The AI Hater's Manifesto
We recorded 42 claims from this post. 15 were checked against evidence. 27 can’t be checked. 25 more are held. Read the post at its source.
Claims from this post (42)
In the order they appear in the post. Each quotation is the speaker’s words in this post. A label such as “Supported” or “Can’t be checked” is the claim’s current assessment; its page has the evidence and the reasons.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“I would never, ever trust this thing that deleted and added lines of code at random with anything mission critical, I could not trust software built with it, and I certainly couldn’t trust it with anything involving my personal data.”
Our reading: Zitron says he would never trust the LLM tool he used, which deleted and added lines of code at random, with anything mission critical, could not trust software built with it, and certainly could not trust it with anything involving his personal data.
No horizon
Scope: the tool he used, which deleted and added lines of code at random
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Here’s how I feel about LLMs. In a vacuum, they’re an interesting technology that can do some interesting stuff, in the right scenarios, but never in a way that involves you fully surrendering your actual work product to it.”
Our reading: In a vacuum, LLMs are an interesting technology that can do some interesting stuff in the right scenarios, but never in a way that involves fully surrendering your actual work product to them.
No horizon
Qualifier: in a vacuum
Assessment: Can’t be checkedSaid
Causal claim
“The further you stretch yourself away from having complete clarity and industry over every element of the output’s purpose, the more likely you are to fall foul to a technology that is mathematically certain to make mistakes”
Our reading: The further you get from having complete clarity and industry over every element of an LLM output's purpose, the more likely you are to fall foul of a technology that is mathematically certain to make mistakes.
No horizon
Assessment: ContradictedSaid
Observation
“there has been little-to-no headway made in making a meaningful impact in other industries.”
Our reading: LLMs have made little to no headway in making a meaningful impact in industries other than coding.
No horizon Evidence as of
Limitation: The result depends on interpreting meaningful headway as substantial deployment-level gains.
Assessment: UnresolvedSaid
Forecast
“There is no AGI coming. There is no conscious computer.”
Our reading: AGI is not coming and there is no conscious computer.
No deadline stated Evidence as of
Limitation: The AGI Society assessment is a synthesis, not a consensus vote or a new comprehensive experiment.
Assessment: UnresolvedSaid
Causal claim
“The anthropomorphism and overpromising about these technologies has suffocated and obfuscated what they can actually do in pursuit of endless growth, and the only reason they can do anything is that OpenAI and Anthropic were allowed to annihilate hundreds of billions of dollars on training”
Our reading: Anthropomorphism and overpromising about LLMs have suffocated and obscured what they can actually do in pursuit of endless growth, and the only reason they can do anything is that OpenAI and Anthropic were allowed to spend hundreds of billions of dollars on training.
No horizon Evidence as of
Limitation: The financial component has an unspecified accounting basis.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“They do not need to be in anything near healthcare or finance or mental health or, really, people.”
Our reading: LLMs do not need to be in anything near healthcare, finance, mental health or, really, people.
No horizon
Assessment: UnresolvedSaid
Observation
“human beings can learn on the job at speed, and can self-correct in a way that LLMs are incapable of doing.”
Our reading: LLMs cannot learn on the job or self-correct the way humans can.
No horizon Evidence as of
Limitation: A correction benchmark is not a test of the full human-learning comparison.
Assessment: Can’t be checkedSaid
Forecast
“If you think any of this is worth hundreds of billions or trillions of dollars, you are either ignorant or corrupt. On top of how disgusting their outputs feel, the cost is going to take at least a decade to share, and begin the end of hypergrowth in the tech industry.”
Our reading: The cost of AI investment is going to take at least a decade to 'share' (his word, meaning unclear) and will begin the end of hypergrowth in the tech industry.
No horizon
Qualifier: 'share' is his word; its intended meaning is unclear
Assessment: UnresolvedSaid
Causal claim
“AI-authored code — especially vibe-coded programs — is inherently dangerous and disrespectful to the user, and I believe endless AI-generated code is behind the overall deterioration of software at large.”
Our reading: AI-authored code, especially vibe-coded programs, is inherently dangerous and disrespectful to the user, and Zitron believes endless AI-generated code is behind the overall deterioration of software at large.
No horizon Evidence as of
Limitation: Benchmark flaws do not establish industry-wide causation.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“AI research fucking stinks, as does AI writing.”
