Microsoft May View AI As a Super App
June 4, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
One of the simplest tricks in debate has a bunch of names. Game players like to say “pop up a level” or just “level up, dude.” Those with exposure to the rigors of traditional indexing say, “Just put it in a different cluster.” Silicon Valley types infused with the zeitgeist of incomprehensibility offer, “Do the meta play.” A rose by any other name still smells like an old age home on Sunday morning.
A great thinker wonders if a rose is actually a cat. Perhaps he can sell this concept and make lots of money? Thanks, MidJourney. Good enough.
The concept is that one has a thing and one wants to make that thing appear more profound, new, zippier, or imbued with qualities a person unfamiliar with “level upping” does not understand, see, or comprehend.
I want to point to an example described in “Exclusive: Microsoft Is Building a Super App That Combines Coding, Chat, and Other Copilot AI Tools.” (I like the exclusive part because by my count there are about a dozen outfits building super apps; however, as a dinobaby, I am generous. I think the “exclusive” is that no other big time real business news outfit has ever reported on Microsoft doing the clustering play or the “pop up a level, dude” tactic.)
Be that as it may, the “exclusive” write up reports:
The software giant is working on a one-stop shop that would connect its GitHub Copilot coding assistant, Copilot chat function, Copilot Cowork tool, and a new agentic workflow capability internally named Autopilot into a single app, according to two sources familiar with the project, who spoke on the condition of anonymity to discuss a platform that hasn’t yet been released.
Yep, a meta-app that will deliver a “cohesive product.” Isn’t this an admission that after three years of AI innovation, users of Microsoft software, hardware, and services are generally confused about what Copilot does what and how. The effort to data has caused a number of Microsoft users to want a way to turn off AI everything in Microsoft “experiences.” Some third parties have been paid big bucks to make the Copilot thing more obedient. But for most users, AI has begun to morph from “gee, this is cool” to “gee, this is a bit of a problem.”
The write up continues:
There may also be a toggle function for a user to go back and forth between their personal and enterprise 365 Copilots. A user will still be able to access their Copilots outside of the super app.
If this statement is indeed true, doesn’t this suggest that Microsoft is just foolin’? There will be many Copilots, not a single super app like the ones available from a Chinese outfit or from the alleged criminal Pavel Durov of Telegram and VKontakte fame. I won’t mention the others because then I will be revealing some of my upcoming lecture for the cyber fraud folks in Virginia in a few weeks. Some of my information, in my dinobabyish opinion, is indeed “exclusive.”
How lost in AI weirdness is Microsoft? May I suggest that this statement from the Fortune article helps answer the question? Here you go:
Microsoft in the past year has undergone one of the largest corporate reshuffles in its history that has included a string of high-profile departures and reorgs throughout its businesses. In April, it announced its first-ever employee buyout offer, aimed at its most long-tenured employees. At next week’s Build conference, Microsoft AI Chief Executive Mustafa Suleyman is expected to unveil new proprietary AI models. Suleyman, who once led consumer Copilot, has focused on models since the restructuring in March.
This is not recalibration in my view. This is scrambling and then trying to figure out how to put the eggs back in their shells.
Can Microsoft level up? Will the company survive the other super app developers? Will there be a reframing of the Microsoft AI strategy that includes security, defense against open source options, and the predatory instincts of Googzilla-type organizations in the US and from — yes, it is really true — other countries?
Confident in your answer? Hit those prediction markets.
Stephen E Arnold, June 4, 2026
Google AI: Speed Means Googlers Do Not Know What Is Happening to Users
June 3, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
I just read “Gemini User Hits 5-hour Usage Cap after a Single Prompt, Google Responds.” The article explains that the many changes to Google’s smart services are not understood by users or Googlers. Remarkably the article explains that no one knew AI usage limits would cause prompts to fail. Plus the write up includes a stunning quote from a Googler; to wit:
“Yikes, let us take a look!”
