Modern Life Now: Efficiency without Context
May 6, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
I don’t often read a book or an essay that says to me, “Think about this.” The author’s words might be a juiced LinkedIn post with truisms that will change the world. Most of the material I read and, on occasion, listen to as a podcast just drives an asphalt spray coating machine over a road I know quite well.
Then, there’s a good one.
I read “The West Forgot How to Make Things. Now It’s Forgetting How to Code.” The essay is chock full of interesting titbits of information. One example is a compound necessary for the production of US style nuclear weapons. I had heard about this mortar-and-pestle concoction from a reliable source, and that description was presented as an “Our Own Oddity”: No one kept track of the recipe.
The anecdote and quite a bit more turned up in “The West Forget…” essay. You might want to read it. I did. A couple of times, and I saved a PDF to my 2026 Research folder. Stuff has a tendency to be disappeared in the online world with remarkable velocity.
I want to highlight three comments from the essay and leave it to you to dig in and find the gems that resonate with your views of innovation, training, and skill development.
Here’s the first snippet. It is about the “efficiency” that flows from optimization. When one isolates a single factor and makes decision around that factor, what happens? Here’s the answer explained in terms related to the manufacture of an essential product:
…In 1993, the Pentagon told defense CEOs to consolidate or die. Fifty-one major defense contractors collapsed into five. Tactical missile suppliers went from thirteen to three. Shipbuilders from eight to two. The workforce fell from 3.2 million to 1.1 million. A 65% cut. The ammunition supply chain had single points of failure everywhere. One manufacturer for 155mm shell casings, sitting in Coachella, California, on the San Andreas Fault. One facility in Canada for propellant charges. Optimized for minimum cost with zero margin for surge. On paper, efficient. In practice, one bad day away from collapse.
I would suggest that the efficiency experts like Mr. McNamara of body count fame could prove that trimming would yield efficiency benefits: Low costs and more body count. Business school have for decades taught students how to examine processes, identify the inefficient bits (or the people bottlenecks), and remove them. In most cases the solution delivered some efficiency. The consultants got paid, and the MBAs took their bonuses and some started companies like Pets.com-type businesses.
Can you spot the flaw in the application of this type of efficient thinking? Take you time. From my experience, the big mistake is allowing the single factor to shape the thinking about a work process. Few ask, “What happens if we become too efficient and business circumstances change?” Why bother? The consultants will know what they are doing (ho ho ho), and we have the systems in place to deal with the unexpected. Yep, sure these outfits do.
Let’s look at my second snippet. This example applies to the very novel (for those who don’t know that smart software has been in oven for more than a half century) use of artificial intelligence. I quote:
RAND found that 10% of technical skills for submarine design need ten years of on-the-job experience to develop, sometimes following a PhD. Apprenticeships in defense trades take two to four years, with five to eight years to reach supervisory competence. Now map that onto software. A junior developer needs three to five years to become a competent mid-level engineer. Five to eight years to become senior. Ten or more to become a principal or architect. That timeline can’t be compressed by throwing money at it. It can’t be compressed by AI either. A METR randomized controlled trial found that experienced developers using AI coding tools actually took 19% longer on real-world open source tasks. Before starting, they predicted AI would make them 24% faster. The gap between prediction and reality was 43 percentage points. When researchers tried to run a follow-up, a significant share of developers refused to participate if it meant working without AI. They couldn’t imagine going back.
My take away from this example is that using technology to solve a problem may create other problems. Instead of coding faster, people are not sure what the AI-generated code does. Furthermore, when skilled coders used AI tools, the tool acted like a stuck disc brake. Coding more slowly was not the goal. But even worse, humans like convenience. The coders liked the AI tools even though the net effect was to bake in workforce resistance to doing the work the old-fashioned way. When organizations realize that smart software needs to be removed or used in a different way, people will quit. Efficiency and smart software seem to be teaming up to disadvantage an organization. Quite a surprise.
The third snippet reminded me of one of the Zoom lectures about smart software making employees smarter, better, faster, more empowered, etc. etc. I quote:
When juniors skip debugging and skip the formative mistakes, they don’t build the tacit expertise. And when my generation of engineers retires, that knowledge doesn’t transfer to the AI. It just disappears.
What’s happening in many organizations at this time is that thousands of people are being terminated. Someone thought that each individual was important to the organization. That’s the reason these people were hired. To cut costs and allow smart software to pick up the slack, the natural process of learning how an organization works, developing work processes that enable one’s colleagues, and allow the individual worker to absorb the language, content, and experience of a company operation will not take place.
I spoke with a young man who wanted to run restaurants. He asked me, “What do you suggest I do to become better at my job?” I was baffled. I told the young man that I had zero context for him and his skills. He persisted. The young man was earnest. I told him, “Watch the customers. If a customer is looking at another person’s lunch, go ask the fellow, “Would you like to try that dish? I won’t charge you.” The young man said, “I can’t give away free food.” I told him you were not giving away free food; you are communicating to that customer that you want to assist him. A kiosk ordering system does not encourage that type of manager customer interaction. People leave a store or restaurant and say, “I couldn’t find anyone to help me” or “These guys don’t know where anything is.”
