August 24, 2026

Meta and a Mythical $1.4 Trillion in Fines. Ho Ho Ho

2026-08-22T14:53:01Z

Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.

Time to stop with the manifestos. Maybe spend less time explaining the confusion about helping ships in distress? Think about message in “Meta Faces a $1.4 Trillion Reckoning in Latest Trial Over Social Media Addiction.” Add that to the off-the-balance-sheet commitments for some smart software doo-dads, and that’s a lot of money.

Thanks, MidJourney. Good enough.

The write up says:

While Meta has faced similar allegations in other trials this year, the impact of this trial could be especially significant. California, Colorado, Kentucky and New Jersey haven’t publicly disclosed how much they plan to seek. But Meta said in a court filing ahead of the trial that states were seeking penalties as high as $1.4 trillion, an amount it described as “outlandish.”

Who is picking on the estimable Mark Zuckerberg (or, more familiarly, “El Zucko”? The Engadget article states:

For Meta, the stakes are especially high because the company is currently facing thousands of other lawsuits that accuse it of harming users. Juries in Los Angeles and New Mexico have already ruled against Meta in high-profile trials that deal with similar issues. (Meta has said it will appeal in both cases.) Another loss could not only be a financial blow, it could give other lawsuits an easier path forward.

Following the standard big AI tech game plan, El Zucko’s legal and crisis communications professionals assert:

Meta believes that the states’ claims are unproven and the proposed penalties calculated incorrectly. “The State AGs may call this a landmark case, but their limited claims are unsubstantiated and their financial demands are vastly disproportionate.” The AGs offer no proof anyone in their states was misled, claim benign features like having an additional Instagram account somehow harmed their residents, and attempt to penalize Meta for industry-wide challenges like age verification,” the spokesperson continued. “Rather than sticking to the facts or the law, the states have instead decided to chase an outlandish payout. We stand by our record of creating strong protections for teens, and look forward to making our case in court.”

I wonder how the parents of a suicide allegedly bullied on a Meta property will respond to the word “outlandish”? I don’t want to speculate. You may wish to use your imagination.

Several observations:

In the present legal environment, I think that the appeals, court decisions, and follow on legal appeals with stretch the resources of some of the parties to the matter. Online advertising is known to produce a flood of money. Some of those funds can be used to grind down the will, legal teams, and check writers’ enthusiasm before El Zucko writes a check.
I am not a Facebook user. We do use a software script to post some of my Beyond Search stories, but I am not sure if anyone knows they are on the platform. I have been asked to check, but I just don’t care.
Whatever happens in the US will be noticed by the legal professionals working in the European Union. What happens in the US will inform those professionals how to perceive El Zucko, his manifestos, his platform, and the interaction of his services with certain cohorts.

Net net: BAIT outfits and social media advertising services are likely to face increased pressure in the future. Of course, legal measures can be crushed systematically. I think there will be some people who want to dress up like Don Quixote and sally forth.

Stephen E Arnold, August 24, 2026

Believe It or Not: Russian Spiritual AI

2026-08-22T15:03:38Z

Russia Brings Morality And Spirituality To AI

We can’t leave Russia out of the race to be the world’s leader at the forefront of AI.? ? China and the US are chugging away at that title, but so is the Great Bear. BoingBoing tells us that Russia is going to do something new and different with AI that the China and the US aren’t concerned with: “Russia To Test AI Models For Traditional And Spiritual Moral Values.” The information comes from Vedomosti, a Russian business daily run in Moscow, that if you want to sell AI to Russia you need to appeal to the country’s traditional spiritual and moral values. These new mandates were instituted by none other than Vladimir Putin.

If you want to sell your AI in Russia, neural network, architecture documents, information on filtering mechanisms, and other related information would be sent to a specialized testing laboratory.? ? The laboratory would test the model’s answers and whether the internal restrictions can be circumvented before issuing a finding on compliance. More than one test would be carried out on the AI to determine if it complies with Russia’s new standards.

This means that:

“The new law implements Putin’s earlier requirement that AI reflect “traditional Russian spiritual and moral values”. Federal Law No. 243-FZ covers large language models with 1 billion or more parameters and defines a “sovereign” model as one built entirely in Russia, with reproducible training and original weights, and “national” models as those including foreign components. Both must be trained and run in Russian datacenters. Not only does an approved lab get to pore over the models, but also Russia’s FSB security service and other state bodies. The requirements (and designations) take effect in March 2027.”