Okay, but it is clear from the story, that Googlers just released changes. No one took a “look.” Isn’t this interesting? Let me share some thoughts about why unintended consequences are rarely considered and why those involved in online products and services don’t bother to do what I would call “real work.”
Thanks, Midjourney. Good enough.
Don’t get me wrong. I understand the management problem. When I was in high school, I took classes at the University of Illinois. The lead professor (Dr. Shattuck, if I recall correctly) told us, “You are exceptional.” Who cared? We were living in a dorm, not a parent within miles. Then when I was talking with a smooth operator at the blue chip consulting firm, Ron Something said, “You are one in 10,000.” Both of these people were in the cow herd waste output business. I think the Google does the same thing. I had the thrill of being in a meeting at the Big Dog of Search when one senior manager said, “Here’s our GLAT.” I looked at the weird green document and replied, “It looks like the SAT. What is it?” The Googler laughed and said, “Some people who want to work here have to take the test because we’re Google.”
Yeah, right. In each of these cases, the people desperately want to convince themselves that they are gifted, the blessed ones, or psychologically off the norm. What happens in each of these “organizational constructs” that blend self-delusion with a task and those involved in the task is the following:
People do their think. These individuals know they are special. Therefore, just do it. One person with whom I worked at an outfit in Columbus, Ohio, loved the word “decider.” These people are deciders. What’s the result of an organization of deciders? Here are some characteristics:
- Most people in the organization have no idea what the other person in the adjacent cube or chair is doing.
- The rarely attentive middle managers whether in person or in some hippy dippy digital form like a Slack or a Zoom don’t know what questions to ask. Consequently, only a big win or a catastrophic screw up captures their attention. Most of the time, modern managers are clueless and “don’t have time” to be enriched by clues.
- The leadership of the organization floats weirdly in a bubble which blocks most negative or unpleasant inputs. Thus, most “leadership” slips into or actively embraces the “I’m special” thought process. Rules and social norms are not that important.
The Google craziness is not localized in Google. It exists, to greater and lesser degrees, at the BAIT outfits (BAIT is my lingo for big AI tech). These firms want to make the rules for everyone else. Most of them operate as if they were sovereign powers. This explains the disdain Elon Musk seems to display to the French judiciary’s criminal charge against X.com/Grok and why Meta’s cooperation with UK authorities is not exactly a priority when one’s firm is struggling in the AI World Cup.
The result, if my analysis is accurate even if based on my personal experiences, are situations like the one described in the article. The same approach has manifested itself in Microsoft’s crawfishing on some of its AI initiatives. The key point I want to make is that these type of management methods are now the norm. Whether it is the hapless owner of a Dodge RAM truck or a person who buys no name ear buds from an Amazon-type vendor, the new management methods virtually guarantee problems for a client, customer, or user.
Whom can one blame? Sorry. There is no one. The failure to do the basics of a “good job” have been displaced by the idea that superior individuals exist, can make decisions in a vacuum or a small circle of contacts, or just don’t care due to incentives. I am not blaming anyone, not even the surprised Google AI wizard. Welcome to the datasphere circa 2026. I am glad I am old.
Stephen E Arnold, June 3, 2026
AI: The Political Tool on Steroids
June 3, 2026
Dictators and other authoritarian bad actors use whatever tools they can for propaganda. Propaganda is an extremely powerful way to shape people’s minds. It’s only gotten more powerful with the advent of social media and now AI algorithms. Vox reports that there are new and exciting ways the manipulate people with AI: “The Hidden Way Dictatorships are Shaping What AI Tells You.”
Here’s an alarming fact: more than a billion people turn to chatbots for advice and information, erotica, robot-plagiarism, and other stuff. ChatGPT has 900 million weekly users.