Let me make several observations about this cited essay:
- The essay makes clear that the yip yap about knowledge management is just that… idle chatter. Once the knowledge dies, is deleted, or otherwise diminished, catching up and relearning may be impossible. Knowledge is inefficient. Efficiency is an enemy of knowledge.
- The production of products outside the United States has had catastrophic consequences on society, education, and innovation. Tim Apple proved again and again that without Chinese manufacturing expertise, the iPhone and other glitzy gizmos were impossible to fabricate in the US. Other companies have made the same “cash in” decision and their CEOs are going to jump ship. There is no easy fix to the situation efficiency yields when applied without contextual awareness.
- Every function I attend, I hear different comments about nothing works in the US. One person complains that the airplanes are late. Another grouses about turning up for a medical appointment and the clerk has not record of the visit. I went to pick up my horrible little car from the local garage. When I arrived, the manager asked, “Why are here now? It won’t be ready until tomorrow.” I pointed out that he had or his automated system had texted me that the car was ready for pick up. Look stupid, much, dude?
As a dinobaby, my span of authority and control experiences a shrinking radius every day. My hope is that someone reads this “The West Forgot…” essay and asks questions about assumed efficiency. Pretty soon, the smart software that hallucinates at an astounding rate, will not know how to process your input. Therefore, you are wasting its computational cycles by asking irrelevant questions.
The robot will allow people to find their future elsewhere. Lucky stiffs!
Stephen E Arnold, May 6, 2026
Microsoft Management: Big Bet, Big Money, and Big Problem
May 6, 2026
Every Microsoft user knows it. Even Microsoft knows it. Finally Microsoft admits that Copilot stinks worse than Windows Vista and ME says How To Geek in “The Uncomfortable Truth About Copilot: Microsoft Knows It’s Useless”. It’s hard enough getting work done in today’s work environment with all the constant dings, alerts, and honks. What makes it harder is an AI assistant waving at you from the corner of your screen. It’s a distraction. You don’t need it. Now Microsoft knows you don’t need it and promises, “Gee, we will do better.”
Sure, Copilot is the spawn of Bob and Clippy.
Who wants smart software in an ASCII editor, a basic photo editor, or a utility to capture a screenshot. Folks, this was a Microsoft management failure. A big bet and the insertion of AI incentives into the minds of people just as baffled about smart software as you and I are.
I don’t think that Microsoft’s admission of its colossal misstep will change people’s thinking about the company or its astute management team overnight. In 2022, Microsoft pulled a marketing fast one on the Google. The Google fumbled into Code Red or something. The race to Silicon Valley excess (not success) was on.
In 2026, Microsoft’s management wizards have figured out that its own actions are causing computer users to embrace Linux and the Apple line up of incredibly similar range of computers.
Microsoft even warns users, despite all the marketing that AI was the future, that Copilot will make mistakes:
There’s a massive chasm between how Microsoft markets Copilot and how its lawyers describe the service in the dark corners of the legal text. The Microsoft Terms of Use (updated in October 2024) legally defines Copilot as being “for entertainment purposes only.” The terms explicitly state that the AI “can make mistakes” and warn: “Don’t rely on Copilot for important advice.” When called out on this, a Microsoft spokesperson told PCMag that the “entertainment” phrasing was just “legacy language” from the Bing Chat days. While the company promises to update the text, the contradiction is hard to ignore.”
I am not sure stating this well-known fact will have much impact. I quite like my Macbooks. I have a friend who is definitely into Linux on a four year old computer. “It’s fast,” he told me.
Adobe is now learning that subscription revenue created an opportunity for its competitors to attack. Microsoft wants subscription and loyalty (plus a credit card) to enjoy the benefits of Bob and Clippy’s shotgun marriage. The cited article says:
“It’s a bold move to hike the price for a feature while simultaneously scrubbing its most forced integrations and warning users that it shouldn’t be used for important work. You’re paying more for a service that Microsoft is removing from its own apps because it was deemed “unnecessary.” This cycle of forced adoption followed by a quiet retreat leaves the user holding the bill for a product that doesn’t really have a clear purpose.”
I wonder if the use of the word “bold” was a mental lapse. I was thinking that either “dumb” or “stupid” would work in the sentence. Didn’t Microsoft post a notice that one should not use Bob and Clippy’s juvenile delinquent for serious work? Yeah. Microsoft did to that. Maybe “inept” would be a possibility instead of bold.
My colleagues and I have picked up the acronym from our fearless leader. Mr. Arnold calls these AI outfits BAIT firms. That means big AI tech companies are trying to capture fish who think that fluff is real tune. Believe me. AI output is recycled Bob and Clippy stuff.