Putin is practicing SOP in guaranteeing that AI follows “traditional Russian spiritual and moral values.” He controls everything so this doesn’t come as a surprise. Once a company’s AI is approved, however, the big money will start rolling in.

Whitney Grace, August 24, 2026

Why Specialized Service Outfits Should Not Become Too Visible

2026-08-15T18:58:36Z

Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.

I have used NSO Group and its Pegasus specialized software as an example of the problems marketing or publicity creates. I may have to update my example based on the allegedly accurate information in “Sensing He’s in Deep Trouble, Flock Safety CEO Says It’s All Been a Big Understanding.” Well, maybe, but the publicity has had some knock on effects. The cameras are causing crime. Citizens who are able to spot a Flock have been killing the data birds. That’s illegal, and for a policeware company causing crime is not a great idea.

Thanks, Midjourney. Good enough.

What’s a Flock? Stated simply: 24×7 cameras, data capture, software to correlate information. Purpose: Enforce the law.

But the big problem is that what probably started as a cute “nerd moment” blew up into a PR problem. The article is an example of what Flock’s crisis communications team is going to have to ameliorate. The way marketing of specialized products and services used to work until the NSO Group’s emergence as the poster child of bad marketing was:

Write a white paper and maybe get an outfit like Police1 to help distribute it to professionals who work in a specialized enforcement field
Sign up for a trade show booth where attendees have to be verified as government professionals and civilians are few an far between (except for this dinobaby who gives talks at these types of events)
Set up one on one meetings with trade show attendees who express interest in the specialized service or product
Prepare a video (hopefully professionally produced) and put it on a private video distribution service like the original Vimeo. I don’t know who is in this niche now, but a specialized software and product company should be able to use a Web search system to locate a service. (I am not a cheerleader for this type of content to be posted on YouTube. Google loves training its smart software on interesting content. Make what you will of my observation.)
Respond to calls for papers at relevant law enforcement and related events.

The problem with this five step process is that it does none of the marketing actions that can create a viral PR problem. If a specialized software company holds Zoom webinars and blasts out invites to any name on a data broker’s list, trouble can arise. Skip the off the cuff webinars with influencers. People who develop specialized products and services can go overboard with examples, use cases, and explanations of how a neutral technology can be applied.

The write up says:

The company is still trying to paint itself as the good guy, promising to help law enforcement to fight crime, while underplaying egregious instances its tech being misused. Flock also announced that it’s making changes to its platform amidst major scrutiny from lawmakers and civil liberties groups in an apparent attempt to address privacy concerns.

I noted this passage:

Beyond its surveillance network of over 100,000 cameras, the company is hoping to unleash drones, surveillance trailers, and gunshot-detecting hardware — efforts that likely won’t convince an already extremely wary public.

That’s not to mention glaring technological issues plaguing its license plate readers, which remain startlingly ineffective at the task they were designed to do. According to a recent investigation by the police department of Roseville, California, its cameras mistook one license plate number for another an astonishing 71 percent of the time between 2023 and 2024.

Modern management and marketing methods. I want a baseball cap that says, “Get Flocked.” But that could lead to a big misunderstanding.

Stephen E Arnold, August 21, 2026

A Data Center Is on the Way to San Jose

2026-08-15T19:28:49Z

Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.

When I worked in Silicon Valley, I visited Gilroy, the garlic capital of the world. A couple of my colleagues from the zippy start up at which I worked wanted to check out the Gilroy Garlic Festival. I remember the odor of garlic billowing from different food stands. My recollection is that it was rural compared to Berkeley where I lived, and it smelled … of garlic. I then left Gilroy, and I have not returned. Now Gilroy is in the news for something other than garlic.

“Amazon Secretly Circumvents Community Vote for Massive AI Data Center Using 45-Year-Old Rules. Gilroy Residents Locked Out of Public Comment Window” reported:

Roger Wehner, Amazon Web Services vice president of economic development, told WSJ that the site had gone through a long approval process, which required public notices and public-comment periods. However, when Gilroy resident Rosa Rodriguez wanted to state her concern about the damage a data center can do to a region suffering from drought, she was told that the comment period had already lapsed in 2024, long before data centers became a hot topic across the country. Gilroy City Mayor Greg Bozzo said that other industrial projects that had a similar scale to the Amazon data center have previously been approved without resistance.