As more people use AI chatbots, they will be able to shape how people perceive information and society. They’re already changing how people perceive things. The biggest fear is how bad actors could use AI chatbots for propaganda:
“This has generated fears about chatbots’ potential to spread state propaganda. Such anxieties generally center on the prospect of major AI labs consciously designing their LLMs to favor pro-regime perspectives while suppressing dissident ones. And there is some basis for this worry: The Chinese AI company Deepseek programmed its model to evade discussion of the Tiananmen Square massacre and other topics inconvenient to the Chinese Communist Party.”
How are governments responding to AI? I shudder to think.
Whitney Grace, June 3, 2026
Traditional Consulting: Becoming Just Less Average
June 2, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
I have an AI bias. My personal experience with different smart software systems is what I would call “mostly average.” Somewhere along the line, I learned about research that demonstrated the obvious. Run the same query set against a group of AI models. Run the queries every couple of weeks. At the end of six months. Look at the outputs. The person who explained this research to me wanted me to tell him what the test showed. I said, “Good enough similar outputs.” He looked at me and asked, “How did you know? Did you read the study?” I said, “The methods produce a good enough because that’s how the math works.”
An AI-centric management consulting firm learns a painful lesson. The work was just good enough but not good enough to prevent immediate termination. Thanks, MidJourney. You are indeed good enough.
I encountered this “good enough” when we worked on some projects for a then-famous enterprise search system based on Bayesian statistics. Train up the system, ask queries, and watch the precision and recall diminish over time. Re-train the system and one can observe the same degradation. Why? The answer is that Bayesian and Bayesian-related methods “drift.” Users have novel problems, but the model is not aware of the user’s context and how the words or other symbols operate in the unknown context. Therefore, the outputs have to converge. This is the engine of drift.
I thought about this “drift” and the convergence to the mean when I read “How AI Threatens the Giants of Consulting.” (Heads up: The story sits behind the weird orange newspaper’s paywall. Don’t complain to me. Just subscribe.) The article is interesting, but it did not touch upon this “good enough” angle, nor did it explain where the “value” comes from when hiring a blue-chip consulting firm. That’s okay. To an outside observer, one sees AI replacing a certain type of employee at a blue chip consulting firm. The obvious impact is staffing, billing, and how the work is completed.
The write up makes some good points; for example:
- Blue chip firms developed a system to scale to handle big projects. The write up does not explain what scale means in a large, global study on a volatile topic. Let me add some color: Low tier and mid tier consulting firms struggle, even with AI, to handle certain types of projects. AI would not be particularly helpful in the study of world economic change I worked on years ago. Consultants had to interview individuals in specific institutions and use that information to inform other components of the research project spanning a dozen countries. AI can do some things, but it cannot do this type of knowledge work. Maybe some day? But today to pull this original research off, one has to work for a blue chip outfit.
- AI makes it seem that anyone can set up, market, operate, and avoid litigation by become a next-generation consultant. Okay, maybe. Let me put in my two cents: Setting up an AI centric consulting firm or boutique and staying in business are two different things.
- Today’s consulting firm clients demand fee linked to outcomes, not just hours on time sheets. My observation: That’s true for certain engagements and certain types of firms. However, blue chip firms once enjoyed client repeat business in the 80 to 85 percent range. This means that the blue chip firms capture a client and can bill equivalent fees and present the invoice any way the client requires. A job for a Fortune 50 CEO may be a single line item: “Consulting services for April 2026, $2.4 million.” A government job uses pay grades, hours, engineering change orders or scope changes to generate an invoice in line with he funds appropriated for a project. For a very large client of a blue chip firm, a project can be buried in another on-going project for that firm and described as “research.” Billing has been flexible for as long as the industry has existed.
- AI enabled firms can be fleeter of foot and more like a Cirque du Soleil act. My view: If an engagement requires speed and agility, find a firm that has specific strengths. Blue chip consulting firms, even when AI enabled, have specific, proprietary systems and methods. Sure, a blue chip consultant can jump to another firm, but that transplant has to learn the specific systems and methods at the different consulting firms. Most of the consulting firm’s recipes contain quite specific ingredients in appropriate proportions. Most of this stuff is not taught in MBA school. Feel, style, and selection of ingredients are unique to each firm. That’s why clients of blue chip firms lock in on one or two blue chip firms for the majority of their high end consulting work. The firm delivers the dish the client likes without surprises.