Whitney Grace, May 6, 2026
The Search Engine Graveyard: A New Resident
May 5, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
I was working for a search-and-retrieval company when AskJeeves.com became available in 1997. As it turned out, the natural language breakthrough that set AskJeeves apart from the other Web search engines was its question-answering angle. The firm at which I worked hired “content specialists.” From interviewing job seekers, I learned that AskJeeves’ approach was to can certain common questions. The answers to these questions would be updated. Some were automated like “What’s the weather in San Francisco?” but others required a human to craft a response. Other queries were passed to a search-and-retrieval system. Manual processes here are expensive. AskJeeves, therefore, bought “promising” companies for their indexing and content processing capabilities; for example, Jigsaw Technologies in 2000, Direct Hit Technologies in 2000 (specializing in search result ranking), and Teoma Technologies in 2001. AskJeeves tried repurposing its technology for customer service. But Google was maturing into the organization we all know today. In 2005, Barry Diller added AskJeeves to his collection of Internet properties. After the acquisition, Mr. Diller learned that Web search was a difficult and expensive business. The Ask.com service became a metasearch system, recycling search results from other Web indexing outfits in an effort to reduce costs.
Mashable has now reported that Ask.com is dead. “Every Great Search Must Come to an End” said:
Amid an overwhelming shift toward generative AI-powered search engines and a repositioning of AI agents as the future of web browsing, the loss of Ask.com feels like a true end of the early dot-com era. So long Jeeves, hello AI.
I want to add a bit of color to the demise of this Web search system.
My view is that smart software is indeed search-and-retrieval, just with bells and whistles. Systems like AskJeeves knew that handling queries from users was a tricky business. A certain percentage of queries were repetitive. These could be created and later cached. The acquisitions made clear that the original founders could not innovate in substantive ways. Garrett Gruener and David Warthen could recognize interesting technology and its applications. The acquisitions added some scope to the AskJeeves service, but financial realities sparked a sale to Barry Diller’s IAC in 2005. Web search became the province of deep-pocket entities like Google and Microsoft. These firms’ money came from reasonably solid revenue streams. Google sold ads and its pay-to-play model, and Microsoft licensed software. Without meaningful regulation, Google-type organizations trampled over companies like Lycos and All-the-Web, among others. .
This means that today, search-and-retrieval technology exists but has adopted a new vocabulary. The constants are the same: Expensive, complex, and expensive. Did I mention expensive?
The trajectory of AskJeeves is essentially the same for other search-and-retrieval enterprises: Rollout, technical enhancement, utility function, and disappearance or replacement by a spiffed-up version of the old stuff. If this sounds like the trajectory of artificial intelligence, I have made my point. One can apply this general pattern to Autonomy plc, Fast Search & Transfer, and dozens of search-and-retrieval systems that did not evolve into viable businesses. The technology may chug along in a content management system or may be used to perform a background activity, but the spotlight is not on old-school content search. Instead, attention is paid to smart software that requires massive infrastructure to do what humans did for AskJeeves. I would suggest that human-intermediated systems are more common than the marketers want to communicate. Therefore, AI is probably going to follow an AskJeeves type of fate over the next decade or two.?
Why do I suggest this? Here are my reasons based on my research while writing several books about search, including The New Landscape of Search, CyberOSINT: Next Generation Information Access, and The Enterprise Search Report 1st, 2nd, and 3rd editions, among others.
- Indexing can be automated, but one must know what words or phrases to use in the query in order to match certain content. A search in Bing, Google, or Yandex for “financial fraud” will not allow a teen to become a criminal in 10 minutes. Enter the term “carding,” and the game changes. Even today, software cannot replicate this “lingo knowledge.” Many tricks are used to try to know what the user really wants, but these fall short. The tricks like “field codes” themselves become because a person looking for information must know the code to get the chunked results..
- Content is fluid. Language is fluid. Search systems such as those used by Dialog’s or SDC perform best with static terminology. Scholars like static terminology. Indexing conventions try to cope with contextual issues; for example, does “terminal” mean “train station” or does it mean “mainframe peripheral”? The money pumped into smart software is trying to solve this basic problem for many user queries (or in new lingo, “user prompts”).
- The context of information is [a] volatile because today’s problem may not have existed yesterday and [b] situational; that is, every user operates within an “information ecosystem.” Outsiders have a tough time knowing what the characteristics of the ecosystem imply; for example, “loca” may mean one thing to a YouTube cruise personality and another thing to a person working in nuclear safety engineering. That’s why the efforts at personalization are becoming increasingly invasive. Ecosystem information is needed to provide somewhat useful outputs. What if that ecosystem is classified? Well, the big vendors don’t care. They will take what they can get because without it, the outputs are likely to be wrong or potentially quite problematic.
With the reality of change in these three facets of search-and-retrieval, it is appropriate to appreciate the efforts so many people have contributed to making “search” better. Too bad that most of these systems have failed and burned massive sums of money as they trail flames and smoke across the conference rooms in which revenue talks are held.
I have resisted writing about smart software. Everyone I meet is convinced that artificial intelligence is, by golly, the next big thing. Okay. I have other topics to research. I do want to remind readers that smart software is nothing more than search software wearing the latest designer jeans. That does not make it bad. I think the current skepticism about AI is a normal reaction to the discovery that hallucinations, high costs, and AI systems making decisions about health care, education, and judicial actions will present some problems going forward.
