Tech industry has always been somewhat vulnerable to success theater. I worked through the service oriented architecture mania where our company turned FTP batch scripts into services running on IBM's BPEL monster that was about .001% as efficient as the original FTP script.
This was done only so the company had marketing material to claim "success" with these expensive tools.
Now we have the AI mania, but it is the biggest mania of them all. Nothing compares. The collapse is going to be equally devastating.
If it aint broke force IT staff to replace with a flashy new product from a well funded and well connected vendor that gives the C suite under the table bennies and strokes their egos. Then the execs can create a narrative that they're tech leaders while driving their staff to thoughts of kicking them down the stairs.
We really need to be better about asserting and protecting our rights as professional employees with valuable skills.
There's also the robotics mania, which seems to be picking up steam again. Watch as everyone needlessly freaks out screaming Terminator over humanoid robots - the least efficient of robot designs. I have no doubt robots, including humanoid robots will continue to advance, but it's folly to compare them to AI. For the foreseeable future, you're gonna be seeing humanoid robots as overpriced clunky hardware that may just keep your Roomba company. Moreover, right to repair and supply chains will make these bots a pain to actually troubleshoot and maintain as they age.
Companies have always engaged in "success theater". It's a big part of the game. For the bigger companies, it is practically the CEO's only job. What is perhaps different in the AI realm is the gullibility of the audience. A car company that claimed its new model will reach 500 mph would be laughed out of town. When an AI company hints that their next version might be AGI, they can get away with it because AGI is a vague concept and the media environment is full of loose talk about how AI is going to take our jobs, cure cancer, etc.
I'm sure some are but I think most don't know enough to evaluate the claims. I'm still seeing posts on social media talking about what AI's "want" coming from otherwise educated people.
The Anthropic CEO said at Davos this year that by 2030, AI will double the human lifespan. I have no idea how the entire audience did not laugh hysterically. What an amazing comedy show.
Partly that seems to be because we are being drowned out in such claims, from doubling the lifespan, to 48 cancer vaccines, to AI doing better than all humans on all things, and what not. The hyperbole has been normalized like much else in our social discourse these days unfortunately.
For some examples, check out my post here (under bullet 5 of "Are we asking the right questions" section towards the end):
Success theater is a great description for the GenAI hype train.
When some CEO finds a technology that actually gives their company a significant productivity boost, they don't immediately run to Twitter to post about it. That tecnology becomes something that drives their competitive advantage. Why would they advertise it to their competitors?
Gary, do you think it's worth your writing about how much of "AI" talent is spent on gussying up responses with personalization ("That's an interesting question, Stuart") and tweaking prompts and emulating writing styles in ways that are not direct raw outputs of the models?
Thanks for your investment in learning so much about AI and its investors.
I've had AI responses change over the course of a few days, from obviously incorrect output (Reggie Bush wore number 15) to correct. There appears to be some process in place to keep tweaking the output.
Just yesterday I googled "hide your light under a bushel" and was told, "It suggests that people should share their skills and gifts with the world." Whoopsie!
ServiceNow may be the canary in the coal mine here. If they can't boost productivity as advertised, their sales don't grow, and the whole thing can collapse. Down 12% so far this year.
I've had to use it in two different places. Either it's one of the worst ticketing systems I've every used, or both locations implemented it really badly. And they have lots of competition in that market, so I'm not sure they're a good canary. Canary might have had a pre-existing condition.
klarna walking back on ai was hilariously "success-theatred". Not acknowledging that support sucked because of the chatbots, the narrative was success-theatred to something like: "chatbots provided the capital to hire more humans"
To this, I say "good". I want the oligarchs to have to pay people. I don't want all money to flow to a handful of psychopaths who are hellbent on imposing their messianic, pseudo-religious ideas (which is what the entire question of AI is to me) on the rest of us, to their enrichment, while we are enslaved to the techno-feudalists.
I want Silicon Valley to fail. I want their hubris, arrogance, and wrong ideas to be laid bare for the world to see. I want their sinister intentions to be laid bare as well, and I want it to force the rest of us to ask ourselves, "What really matters, what can we really control, what should we really control, and what are our real choices?".
Whether public or private, there appears to be a strong tendency to get on board with the hype or else have your public stock price/private valuation negatively affected....and none of them want that.
Marcus, is the smooth development of AI being hampered by the abundance of lies and contradictions that are enmeshed in the data that are in publications, statistics, etc?