What the article suggests is that the world of blue chip consulting will be roiled yet again. I don’t agree. Blue chip consulting firms did not become blue chip by being “good enough.” Delivering a report anyone can do with a prompt says more about the client’s expectations and the firm doing the work than relevant information about a blue chip firm.
The blue chip consultants have been affected by the corrosive effect of online information and modern technology. Modern business produces winners and losers. That means that firms disappear, get acquired, or decide to merge. The same forces apply to Web search and to the US auto industry.
There are what I call azure chip consulting firms or mid tier outfits. LinkedIn runs ads when one of the better known mid tier firms says it wants to hire an expert in technology X or business sector Y. These firms take work that is too small to be of interest to the blue chip firms. In Manhattan one can hire a research firm to do a study. The firm does okay work, and its clients usually don’t know the difference between what a blue chip firm would have delivered and what the mid tier outfit delivered. Mid tier consultancies have always been cheaper.
There are outfits like Gerson Lehrman-type outfits who “rent” former blue chip consultants to companies working on a budget. These organizations have been around for years and have not killed the blue chip consulting firms. My hunch is that these rental agencies will be able to deliver more value because the individual experts they rent will avail themselves of AI to make a one hour Zoom meeting more useful.
The idea that AI will light the fire of lower cost, AI consulting firms is accurate. However, most of these will not survive. That’s because start ups, regardless of business sector, fail at a rate that would make moms, friends, and clueless family offices sick to their stomachs.
I think the article is interesting. I am not sure that the angle of attack is matched to how the management consulting business operates and has operated for more than a century. I think Booz, Allen fired up in 1917 or so with a project for Sears (may it rest in peace).
AI is a tool. Management consultants are knowledge value professionals who use tools. A tool provides something of use to a skilled user. By itself, the tool does not do much of anything. It’s the know how that produces a hierarchy with the best performers at the top and the wanna-bes at the bottom. AI is good enough right now for some things. I am not convinced it is good enough for context sensitive, fast changing, dynamic things.
Stephen E Arnold, June 2, 2026
Google Showers Cash on Consultants: The “Useless” Professionals Are Important Now
June 2, 2026
While Google is cutting costs by firing employees, it is also spending $750 million to fund a Cloud Next 2026 endeavor. This is meant to advance its partners’ development of agentic AI applications, making this the largest single partner investment in history. According to The Next Web’s story: “Google Puts $750 Million Behind The Consultants It Needs To Close The Cloud Gap.” the firing of the money canon signals the start of another AI race.
The AI death match will shift from selling cloud infrastructure to financing systems integrators and consultancies that deploy Google AI. Agentic AI will create a $1 trillion global market and Google wants to own a large portion of it by making its partners the main delivery channels.
Here are more details about the fund:
“The fund is not a venture capital vehicle. It is a mix of credits, co-investment capital, training subsidies, and go-to-market funding designed to get the world’s largest consulting firms building agents on Google’s platform rather than on Microsoft Azure or AWS. The economics explain the urgency: for every dollar a customer spends on Google Cloud, partners capture up to $7.05 in services revenue, meaning the consultancies are not just a distribution channel but a multiplier of Google’s own cloud consumption. Google now counts more than 2,900 services partners, with a 400% increase in new partner entries over the past year and a 250% increase in partner-influenced revenue. The fund is a bet that accelerating the partner ecosystem is the fastest route to closing the market share gap with AWS and Azure.”
Google’s individual partners have added their own money to the fund. Accenture is expanding its Gemini practice and has already built more than 450 agents on Google Cloud. Deloitte has more than one hundred agents on and said the company made the largest investment in Google. KPMG said they’ll allot $100 million to the project, while PwC has $400 earmarked for Google Cloud. NTT Data and Cognizant have 5000 engineers working on agents specifically for healthcare, financial services, and manufacturing.