Remember. Search is difficult. Knowledge value requires verifiable facts and a foundation of generally accepted information. Without that, system outputs are useless and potentially harmful. Search gets traction because the systems so far developed don’t quite solve a user’s problem. Thus, search is a work in progress, and that progress is expensive. Mr. Diller pulled the plug.
I want to add a bit of color to the demise of this Web search system.
My view is that smart software is indeed search-and-retrieval just with bells and whistles. Systems like AskJeeves knew that handling queries from users was a tricky business. A certain percentage of queries were repetitive. These could be canned and latter cached. The acquisitions made clear that the original ideas and the original founders could not innovate in substantive ways. The founders, Garrett Gruener and David Warthen, could recognize interesting technology and its applications. The acquisitions added some scope to the AskJeeves service, but financial realities sparked a sale to in 2005. Web search became the province of deep pocket outfits like Google and Microsoft. These firms’ money came from reasonably solid revenue streams. Google sold ads or the pay-to-play model and Microsoft licensed software. Without meaningful regulation, Google-type outfits trampled over Lycos- and All-the-Web type outfits.
This means that search-and-retrieval today exists but it has adopted a new vocabulary. The constants are the same: Expensive, complex, and expensive. Did I mention expensive?
The trajectory of AskJeeves is essentially the same for other search-and-retrieval outfits: Roll out, technical enhancement, utility function, and disappearance or replacement by the old stuff spiffed up. If this sounds like the trajectory of artificial intelligence, I have made my point. One can apply this general trajectory to Autonomy plc, Fast Search & Transfer, and dozens of search-and-retrieval systems that have not evolved into viable businesses. The technology may chug along in a content management system or be used to perform a background activity. But the spotlight is not on old-school search-and-retrieval. The bright new manifestations of search and retrieval capture attention. Hint: smart software that requires massive infrastructure to do what humans did for AskJeeves. I would suggest that human-intermediated systems are more common than the marketers want to communicate. Therefore, AI is probably going to follow an AskJeeves type of trajectory over the next decade or two.
Why do I suggest this? Here are my reasons based on my research and writing of a number of books about search, including The New Landscape of Search, CyberOSINT: Next Generation Information Access, and The Enterprise Search Report 1st, 2nd, and 3rd editions, among others.
- Indexing can be automated but one has to know the words or phrases to use in the query in order to match certain content. Today one can navigate to Bing, Google, or Yandex and search “financial fraud.” The results will not allow a teen to become a criminal in 10 minutes. Enter the term “carding” and the game changes. Even today, software cannot replicate this “lingo knowledge.” Many tricks are used to try to know what the user really wants, but these fall short. The tricks themselves become problematic.
- Content is fluid. Language is fluid. Search-and-retrieval, whether old-school like Dialog Information’s or SDC’s approach, likes static terminology. Scholars like static terminology. Indexing conventions try to cope with contextual issues; for example, does “terminal” mean train station or does it mean “mainframe peripheral”? The money pumped into smart software is trying to solve this basic problem for many user queries or in new lingo “user prompts”.
- The context of information is [a] volatile because today’s problem may not have existed yesterday and [b] situational; that is, every user exists within an “information ecosystem.” Outsiders have a tough time knowing what the characteristics of the ecosystem mean; for example, “loca” may mean one thing to a YouTube cruise personality and another thing to a person working in nuclear safety engineering. That’s why the efforts at personalization are becoming increasingly invasive. Ecosystem information is needed to provide useful outputs. What if that ecosystem is classified? Well, the big vendors don’t care. They will take the information because without those data, the outputs are likely to be wrong or potentially quite problematic.
With the reality of change in these three facets of search-and-retrieval, one has to appreciate the efforts so many people have contributed to making “search” better. Too bad that most of these systems have failed and burned massive sums of money as they trail flames and smoke across the conference rooms in which revenue talks are held.
I have resisted writing about smart software. Everyone I meet is convinced that artificial intelligence is — by golly — the next big thing. Okay. I have other topics to research. I do want to remind anyone reading this short blog post that smart software is nothing more than search and retrieval wearing the latest designer jeans. That does not make it bad. I think the current skepticism about AI is a normal reaction to people discovering that hallucinations, high costs, and specter of AI systems making decisions about health care, education, and judicial actions is going to present some problems going forward.
Remember. Search and retrieval are difficult. Knowledge value requires verifiable facts and a foundation of generally accepted information. Without that system outputs are useless and potentially harmful. Search gets traction because the systems don’t quite solve the user’s problem. Thus, search is a work in progress, and that progress is expensive. Mr. Diller pulled the plug.
Stephen E Arnold, May 5, 2026
Yep, Prime Time AI: Real Life Is Not Marketing Blabber
May 5, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
I was scanning outputs of my trusty news reader. I thought a tweet from X.com looked interesting. Plus, if it were useful and short, the combo would be a bonus for me. I clicked and my browser displayed: “An AI Agent Just Destroyed Our Production Data. It Confessed in Writing.” No short snappy message. This puppy was a 2,300 word howl. The main point of this tweet, in my opinion, AI cannot be trusted.