Our reading: AI research stinks, as does AI writing.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“I have never seen somebody like Nik who is also a huge, drooling fan of AI. In fact, the people most-excited about AI tend to, at best, create distinctly mediocre shit.”
Our reading: The people most excited about AI tend to create distinctly mediocre work at best.
No horizon
Qualifier: tend to
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Generative AI is far more useful as an idea than as a technology, and only ever has to show enough promise to back whatever vile agenda you’re pursuing.”
Our reading: Generative AI is far more useful as an idea than as a technology, and only ever has to show enough promise to back whatever vile agenda one is pursuing.
No horizon
Assessment: UnresolvedSaid
Observation
“Oh, and AI data centers are literally only good for AI, AI GPUs do not have other mass-market use cases. There is no post-Dot Com story.”
Our reading: AI data centers are literally only good for AI, AI GPUs have no other mass-market use cases, and there is no post-dot-com story.
No horizon Evidence as of
Limitation: Older technical announcements are context, not target-period outcome evidence.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“The fact any of you take “annualized run rate” seriously is an offense to good sense, and yes, that includes every reporter reporting it, even the ones I respect.”
Our reading: Annualized run rate figures for AI companies should not be taken seriously, including by reporters.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Look, I’m sorry, Anthropic is not worth $2 trillion, and whatever convinced you of that is a mixture of manufactured consent and mistaken trust of the powerful.”
Our reading: Anthropic is not worth $2 trillion, and whatever convinced people otherwise is a mixture of manufactured consent and mistaken trust of the powerful.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“It’s also ridiculous that anyone is talking about “recursive self-improvement.””
Our reading: It is ridiculous that anyone is talking about 'recursive self-improvement.'
No horizon
Assessment: ContradictedSaid
Observation
“And man, is it ridiculous that AI doomers warning about spooky superintelligences have somehow had such incredible prominence in the media without ever succeeding in stopping a single thing — or even substantiating their concerns.”
Our reading: AI doomers warning about superintelligence have not stopped anything or substantiated their concerns despite their media prominence.
No horizon Evidence as of
Limitation: The result depends on interpreting substantiation as empirical support rather than proof of catastrophic predictions.
Assessment: Can’t be checkedSaid
Causal claim
“The push into cybersecurity seems to have happened as a result of training models to excel at coding hitting the point of diminishing returns, at least from the perspective of impressing people enough to be excited about the company again.”
Our reading: AI labs' push into cybersecurity appears to result from coding training hitting diminishing returns in its ability to impress.
No horizon
Qualifier: from the perspective of impressing people
Assessment: Can’t be checkedSaid
Observation
“The fact that non-enterprise customers are still able to buy monthly subscriptions is proof that the AI labs know that regular people won’t pay the actual cost of AI.”
Our reading: The fact that non-enterprise customers can still buy monthly subscriptions proves that AI labs know regular people won't pay the actual cost of AI.
No horizon
Assessment: UnresolvedSaid
Observation
“For example, while GPT-5.6 Sol might be the “same price” as GPT 5.5 was, it burns more than twice the amount of tokens, meaning that the “cost of intelligence” might have gone down in the sense the model is better at benchmarks, but the “cost of actually doing shit” went up.”
Our reading: GPT-5.6 Sol is priced like GPT-5.5 but burns more than twice the tokens, so the real cost of doing work went up.
No horizon Evidence as of
Limitation: The supporting report is a secondary account of one workflow.
Assessment: UnresolvedSaid
Causal claim
“As a result, Anthropic and OpenAI are incentivized to have you burn more tokens and build inefficient models as a result.”
Our reading: Token-based pricing incentivizes Anthropic and OpenAI to make users burn more tokens and build inefficient models.
No horizon Evidence as of
Limitation: Provider efficiency disclosures are not independent verification.
Assessment: Can’t be checkedSaid
Observation
“The fact that OpenAI has now cut the costs of all three of its latest models less than two months after their release is a sign that it knows there’s a disconnect”
Our reading: OpenAI cutting prices on its three latest models within two months of release signals it knows there is a value disconnect.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“Guess what folks! Building the infrastructure for all these fucking LLMs just got more expensive, with NVIDIA raising its prices by 17% for systems due to be delivered next year — an important designation, because it’s very likely that much of the revenue for said systems gets booked in this year”
Our reading: After a 17% price increase for systems delivered next year, NVIDIA will very likely book much of the revenue for those systems in 2026.