Google is investing this money with their partners in order to beat Microsoft in the enterprise. Microsoft has a structural advantage with Office 365 and it also happens to be distributed to Fortune 500 companies that Google wants to make their own. Sure, AWS is also a rival to Google, but Microsoft is the outfit the Googlers want to supplant.
Google knows it can run the world. Will the world go along with this McKinsey-inspired big idea? Eric Schmidt, the former CEO of Google, tried to talk about AI in a recent graduation address. He probably said Googley things amidst the boos. Is that a harbinger of Google’s AI vision?
Whitney Grace, June 2, 2026
Modern Management Methods: Zuck-ocles and His Threatening Pink Slip Sword
June 1, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
According to Google in its new and improved form, the Sword of Zuck-ocles — hey, sorry, I meant sword of Damocles — refers to an impending disaster. The operative idea is that the sword can fall at any time and kill a career, smash a planned vacation, or keep one’s child from attending an expensive private elementary school.
The Sword of Zuck-ocles fell on Wednesday, May 20, 2026, and severed an alleged 8,000 surplus humans from their pay check. The blow arrived by electronic mail, and I don’t think any of the former Zuck admirers expired. In fact, in an interesting post-termination write up, that old sword of Zuck-ocles is still out there and ready to go.
The Sword of Zuck-ocles is poised to strike. Thanks, Venice.ai. Good enough.
“Zuckerberg Tells the Tattered Remainder of His Workers That He Won’t Conduct Another a Mass Firing for at Least Seven Months” reports:
Meta employees are currently being forced to train agents expressly designed to do exactly what they do.
But don’t worry, Meta professionals. The article says:
Zuckerberg further noted in the memo that he doesn’t expect another large culling to take place this year. You know, the year that’s already close to halfway over. The not-particularly-comforting guarantee comes as morale at Meta is already wildly low.
In my long and completely undistinguished career, I have been exposed to a number of management methods and techniques. There was the crazed nuclear engineers replicating frantic work like that done for Fat Ralph or whatever the device was called. There was the cookie cutter “that’s how we have done dot points since 1917” approach of the blue chip consulting firm silly enough to hire me. There was the chaos now and tomorrow approach of the start ups with which I have been involved. There was the “everyone in the newspaper business does it this way” approach. And, of course, others.
I cannot recall what I would call the Sword of Zuck-ocles management tactic. The idea is that not one person, not a group, but thousands of people could be terminated in 200 days (maybe more, maybe less, who knows?) I suppose the idea is a good one. Obviously Meta leadership does not have any interest in old fashioned management methods; for example, communicating a specific goal, generating staff support and enthusiasm for that goal, and an fostering an organization-wide desire to solve customer problems.
Like the Greek story, the Sword of Zuck-ocles hangs over the heads of Meta professionals. Instead of being supported by a single hair from a horse, the Sword of Zuck-ocles is supported by marketing collateral about artificial intelligence. I know I would find working with assertions about the value of AI floating over me in my cube in Silicon Valley unpleasant. My reaction just underscores how out of touch this dinobaby is with today’s sophisticated management methods.
Stephen E Arnold, June 1, 2026
The AI – Catholic Church Issue
June 1, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
The New York Times published a short summary of five encyclicals. It is good to know that the NYT knows how to research the history of Papal actions. The story is paywalled, so paying, not praying, will reveal the truth. But the five encyclicals did not get to what I think is the crux of the matter or X marks the real spot.
Galileo Galilei makes it clear that he was very confused about the planets revolving around the sun and most of his other heretical thinking. House arrest made more sense to him than being burned at the stake. Thanks, Venice.ai. Good enough after three tries.
Not surprisingly, the Magnifica humanitas statement emphasizes guardrails, global regulation, and protection of humanoids. Anthropic’s “statement” was, and I summarize:
- Guardrails. Check
- Regulation. Check
- Protect humanoids. Check.