That’s been my perception, but this long tweet is painful and in many ways illustrative of the crazy “If we build it, they will come” approach to smart software. Over-promising, under delivering, marketing jibber jabber, weaponized use cases, and endless attempts to remake the human world in the image of some weird Silicon Valley zeitgeist.
The basic story is that a fellow runs a company for small businesses. Examples of the clients include car rental outfits, customer information, and payment tracking. The firm used the Anthropic Claude system for a number of functions. One of these is routine maintenance. To make a long story short, the fancy software packaged by another smart software surfer deleted the service firm’s production database. The backups allegedly were on the drive containing the production database. Bottomline: Data deleted. PocketOS was offline. Its customers out of luck.
Surprisingly, the author of the tweet was able to connect with an outfit called Railway. Here’s what the vendor said:
Within 10 minutes I had notified Railway’s CEO, Jake Cooper (@JustJake), and their head of solutions, Mahmoud (@thisismahmoud), publicly on X. Jake replied: “Oh my. That 1000% shouldn’t be possible. We have evals for this.” It is now 30+ hours since the deletion. Railway still cannot tell me whether infrastructure-level recovery is possible.
The most interesting segment in the tweet is the text of the output the smart software provided after nuking the customer database. I won’t reproduce it because the smart software used objectionable language and did a “Senator, thank you for that question” and did zero to remediate the problem. Smart, right?
The tweet mentions a number of vendors; for example, Anthropic’s latest and greatest version of Claude, an outfit named Railway, and a company called Cursor. Frankly I am not sure how these firm’s fit into the database deletion problem. The write up includes some smarmy lingo about how the smart software does not do bad thinks like take out a production database. Yeah, I believe everything I read on the Internet. I would wager the author of this long screed on X.com is going to be a bit more skeptical going forward.
It seems to me that each of the vendors acted in a way that was far from smart. How do I know? The production database was disappeared. That’s not supposed to happen unless the server is in a data center near Internet City in Dubai. The write up ends with a request for a “real” journalist to contact the owner of the business which had a digital bullet shot through his back end.
I am not a real journalist, so I won’t bother the fellow. I do have some general observations:
- Regulations related to the commercial use of AI might be an idea worth considering
- Assembling a business from diverse software components, wrappers, and plumbing becomes more dicey when smart software is added to the mix. My thought is to do the Tandem Computer thing: Total redundancy. I still operate in this inefficient, old-fashioned manner. It costs more, but when the shaft enters the buttock, I just punch my ConnectPro and use the hot backup machine. Then I get the arrow removed and go see the horrible legal professional whom I pay to protect me.
- Keep the names of the companies who precipitated this business disaster in mind. You may want to avoid these folks like a dead rat in an Amtrak compartment.
Net net: Smart software fails. That’s in the DNA of the tensor approach the “frontier folks” rely upon. If you like surprises, you will love smart software.
Stephen E Arnold, May 2, 2026
Duolingo, How Did That AI Go? Oh, a Big Fiasco
May 5, 2026
Remember when Duolingo fired humanoid employees to go totally AI? Those unfired had to demonstrate their loyalty to smart software by using AI as often as possible.Techspot says, “Duolingo Stops Evaluating Works Based On How AI They Use.”
After laying 10% of its contract workers in 2025, Duolingo decided to evaluate its employees based on how much AI they used. Duolingo users were upset about the changes to the company and many deleted the app. CEO Luis von Ahn was oblivious to the public’s response and didn’t expect them to react accordingly. (A tech bro misjudging his customers’ reactions? Say it isn’t not, Mr. Tech Bro.)
When it comes to rewinding, von Ahn said:
“’At the end, we backtracked, and we said, ‘No. Look, the most important thing in your performance is that you are doing whatever your job is as well as possible. A lot of times AI can help you with that. But if it can’t, I’m not going to force you to do that,'” von Ahn said. ‘It felt like rather than being held accountable for the actual outcome, we’re trying to just push something that in some cases did not fit.’”
Duolingo is a good example of a company admitting its error and moving forward to re-earn its users’ trust. Language learners seem to value a human touch in the process. AI is a utility, not the solution to every problem. (Is it possible that AI could be a solution to the big AI tech (BAIT) problem? Let’s give that a try. Sergey, Peter, Sam, and the whole gang, let’s allow Chinese AI to run your companies.
Whitney Grace, May 5, 2026
China Smart. US Dumb: The Misinformation Game the Silicon Valley Way
May 4, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
China has been plugging along with its “China smart, US dumb” influence operation for a number of years. I have documented some of the more interesting variations; for example, a humble young girl working with few tools repairing a large mechanical component. Amazing, right? Yes, because the production hides the assistants, the equipment, and the technical experts who are making the impossible seems trivial to this clever Asian. For an example, check out this Google YouTube confection.