Imagine all that money or even a small fraction of it going into STEM for students of all backgrounds. The result would be a far more predictable improvement in engineering productivity as well as a happier, more contented engineering community.
All this reminds me of the typical new tech + professional management exploding cocktail:
1. New promising tech with a lot of hype.
2. Executives see an opportunity to cut headcount and let the remaining staff 'optimise' with the great new tech.
3. Cuts happen without really changing anything about the company.
4. Service levels slowly decline but are covered by the established nature of a company and legacy lock-in.
5. Slowly new-ish companies or established companies that leveraged the new tech better start to eat at their market share (often just by luck, rarely by planning).
6. Executives start hiring again rebuilding previously functioning teams and the cycle repeats...
For some reason, I keep thinking about the Agile hype cycle: "oh, all our teams are Agile now, so we don't need project managers any more!" (fires all the project managers) ... 5 years later - realises that the company has achieved much less and paid much more (mostly on project managers called other things).
The hype is so strong, it's so easy to fall into a doom cycle countering the hype. There is (some) value in the tech but the desire to be contrarian is soooo appealing...
Thank you Gary. Awesome post and insights as always!
Success Theater is part of a love triangle involving Vendors, Customers, and Analyst firms that specialize in business technology.
Especially at the enterprise level, Analysts can be valuable to Customers because the products are complex, difficult to understand, expensive to buy, and it’s costly—in dollars and lost careers—to recover from a bad decision. Vendors participate because they’re exposed to more Customers, they gain intelligence on their competition, and they learn what prospective Customers want. I played for Team Vendor from 1978 to 2014, often meeting with Analysts and Customers.
Analysts own the bed; Customers and Vendors pay Analysts for access to the knowledge that Analysts have gathered from… those very same Customers and Vendors. Everyone has fun.
Trusted advisor? Rentier? Extortionist? Decide for yourself, but Analysts run a garden-variety neoliberal market-based business model: create some IP, wall it off, and sell (I mean license) access to it.
Success Theater (noun: polite synonym for bullshit) derives from everyone’s need to somehow display their prowess in bed. Vendors and Analysts alike offer exposure and potential career advancement to ambitious Customer executives by dangling opportunities such as speaking slots at the conferences they control. In exchange, they extract Success Theater quotes, which are (almost) always written for the Customer.
But it isn’t often that things go as far off the rails as with this one, mostly because enterprise Vendors offer genuine products that perform as advertised.
Ain't no such thing as "generic logic." There are logic systems comprised of a set of axioms, connectives, modifiers, and arbitrary formulas. As an incomplete list: universal logic, existential logic, propositional logic, predicate logic, the logic of predicate functions, first order logic, Aristotelian logic, multivalued logic, e.g. Fuzzy logic, paraconsistent logic, Kripke Semantics, mathematical logic, meta-mathematical logic, modal logics, constructive logic, informal logic, Boolean logic, topological logic, Indian logic, Chinese logic.
I just got a Samsung S25 (the latest version) and I can confirm what you're saying. The predictive text on it is pure garbage. The way that it tries to override me too is enraging. The old Samsung flip phone I used 15 years ago was heads and tails better.
This is remarkable, isn't it? What impresses me is that the predictive natural language processing I dealt with has gotten noticeably more loused up over the last few months. Unmistakable.
Tech industry has always been somewhat vulnerable to success theater. I worked through the service oriented architecture mania where our company turned FTP batch scripts into services running on IBM's BPEL monster that was about .001% as efficient as the original FTP script.
This was done only so the company had marketing material to claim "success" with these expensive tools.
Now we have the AI mania, but it is the biggest mania of them all. Nothing compares. The collapse is going to be equally devastating.
💯
If it aint broke force IT staff to replace with a flashy new product from a well funded and well connected vendor that gives the C suite under the table bennies and strokes their egos. Then the execs can create a narrative that they're tech leaders while driving their staff to thoughts of kicking them down the stairs.
We really need to be better about asserting and protecting our rights as professional employees with valuable skills.
Sounds like we might want to organize together. We can call ourselves the "White Collar Wobblies".
There's also the robotics mania, which seems to be picking up steam again. Watch as everyone needlessly freaks out screaming Terminator over humanoid robots - the least efficient of robot designs. I have no doubt robots, including humanoid robots will continue to advance, but it's folly to compare them to AI. For the foreseeable future, you're gonna be seeing humanoid robots as overpriced clunky hardware that may just keep your Roomba company. Moreover, right to repair and supply chains will make these bots a pain to actually troubleshoot and maintain as they age.