I interpreted the encyclical differently, probably because I worked on my graduate degree at Duquesne University (a Jesuit institution which like the idea of guardrails, regulation, and protection for humanoids. Believe it or not, I taught a couple of classes a requirement of my waived tuition, but I did get some money each month so I could enjoy a truly lavish lifestyle in Pittsburgh, Pennsylvania.)
The write ups about the encyclical did not talk about three interesting church actions. I will not talk about the Inquisition, but, please, keep that in mind if you doubt that the Catholic Church can take steps to clarify the thinking of certain of the church faithful.
I want to offer three examples of what happened when the tech bros did not line up with the Catholic Church.
First, Copernicus came up with the idea that lecturing and writing about his idea that the planets revolved around the sun deserved broad dissemination. The Church just grumbled but in 1616, that pot boiler De revolutionibus orbium coelestium was added to the Index of Forbidden Books.
Second, Galileo Galilei was into the Copernican concept. In 1633, he was tried by the Inquisition an found guilty. Galileo did the pragrmatic thing. He said, “I am sorry, very, very sorry.” The Church placed him under house arrest until he died.
The third example is one that calculus students don’t know much about. The Church determined that infinitesimals and by extension infinity bumped into Aristotelian philosophy. Those “teaching” about infinity were able to find jobs as farmers or buskers in Padua. Math bumped into this problem for decades until other issues pushed infinitesimals into an infinintely small segment of Catholic dogma.
My view of the encyclical about AI is, therefore, based on these historical actions. Thus, several observations can be offered:
- The Catholic Church can take direct and indirect action to suppress or cause information change
- Some of the methods tolerated by the Church involve indirect (house arrest) and direct (Iron maidens, heated fireplace pokers, etc.) to help individuals free their minds from certain thoughts
- The encyclical can spawn other statements that will be disseminated not by the Zuck-type or Telegram-type services. The message will be delivered to about 1.5 billion people. These indivdiuals with log on to the Zuck-type or Telegram-type services and comment about the Church message.
Net net: The Pope’s encyclical is a significant document and highly visible action. Additional communications and manifestations of the Pope’s guidance can be implemented with surprising ease and speed.
Stephen E Arnold, June 1, 2026
The New Dark Ages: A Recent Example
May 29, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
We are now entering the new Dark Ages. I see more posts about stupidism. Authors range from teachers in grade school, middle school, and high school. There are posts by recent college graduates about the impact of AI on college students. College professors grumble about students while some of that cohort use AI to crank out junk research. ArXiv has bluntly said that its editorial process will ban those who submit AI research slop. I like to toss in reminders that the former head of the Harvard ethic department resigned over allegations of recycling other people’s work. Several years ago the president of Stanford University resigned. Yep, intellectual short cuts. But that’s just a function of de-emphasis on reading, writing, and arithmetic and the 24×7 distraction datasphere. How can anyone think with the vivid “noise” pulsing through simple activities like walking. Who walks without a mobile device, a smart watch, ear buds, smart glasses, and a tracking beacon OR one of these gizmos. The Dark Ages thrive on stupidism.
Thanks, Midjourney. Art from your life experience I assume.
The other driver of the Dark Ages is ill-considered decisions. There are others, but I want to highlight one small example. The loss of a small firm’s data, reports, and models seems trivial. You can read the case example in “Disney Erased FiveThirtyEight.” Yep, the mouse is eager to contribute to stupidism.
The write up says:
It’s common to read things like: “what happens on the Internet stays on the Internet”, the notion being that you can never escape your digital past. But this isn’t really true. A Pew study of a random sample of Internet links conducted in October 2023 found significant “link rot”: almost 40 percent of links that had been active 10 years earlier were broken. And that’s probably an underestimate: the study was based on the Common Crawl web archive (the same one that AI labs use to train their models), which is quite comprehensive but probably contains some bias toward more prominent sites. Another study by ahrefs found a two-thirds attrition rate for web links after 11 years.