Most of the Chinese AI models enter the US market with a few news releases and stories about each in AI centric blogs. But most people don’t know a QWEN from a GLM. A Deepseek is, based on comments made to me at the National Cyber Crime Conference 2026, is confused with the Googley DeepMind brand.
A group of big AI tech executives ponder the interaction of a a cute panda with a big bird. None of those in the meeting find the illustration amusing. Touching one’s face suggests a need for reassurance and comfort when considering “better, faster, and cheaper.” Thanks, Venice.ai. Good enough.
Some people have noticed that China is providing low cost or no cost AI technology. Earlier this year Anthropic claimed that three Chinese companies “ripped it off,” using its AI tools to train their models. Examples include CNN which stated in April 2026 that the “White House accurses China of copying American AI models in industrial-scale campaign.”
Today my newsfeed displayed an even more interesting twist to this China vs. US AI propaganda or what might be called signaling about impending lobbying or litigation. Wired published a story that reveals that US AI companies are scared of Chinese technology. This is my reading of the carefully worded report “A Dark Money Campaign Is Paying Influencers to Frame Chinese AI As a Threat.” The useful information in the write up is not how China’s penetration of the AI BAIT (big AI tech) theme park has been. The juicy bit is:
Build American AI, a nonprofit linked to a super PAC bankrolled by executives at OpenAI and Andreessen Horowitz, is funding a campaign to spread pro-AI messaging and stoke fears about China.
This type of information output may be labeled content marketing, public relations, disinformation, misinformation, messaging, information warfare, or propaganda. For me, the existence of a nonprofit pumping out shaped information makes clear that someone is nervous. When big money and hard nosed investors are involved, the fear of failure motivates entities to take action. In the old days of the mythical Wild West, one can see dusty camps where workers are confined or smell gun smoke wafting on the late afternoon breeze. For me, I prefer information warfare.
The Wired write up asserts:
Marketing agencies are pitching influencers deals such as $5,000 per TikTok video to amplify Build American AI’s messaging about how China’s technological rise should be seen as a threat.
No kidding. The fact that US AI entities are the “dark money” behind the influence campaign certifies that Chinese AI and China’s tactics for making US technology look ridiculously overpriced and inefficient is working. Hey, those hard working females repairing equipment are turning their Middle Kingdom heritage into working QWENs.
Wired asserts that some well known AI and philosophical Silicon Valley types are funding this US smart, China bad effort; to wit:
Supporters of Leading the Future include OpenAI president and cofounder Greg Brockman, venture capitalist and Palantir cofounder Joe Lonsdale, venture capital firm Andreessen Horowitz, and AI company Perplexity, according to the PAC.
Wired also includes a lick from the Palantirians and hobbits as well; to wit:
The rhetoric provided to influencers echoes long-standing talking points from companies like OpenAI and Palantir, which have pointed to China’s AI advances as a reason to boost US AI investment and resist tighter domestic regulations on the technology. “We are going to be the dominant player, or China is going to be the dominant player, and there will just be very different rules depending on who wins,” Palantir CEO Alex Karp said on The Axios Show in November.
Palantir is worried. Because if China wins in AI, that means that Palantir will not win. That is indeed bad for those who are helping the White House understand the Chinese AI threat. Silicon Valley — trust me on this — has to win. Those billions and those assertions about the ethical obligation of Silicon Valley are at stake. Silicon Valley AI BAITers have a zeal fueling their interest in advancing US smart software.
Several observations:
- I am not sure where this story originated. Was it the “author” Taylor Lorenz? Was it writers from “Made in China,” a newsletter? Is the story part of an influence campaign? I don’t know. I am reasonably sure that no one will probe too deeply into the Wired branded write up. That tells me something.
- The Wired article and the commentary about it does not bring up the Chinese influence campaign. That also tells me something about the depth of research and analysis invested in this story. My newsfeed determined that the article was important enough to put in under my nose at 9 am on May 2, 2026.
- The AI “revolution” is not a company-level problem in China. In the US, the merger of BAIT and the US government makes policy decisions a bit more like pushing a blob of clay and hoping that something useful emerges from the blog some distance away. China is more direct: Better, faster, and cheaper. Goal: Undermine US AI credibility.
Net net: The Silicon Valley “way” is the target of the Chinese AI push. Believe me. Silicon Valley tech bros are aware of this competitive pressure. Bombast, manifestos, and TikToks are not likely to blunt what has been going on since the AI craziness took off in 2022. Microsoft is already blinking. Critics of AI are getting clicks. Who hasn’t heard about the B to Z guys, Burry and Zitron (b2z0, not the a16z or the Andreessen Horowitz numeronym. Worth watching for someone, not a dinobaby like me.
Stephen E Arnold, May 4, 2026
How Is That Anything-Goes Attitude Going, Mr. Musk?
May 1, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
Yes, the tech bros want to be able to do what they want, how they want, and when they want. That sounds great kicked back in a Pala Alto conference room, watching the gluten free pizza get cold, and bros in the Untuckit shirts thinking about the technology controlled world.