It's all there in Mackay's 1852 "Extraordinary Popular Delusions and the Madness of Crowds."
(tl;dr - fools and their money are eventually parted)
Companies have always engaged in "success theater". It's a big part of the game. For the bigger companies, it is practically the CEO's only job. What is perhaps different in the AI realm is the gullibility of the audience. A car company that claimed its new model will reach 500 mph would be laughed out of town. When an AI company hints that their next version might be AGI, they can get away with it because AGI is a vague concept and the media environment is full of loose talk about how AI is going to take our jobs, cure cancer, etc.
what’s (somewhat) different here is the customers are playing along with the ruse
I'm sure some are but I think most don't know enough to evaluate the claims. I'm still seeing posts on social media talking about what AI's "want" coming from otherwise educated people.
The checks and balances seem to have disappeared but on close examination, even the LLM providers’ narratives don’t seem to add up:
https://www.linkedin.com/posts/mohakshah1_anthropic-wef-ai-activity-7295826612162347009-ohq3?utm_source=share&utm_medium=member_ios&rcm=ACoAAAccQbQBcVe0DcVYj9qIUfaIPC_bWHoxHFc
https://www.linkedin.com/posts/mohakshah1_anthropic-openai-ai-activity-7297623397952327680-dj5b?utm_source=share&utm_medium=member_ios&rcm=ACoAAAccQbQBcVe0DcVYj9qIUfaIPC_bWHoxHFc
Once we try to connect the dots, there seems to be gaping holes
The Anthropic CEO said at Davos this year that by 2030, AI will double the human lifespan. I have no idea how the entire audience did not laugh hysterically. What an amazing comedy show.
https://www.pymnts.com/artificial-intelligence-2/2025/anthropic-ceo-sees-ai-powered-advances-doubling-human-lifespans/#:~:text=Anthropic%20CEO%20Sees%20AI%2DPowered%20Advances%20Doubling%20Human%20Lifespans&text=Anthropic%20CEO%20Dario%20Amodei%20said,get%20this%20AI%20stuff%20right.%E2%80%9D
Partly that seems to be because we are being drowned out in such claims, from doubling the lifespan, to 48 cancer vaccines, to AI doing better than all humans on all things, and what not. The hyperbole has been normalized like much else in our social discourse these days unfortunately.
For some examples, check out my post here (under bullet 5 of "Are we asking the right questions" section towards the end):
https://mohakshah.substack.com/p/the-deepseek-saga-are-we-taking-the?r=224koz
Success theater is a great description for the GenAI hype train.
When some CEO finds a technology that actually gives their company a significant productivity boost, they don't immediately run to Twitter to post about it. That tecnology becomes something that drives their competitive advantage. Why would they advertise it to their competitors?
Gary, do you think it's worth your writing about how much of "AI" talent is spent on gussying up responses with personalization ("That's an interesting question, Stuart") and tweaking prompts and emulating writing styles in ways that are not direct raw outputs of the models?
Thanks for your investment in learning so much about AI and its investors.
I've had AI responses change over the course of a few days, from obviously incorrect output (Reggie Bush wore number 15) to correct. There appears to be some process in place to keep tweaking the output.
Just yesterday I googled "hide your light under a bushel" and was told, "It suggests that people should share their skills and gifts with the world." Whoopsie!
See: "OpenAI Used Kenyan Workers on Less Than $2 Per Hour to Make ChatGPT Less Toxic"
It was only through human intervention the various OpenAI chat bots were prevented from spewing incorrect output.
I am shocked—shocked!—to find that gambling is going on in here!
ServiceNow may be the canary in the coal mine here. If they can't boost productivity as advertised, their sales don't grow, and the whole thing can collapse. Down 12% so far this year.
I believe in the case of AI, it is spelled “ conAIry”
“He’s got the whole world…in his hands”
I've had to use it in two different places. Either it's one of the worst ticketing systems I've every used, or both locations implemented it really badly. And they have lots of competition in that market, so I'm not sure they're a good canary. Canary might have had a pre-existing condition.
klarna walking back on ai was hilariously "success-theatred". Not acknowledging that support sucked because of the chatbots, the narrative was success-theatred to something like: "chatbots provided the capital to hire more humans"
Success Theater?