Online information is not a library filled with non-digital information objects. Sure, the mythical wonders of the library of Alexandria burned, but copies of some of the information persisted. Knowledge regained some traction, and the voids are unknown. But knowledge survived. As the quote makes clear, some information in digital form just goes away. A government agency takes down public information about MIC or LOCAs. A big company decides to get out of a business and tells the system administrator to take down some content. A careless programmer allows an AI agent to nuke a database. A less careless bad actor penetrates a systems created by a careless developer and vaporizes data. Fungible information artefacts can persist. The digital stuff just goes away.
The write up reports:
ABC finally fully shut down FiveThirtyEight in March 2025, 11 years after its debut at Disney. Eleven years is a long time in the media business, and the site covered one of the most tumultuous periods in American political history with its unique blend of analytics, brutal honesty and irreverence. It would be nice if that work could be preserved for the public record. I don’t know what plans Disney has for FiveThirtyEight, if any. But I did approach Disney a year or two ago, through my agent, about acquiring the remaining IP. I’m probably the logical high bidder, though the value is rapidly depreciating as what’s left of the site falls into disrepair. At a minimum, we’d restore the archive, with prominent links to Silver Bulletin. We were told to basically get lost: ABC was annoyed with my critical public comments about their management of FiveThirtyEight. It apparently wasn’t a long conversation, so I don’t have a lot more color to report than that.
I urge you to read the full essay. I do want to focus on my thesis of the new Dark Ages.
First, this case example exists within the stupidism datasphere; that is, business decisions, the application of math to certain topics, and the push for maximum returns (psychological, market share, monetary) results in the decision to delete information or remove it from access.
Second, the impact of stupidism is that it diffuses like spilling royal blue ink on a white sheet. The impacts goes in multiple dimensions and ruins things one assumed would be okay. Hey, it’s just a tiny spill. Sometimes not.
Third, as stupidism becomes institutionalized, working out of the Dark Ages takes time. The problem is that time today has been compressed. Centuries ago humanity had, well, centuries. Today, humanity may have less time. When there is zero way to see what has preceded and been verified multiple times, folks are flying bling and don’t know.
I like phrases like the Jazz Age. I wish I could think of a snappy way of saying, “Yep, it is the Age of stupidism.” I can’t. I will stick with new Dark Ages and stupidism.
Stephen E Arnold, May 29, 2026
AI Can Train Itself. Humans, Take Note, Please
May 29, 2026
Did you know there is a bunch of data that could be used to train AI models but is going un-captured and unnoticed? It is literally every query an AI application processes, especially the corrections. Production workflows are constantly making data that improves AI models but no one is taking advantage by putting it to use. VentureBeat explains that, “Enterprises Can Now Train Custom AI Models From Production Workflows-No ML Team Required.”
Impromptu AI released Alchemy Models with the premise that AI applications are already building their own training data and it’s not being harnessed. Alchemy Models captures the data, routing validated outpours from subjects matter experts into a fine-tuning pipeline that improves the model. Enterprises own the resulting data. Here’s more about how it works:
“It sits in different territory from both RAG and traditional fine-tuning. RAG retrieves external context at inference time without modifying model weights. Traditional fine-tuning changes weights but requires separately assembled labeled datasets and a dedicated ML pipeline. Alchemy does the latter continuously, using the enterprise application itself as the data source.
Companies adopting foundation model APIs face three compounding constraints: inference costs that scale with usage, no ownership of the models their data is effectively training, and limited ability to customize behavior for domain-specific tasks. Empromptu CEO Shanea Leven says those constraints are widely felt but rarely addressed.”