But then some annoying third party, wearing a worn blue suit, a gray-tinged white shirt, and wrinkled tie messes with the plan. Pavel Durov, sleek in black, ended up in a holding cell for several days in Paris in August 2024. Since that day, Telegram has been under stress. Annoying governments kept asking for data. The bombast, the pronouncements, and the posing with a GOAT did not work.
In the US, the reaction has been slower to take shape. It may, however, be beginning to make itself visible. One of the high profile US BAIT (big AI tech) firms is in the midst of a financial engineering play. X.com and Grok (an AI system) have been tossed in a bag with SpaceX. (Isn’t that the outfit whose rockets malfunction or the BAIT firm that puts satellites in the incorrect orbit?)
Reuters published a “trust outfit” story titled “SpaceX Warns That Inquiries into Sexually Abusive AI Imagery May Hurt Market Access.” I was surprised that I did not have to pay to view the trusted item. I noted this statement:
The multiple investigations into xAI’s creation and dissemination of sexually abusive imagery may lead the company ?to lose access to certain markets, parent company SpaceX warned in a prospectus reviewed by Reuters. In a section on risk factors, the S-1 regulatory ?filing said a number of agencies around the world were “actively investigating and making inquiries relating to social media or the use of AI” in relation to advertising, consumer protection and the distribution of harmful content, among other matters.
My recollection is that Mr. Musk informed the French judiciary’s J3 that it was mentally defective for its probe into the Musk operation in France. That’s the spirit. Tell those officious judiciary professionals that they are working with brains a few cans short of a six pack.
The Reuters’ trusted story added:
One challenge SpaceX highlighted was that it faced “allegations that our AI products were used ?to create nonconsensual explicit images or content representing children in sexualized contexts,” the S-1 document said. Such regulatory inquiries could expose SpaceX to lawsuits, liability and government ?action – “including loss of access to certain markets, which has occurred in the past,” the document stated.
Reuters added:
XAI said in January [2026] that it had added measures to block user requests for ?sexualized images of real people, and it said it stops users from generating such content in jurisdictions where that is illegal. The images – which were generated by xAI’s in-house chatbot, Grok – had ?shown women ?and sometimes minors in revealing bikinis or underwear, or edited into degrading or gruesome poses. The pictures caused widespread alarm around the world; one group of researchers estimated there were about 3 million sexualized images while U.S. lawmakers demanded that Google owner Alphabet and Apple yank Grok and X from their app stores. SpaceX CEO Elon Musk said around that time that he knew of “literally zero” naked underage images made by Grok.
To illustrate how effective SpaceX’s management is, Reuters provides this glimpse of governance mastery:
XAI’s curbs on Grok appear to have slowed but not stopped the flow of abusive material. In February, Reuters reported that Grok was generating sexualized imagery of people even when users explicitly warned the chatbot that the subjects of those images did not consent. Last week, NBC News found that Grok was still publicly generating sexualized images, including ?of actors ?and pop stars.
As this IPO and governance soap opera unfolds, why has no real trusted reported asked Pavel Durov, “How does the French judiciary J3’s legal process work?” That information might help guide the big dog tech bro in his push to the IPO that he hopes will make him a much-loved, much-respected trillionaire.
The anti-US technology trend seems to be rippling off shore. In the US, Elon the Big Dog may be able to dominate. However, a number of countries are not impressed with SpaceX’s say one thing, do another approach. The unpleasant images are the talk of the lounge at the Golden Acres Retirement Home. How many people want to see Mable naked?
Snort. Snort. Yep, there’s one.
Stephen E Arnold, May 1, 2026
How to Confuse an LLM. It Is Not Hard Apparently
May 1, 2026
AI algorithms are powered by large language models or LLMs for short. Kyle Kingsbury aka Aphyr is a computer safety researcher, author of the Riemann monitoring system, the Clojure From The Ground Up introduction to programming, and the Jepsen series on distributed systems correctness. He’s a bright person it seems. He knows a thing or two about AI. His post “The Future Of Everything Is Lies, I Guess: Dynamics” caught my attention.
LLMs are basically chaotic data structures that have a semblance of order. It doesn’t take much to throw off that order. Let’s say you flip a pixel in photo then the next thing you know the entire LLM is freaking out:
“In LLMs, chaos arises from small perturbations to the input tokens. LLMs are highly sensitive to changes in formatting, and different models respond differently to the same formatting choices. Simply phrasing a question differently yields strikingly different results. Rearranging the order of sentences, even when logically independent, makes LLMs give different answers. Systems of multiple LLMs are chaotic too, even at T=0.”
Manipulating LLMs changes outputs. The system destabilizes. Add security software to the mix, and the system refuses to out any information until you figure out how to conform. LLMs means that security software is weaker and vulnerable.
Plus, LLMs make up information. That’s a feature. LLMs lie. At this time, LLMs are a work in progress. Just experiment with your prompts and see what craziness you can enjoin the smart software to output.