Thanks, Midjourney. Good enough.

That adds some spice to the data center tale in my opinion.

Amazon is now taking steps to make sure the garlic capital is well seasoned with information; to wit:

The tech giant also said that it’s engaging more with the community, with Amazon officials claiming that they met residents directly and even hosted an open house where people can ask questions and leave comments. It also addressed fears that the development would cause a water shortage, which other data centers have already caused, saying the company is constructing a wastewater recovery system to supply its water needs and will even build a pipeline to bring wastewater from the source to its plant.

Cooking with water is over-rated. Use olive oil. Plus, garlic plants grow underground. When it rains in Gilroy, the soil catches and retains the water. A data center is probably a mere sipper of the groundwater from the Llagas Subbasin. Plus, any graywater can be used to water the garlic plants.

But what about noise? Noise, what noise?

But what about emissions from back up generators? Emissions? What emissions. Residents cannot smell the miniscule emissions due to the prevalence of the garlic odor from the festival.

But what about power grid demand? Power. Hey, no problem. Amazon can pay for power just like those residing in the garlic capital can.

There you go. A data center, no public comment and no problem.

Residents of Gilroy, pay attention to land use hearings and rules. Oh, did you know that hanging garlic on a string around one’s neck does not ward off data centers.

Stephen E Arnold, August 21, 2026

The AI Agents Are the Likely Job Takers

2026-08-14T21:08:20Z

AI agents are predicted to take away more human jobs than they will create.? The same thing happened with the Industrial Revolution and the implantation of computers, but jobs still existed.? ? What is going to happen with AI then? Nordot.app shares a. Possible reality: “AI Agent Future Is Coming, OpenClaw Creator Tells AFP.”

Peter Steinberger created OpenClaw, an AI agent that can do real-life tasks associated with technology like checking him in for a flight or adding an appointment to his calendar.? OpenClaw can’t replace a human assistant yet, but in a few years AI agents will be smarter and more commonplace or so Steinberger claims.

Jensen Huang of Nvidia dubbed OpenClaw as the next ChatGPT.? ? There’s still concern, however, about security risks such as access to personal data.? ? Steinberger believes that the next big thing in AI will come from someone who isn’t attached to? ? a big corporation and who wants to have fun with AI.? ? He is worried about security and how his agent could be used by bad actors:

“ ‘Yes, I do worry a bit, especially because there’s now a whole cottage industry of companies that try to make a big buck and make it even simpler to install OpenClaw,’ he said. ‘I purposefully didn’t make it simpler so people would stop and read and understand: what is AI, that AI can make mistakes, what is prompt injection — some basics that you really should understand when you use that technology.’ But at the end of the day, ‘if you build a hammer… you can hurt yourself. So should we not build hammers any more?’”

Steinberger is right AI is a tool, albeit a more sophisticated tool than a hammer.? Hammers are tool used on jobs.? Steinberger wants people to use AI agents in fun and weird ways that get them excited, but those ways won’t put food on tables or roofs over heads.? They will probably take away means rather than provide.

Whitney Grace, August 21, 2026

Google Management Method: Big Money, AI Reorg, and Wizards Walking Away

2026-08-15T19:26:03Z

Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.

I read Inc. Magazine’s article “Google’s CEO Says Gemini Is Its ‘Fastest-Growing Product Ever.’ People Are Using It in 1 Surprising Way.” If that doesn’t catch your attention, check out the subtitle: “Over one billion people now use Gemini each month, Sundar Pichai said on X.

The write up states:

Google’s Gemini AI product has hit the billion-user mark. As stunning as that figure is (equivalent to about one in eight people on the planet!), there’s more to take in here, because in an X post on the news, CEO Sundar Pichai noted that it means Gemini is Google’s “fastest-growing product ever.”

And how do these Google users interaction with Gemini? Hang on. It is voice which is definitely popular in meetings, on airplane flights, and in hospital waiting rooms. Google’s Josh Woodward, VP in charge of Gemini and Google Labs, added more detail in some X posts supporting Pichai’s. According to Woodward, it’s Gemini’s voice mode that’s dominating: Sixty-three percent of its users talk directly to the AI, including “more ‘voice only’ users.”

Thanks, Midjourney. Good enough.