Let's just call it out as the bullshit it is.
To this, I say "good". I want the oligarchs to have to pay people. I don't want all money to flow to a handful of psychopaths who are hellbent on imposing their messianic, pseudo-religious ideas (which is what the entire question of AI is to me) on the rest of us, to their enrichment, while we are enslaved to the techno-feudalists.
I want Silicon Valley to fail. I want their hubris, arrogance, and wrong ideas to be laid bare for the world to see. I want their sinister intentions to be laid bare as well, and I want it to force the rest of us to ask ourselves, "What really matters, what can we really control, what should we really control, and what are our real choices?".
Whether public or private, there appears to be a strong tendency to get on board with the hype or else have your public stock price/private valuation negatively affected....and none of them want that.
Keep up the healthy skepticism, Gary!
Marcus, is the smooth development of AI being hampered by the abundance of lies and contradictions that are enmeshed in the data that are in publications, statistics, etc?
Imagine all that money or even a small fraction of it going into STEM for students of all backgrounds. The result would be a far more predictable improvement in engineering productivity as well as a happier, more contented engineering community.
Keep up the great work. I love success theater
All this reminds me of the typical new tech + professional management exploding cocktail:
1. New promising tech with a lot of hype.
2. Executives see an opportunity to cut headcount and let the remaining staff 'optimise' with the great new tech.
3. Cuts happen without really changing anything about the company.
4. Service levels slowly decline but are covered by the established nature of a company and legacy lock-in.
5. Slowly new-ish companies or established companies that leveraged the new tech better start to eat at their market share (often just by luck, rarely by planning).
6. Executives start hiring again rebuilding previously functioning teams and the cycle repeats...
For some reason, I keep thinking about the Agile hype cycle: "oh, all our teams are Agile now, so we don't need project managers any more!" (fires all the project managers) ... 5 years later - realises that the company has achieved much less and paid much more (mostly on project managers called other things).
The hype is so strong, it's so easy to fall into a doom cycle countering the hype. There is (some) value in the tech but the desire to be contrarian is soooo appealing...
Thank you Gary. Awesome post and insights as always!
Success Theater is part of a love triangle involving Vendors, Customers, and Analyst firms that specialize in business technology.
Especially at the enterprise level, Analysts can be valuable to Customers because the products are complex, difficult to understand, expensive to buy, and it’s costly—in dollars and lost careers—to recover from a bad decision. Vendors participate because they’re exposed to more Customers, they gain intelligence on their competition, and they learn what prospective Customers want. I played for Team Vendor from 1978 to 2014, often meeting with Analysts and Customers.
Analysts own the bed; Customers and Vendors pay Analysts for access to the knowledge that Analysts have gathered from… those very same Customers and Vendors. Everyone has fun.
Trusted advisor? Rentier? Extortionist? Decide for yourself, but Analysts run a garden-variety neoliberal market-based business model: create some IP, wall it off, and sell (I mean license) access to it.
Success Theater (noun: polite synonym for bullshit) derives from everyone’s need to somehow display their prowess in bed. Vendors and Analysts alike offer exposure and potential career advancement to ambitious Customer executives by dangling opportunities such as speaking slots at the conferences they control. In exchange, they extract Success Theater quotes, which are (almost) always written for the Customer.
But it isn’t often that things go as far off the rails as with this one, mostly because enterprise Vendors offer genuine products that perform as advertised.
Ain't no such thing as "generic logic." There are logic systems comprised of a set of axioms, connectives, modifiers, and arbitrary formulas. As an incomplete list: universal logic, existential logic, propositional logic, predicate logic, the logic of predicate functions, first order logic, Aristotelian logic, multivalued logic, e.g. Fuzzy logic, paraconsistent logic, Kripke Semantics, mathematical logic, meta-mathematical logic, modal logics, constructive logic, informal logic, Boolean logic, topological logic, Indian logic, Chinese logic.
I just got a Samsung S25 (the latest version) and I can confirm what you're saying. The predictive text on it is pure garbage. The way that it tries to override me too is enraging. The old Samsung flip phone I used 15 years ago was heads and tails better.
This is remarkable, isn't it? What impresses me is that the predictive natural language processing I dealt with has gotten noticeably more loused up over the last few months. Unmistakable.
I was watching this very process in action this afternoon.
I'll spare you the details, but I'm growing convinced that natural language processing is getting crappier by the day. Way worse.