Most training processes require organizations to collect, clean data, and add metadata before training. Alchemy Models cleans the data itself. Here is more of the nitty-gritty:
“The mechanism runs through Empromptu’s Golden Data Pipelines infrastructure in two stages. Before an app is built, enterprise data is cleaned, extracted and enriched so the application starts with structured inputs. Once it is running, every output it generates goes back through the pipeline, where subject matter experts inside the organization review and correct it. That validated output becomes the training data for the next fine-tuning run.”
Governance is built into Alchemy Models so data goes through a process of compliance controls, guardrails, and evaluations within the same pipeline. This means that enterprises can use any data and transform it into training models. The training models become more accurate and cleaner with every use.
But what about the humans? You know the answer.
Whitney Grace, May 29, 2026
Has AI Failed Salesforce?
May 28, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
The Next Web published “Salesforce Is Selling the AI Future Harder Than It Is Delivering It.” This write up fits in the category of “AI Is a Loser.” From my point of view, AI is a utility. The problem is that hyperbole about a utility leads to disappointment. How many times have you worn a T shirt with the logo of the firm providing aluminum sulfate to your local water treatment plant operator?
An AI misstep can have some significant consequences. Thanks, Venice.ai. Good enough.
The write up doesn’t view AI as a utility software. I think the publication believes it bought into the AI hype and now appears to balance its old AI information with some “maybe it is an issue” coverage. Salesforce is a good case. Its forward-leaning sales pitches strike me as embarrassing. The Next Web article states:
The company has closed 29,000 Agentforce deals and claims $800 million in ARR, but its stock is down 30 per cent this year and the showcase demos keep turning out to be works in progress.
Even my team and I can make demos work. We define narrow functions. We shape the information. Then we make the system look like it is really pulling the cart around the circus ring. In the somewhat fluid and usually chaotic world of an operating organization, AI hallucinates, and “fixes” are not working.
The write up says:
SharkNinja, the maker of Shark vacuums and Ninja kitchen appliances, was another [Salesforce] headline customer. Salesforce said the company would use Agentforce to streamline customer service. Bloomberg reported a 20 per cent reduction in support calls as part of the pitch. But the deployment described was forward-looking, with agents expected to “guide customers through the buying process” and “manage returns,” not a report on outcomes already achieved.
Okay, “not a report on outcomes already achieved.” I interpreted this to mean that the Salesforce pitch was great. The real-life operation was a bit of a high-profile problem. Oh, oh. Another corporate president gets squeezed because of AI jazz talk.
To make clear that Salesforce has created a problem for itself, the story reports:
Salesforce’s financial trajectory adds another layer. Revenue growth has slowed from roughly 25 per cent a few years ago to about 10 per cent in fiscal 2026, when the company reported $41.5 billion in total revenue. That is still a large business, and the company delivered a strong fourth quarter with 12 per cent growth. But the deceleration is exactly what investors fear when they hear that AI agents will compress the number of human users who need software licenses.
Yeah, bubbling under the surface is the small matter of pricing. I would suggest that the blue chip consulting firms are staring at this brick wall as well. It costs a lot of money to replace a high roller client. Careers are ended when the AI initiative craters a traditional model young CEOs don’t understand or respect.
What’s Salesforce’s future look like? I think there will be lower cost, good enough options to replace Salesforce. My thought is that few in Salesforce’s leadership will agree with my statement. That’s okay. But I think this write up from The Next Web is just the first of others that document the issues surfacing with a utility that screws up a function instead of working well and reducing costs, thus attracting more customers. The case studies of the slip betwixt cup and lip will become more prevalent. Then they will stop. Utilities have to work. AI hallucinates, makes stuff up, and creates code that may detonate at an unknown point in the future in a way that no one wants to think about.
At Salesforce, the commitment to AI has been made. Stakeholders hope that the delivery of promised features and functions will happen. And once deployed, these smart services make the customers really happy. Otherwise, there will be a very large, very conspicuous office building in the San Francisco area with dark windows.
Stephen E Arnold, May 28, 2026
-
- Subscribe to Beyond Search
Feature archive
News archive
-