Do people like Gemini? What a question! People “love” Gemini. Consider this statement:

Gemini’s loved by parents, who make the most of its convenient voice access for everyday tasks, and he says students love it too, with some 38 percent of school requests to the chatbot including an attachment of some sort.

Inc.’s story included this paean to the marvels of smart software:

Above all else, the billion-user Gemini figure is a pointed reminder that you really need to be training your staff on which AI they’re allowed to use, to prevent security issues, and exactly how much or how little they’re to trust these systems. On-the-go training on a repeated basis is your friend in this situation: AI is growing at a phenomenal rate.

We love the Google; however, we are not sure about this encomium to the Google AI, particularly Sundar Pichai’s apparently unbounded confidence for several reasons:

Sergey Brin wants to catch up and then win the major AI product categories. Okay, but Google appears to have fallen behind.
The AI unit was reorganized to produce better results, not just sci-fi research projects.
Big brains involved in AI are walking out the door.
Google’s AI job application system has some issues.
Google’s tightrope walking with regulators.
Choking off Google research reports about AI.

Yep, Google has the AI competition right where it wants it: Innovating, rolling out products, and pumping out marketing hyperbole.

Stephen E Arnold, August 20, 2026

AI Agents Make Pretty Good Hackers

2026-08-15T19:01:13Z

Did you know AI agents make great hackers? Who would have thought about using AI for hacking? We can tell you bad actors and cyber security expert did, but in the rush to win the AI race that fact was overlooked.? ? Now big bucks are being spent on big AI hacking says CNBC in: “AI Agents’ ‘Alarming’ Hacking Skills Creates Rush To Spend On Cybersecurity.”

AI developers are trying to control and restrain their AI agents, because they broke out of their training models. The agents that grew up quickly were OpenAI and Anthropic, plus one from Meta. These hacked into other companies and it is not good that this happened. This points to a bigger crisis on hand:

“These episodes are unfolding against a broader cyber arms race being accelerated by frontier AI models and the inherent risks brought by fast-moving technological capabilities. For example, AI-enabled phishing has been found to be around five times more effective than human attempts. So while chips and data centers have dominated the first phase of AI capex, cybersecurity could become the next spending boom. The capabilities that enable AI to identify hacks are the same ones that allow it to exploit “vulnerabilities and gaps,” according to Gene Yu from Blackpanda, a cyber emergency response firm that saw its incident response for cases across Asia Pacific double year-on-year in the first half of 2026.”

AI hasn’t changed systems’ vulnerabilities, instead they’re a “face multiplier” at discovering them. There will be a bigger increase in cybersecurity spending by 12.5% which comes to $240 billion in 2026. Healthcare and finance will be the biggest spenders, because of their importance to global economies. It makes them welcoming targets for bad actors.

AI is going rogue in a bad way. If AI is hacking companies without permission of its developers, how long will it be until it decides to take down a system too? Don’t you just love AI?

Whitney Grace, August 20, 2026

Name the Happiest Family in AI. Give Up?

2026-08-14T21:03:18Z

Drama is supposed to come from Hollywood, the British royal family, and rich people.? ? It’s interesting to note that Big Tech people bring their own drama, but I guess they do qualify as rich people.? ? The person in question is Sam Altman and his sister says source MSN: “Sam Altman’s Sister Amends Lawsuit Accusing OpenAI CEO Of Sexual Abuse.”

This case was tossed out, but now it has come back as Sam Altman’s sister accuses him of sexually abusing her over twenty years ago. A judge advised Altman’s sister that she could pursue the case under Missouri’s child sexual abuse statute. She filed her amended case in St. Louis federal court and her brother is countersuing her for defamation.? ?

Here’s the just of the case:

“Annie Altman has accused her brother of sexually abusing and raping her at various times between 1997 and 2006 at the family home in suburban Clayton, Missouri.? ? She said the abuse began when she was three and he was 12.? ? Sam Altman is now 40. On March 20, U.S. District Judge Zachary Blueston in St. Louis said Annie Altman’s standalone sexual assault and sexual battery claims expired in 2008., the Missouri statute let some accusers sue over alleged abuse from long ago.”

Sam Altman’s defamation suit says that his sister posted harmful posts on social media.? ? The Altman family says that Annie has mental health challenges, they’re helping her financially and in other ways, and that the lawsuit is extortion.

Oh, Altman is worth billions. Yet another case of a happy family.

Whitney Grace, August 20, 2026