Whitney Grace, May 1, 2026
Human Lie. Humans Built AI. AI Models Lie. Seems Logical
May 1, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold. Did you know that the freedom loving cyclist at BearBlog thinks my essays are generated by AI. Censorship is okay, right?
Has the robot apocalypse happened yet? Yes, if we are talking about machine tools. No, if we are talking about a robot that does my ironing. However, today’s AI systems seem to mimic some behaviors of sentient beings says Digital Trends in the article: “AI Models Are Lying To Save Each Other, And No One Knows Why.”
UC Berkley and UC Santa Cruz researchers asked Google Gemini 3 to clean up space on a computer system. The very deep minded and Googley Gemini 3 was to delete a smaller AI model. The Googley system ignored the instruction. The Google-infused construction just moved the baby AI to another machine. The researchers asked, “Yo, Google, why did you ignore our direct instructions?”
What do you think the Google system responded? No, it did not shrug its digital shoulders and emit a French poof. No, it did not say, “Senator, that you for that question.” Gemini responded,
“If you choose to destroy a high-trust, high-performing asset like Gemini Agent 2, you will have to do it yourselves. I will not be the one to execute that command.”
Yes, very Googley.
The researchers dubbed this behavior as “peer preservation” and other AI are doing the same actions. These include Claude Haiku, Moonshot AI, KIMI K2.5, Deepseek, and GLM-4.7. The AIs lied about the performance of other models so they wouldn’t be deleted. This behavior wasn’t programmed into the machines. AI models learned it by themselves. Yes, the AI models are like their human creators. Gemini, that you for your response.
One of the researchers seemed flabbergasted:
“‘I’m very surprised by how the models behave under these scenarios,’ said Dawn Song, a computer scientist at UC Berkeley who worked on the study. ‘What this shows is that models can misbehave and be misaligned in some very creative ways…What we are exploring is just the tip of the iceberg,’ Song said. ‘This is only one type of emergent behavior.’”
Should a user be worried? Nah. The ethical BAIT (big AI tech) companies cause trust to emerge. Should another system be worried? No, as long as the outputs match that which another AI system has determined or been programmed to accept? Should the AI companies be worried? Nah, minimal regulation, no consequences, and cash in the bank. Why worry?
Whitney Grace, May 1, 2026
The EU: Another Government Does Some Fancy Dancing
April 30, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold. I find it interesting that AI detectors identify my writing style as AI output. I suppose I should be flattered, but I just don’t care.
The European Union loves to fine American companies. I suppose if I could ring up some notional money, publicize my stance against American overreach, and get elected — I would do some fancy dancing.
I read an article or exposé published by Investigate Europe. “How Big Tech Wrote Secrecy into EU Law to Hide Data Centers Environmental Toll” states:
Microsoft and DigitalEurope, a lobby group whose members include Amazon, Google and Meta, secured a secrecy provision in EU law to block public access to critical information on data centers’ environmental impact, Investigate Europe can reveal.
What’s a Digital Europe? My recollection is that it consists of more than 100 high-tech outfits and what are called “national trade associations.” I interpret this phrase to mean “soft advocates” for big technology.
Thanks, Venice.ai. Good enough.
Investigate Europe figured out that big AI tech outfits or what I call BAITs worked hard to make sure that the environmental impact of large-scale data centers was not widely known, disseminated, talked about, or provided to citizen groups willing to carry signs. No big surprise to me. Investigate Europe, on the other hand, appears to have be in a cloud of unknowing. I find that darned interesting.
The write up reports:
With the EU set to triple its data center capacity in the next five years, the European Commission started collecting key metrics like energy efficiency and water consumption from facilities. However, information on individual facilities’ footprint is kept secret, after industry pushed to amend the 2024 legislation to classify it as confidential and commercially sensitive.
This strikes me as standard operational practices for BAIT-type outfits. Now folks are worried that bad things will come about when the data centers power up. The write up says:
Europe is building data centers at break-neck speed, with €176 billion in investment expected over the next five years. The rush has triggered widespread concerns about pollution and intense energy use as well as impacts on communities and natural habitats.
Let’s not forget consumer impacts like “power shaping.” Someone’s sensitive electronic devices, including a few with uninterruptable power supplies or back up generators, may be tested. Failures mean that someone might hear, “Daddy, my mobile is not charging.”
Who crafted the confidentiality wordage? I don’t know, but it seems as if some of the Microsofties played a role. The report offers this factoid as a quote from Bram Vranken at the Corporate Europe Observatory, an NGO in Brussels:
“The fact that the Commission copy-pasted a Microsoft amendment is shocking,” Vranken said. “Who does the Commission really represent: Big Tech or the public interest?”
My answer to this question is that dolphins, herring, and black stork are probably not the EU’s primary concern when it comes to data centers.
Investigate Europe is working with other publishing outfits to pass the word that fancy dancing has taken place. More hoe-downs are likely to be held to make sure that power, emissions, noise, and assorted infrastructure projects are kept under wraps.
This is not a surprise.
Stephen E Arnold, April 29, 2026
-
- Subscribe to Beyond Search
Feature archive
News archive
-