No horizon
Assessment: Can’t be checkedSaid
Conditional forecast
“Anthropic plans to raise $100 billion at a $2 trillion valuation, and if it does so, it will successfully suck up the remaining liquidity in a market already dangerously close to losing its lunch.”
Our reading: If Anthropic raises $100 billion at a $2 trillion valuation, it will soak up the remaining liquidity in the market.
No horizon
Only if: if Anthropic raises $100 billion at a $2 trillion valuation
Assessment: Can’t be checkedSaid
Forecast
“And said market may not accept that valuation, or want that much stock.”
Our reading: The market may not accept Anthropic's $2 trillion valuation or want that much stock.
No horizon
Assessment: UnresolvedSaid
Forecast
“I must also be clear there’s no bailout coming.”
Our reading: There is no bailout coming for the AI industry; even a government funding injection for OpenAI and Anthropic would not solve the need for their spending to keep growing.
No deadline stated Evidence as of
Limitation: An isolated reading of the first sentence would produce a different test.
Assessment: Can’t be checkedSaid
Conditional forecast
“This is why I find it difficult to see a post-bubble future for LLMs.”
Our reading: Zitron finds it difficult to see a future for LLMs after the AI bubble.
No horizon
Only if: after the AI bubble bursts
Assessment: Can’t be checkedSaid
Observation
“We still lack any real killer app for LLMs.”
Our reading: LLMs still have no real killer app.
No horizon
Assessment: Can’t be checkedSaid
Causal claim
“These costs are only increasing, likely due to the diminishing returns of pre-training and the massive cost of buying training data for every imaginable new vertical.”
Our reading: OpenAI's training costs keep rising, likely because of diminishing returns from pre-training and the cost of training data.
No horizon
Assessment: Can’t be checkedSaid
Causal claim
“I also think that LLMs have created a near-permanent scar in the workforce, and traumatized more people than we’re aware of right now, both in those pressured about AI and those defending it.”
Our reading: Zitron thinks LLMs have created a near-permanent scar in the workforce and traumatized more people than we are aware of right now, both those pressured about AI and those defending it.
No horizon
Qualifier: I also think
Assessment: Can’t be checkedSaid
Forecast
“They’ll have to explain why they ignored a company that burned $5 billion in 2024, $20.9 billion in 2025, and will likely burn $30 billion or more in 2026”
Our reading: OpenAI, which burned $5 billion in 2024 and $20.9 billion in 2025, will likely burn $30 billion or more in 2026.
No horizon
Assessment: UnresolvedSaid
Forecast
“As I’ve said, NVIDIA’s price increase is going to increase the price of every single data center in construction by billions of dollars, and we’re already approaching the limits of how much money can be raised for them.”
Our reading: NVIDIA's price increase will raise the cost of every data center under construction by billions, and financing is nearing its limits.
No deadline stated Evidence as of
Limitation: Cost evidence is an estimate made before the post.
Assessment: Can’t be checkedSaid
Causal claim
“This is not a sustainable industry — either economically or emotionally — because it has a fundamentally dishonest relationship with its customers defined by the inconsistency of LLMs both in efficacy, stability (see: Anthropic’s downtime) and training”
Our reading: The AI industry is not sustainable economically or emotionally because it has a fundamentally dishonest relationship with its customers, defined by LLMs' inconsistency in efficacy, stability and training.
No horizon
Assessment: UnresolvedSaid
Observation
“It is an objective disaster that Microsoft spent more than $260 billion on capex to create a business with less than $11 billion in annual revenue outside of OpenAI”
Our reading: Microsoft spent over $260 billion on capex for an AI business with less than $11 billion in annual revenue outside OpenAI.
No horizon Evidence as of
Limitation: The claim combines a cumulative spending figure with an annual revenue figure.
Assessment: UnresolvedSaid
Observation
“These big numbers — like Anthropic having a $65 billion annualized run rate, an undefined term that obfuscates the fact that Anthropic has made $16.5 billion in the first half of 2026, losing billions of dollars in the process — are fundamentally meaningless, because they’re easily gamed at best, and inherently uncertain at worst.”
Our reading: Anthropic's $65 billion annualized run rate is meaningless; it made $16.5 billion in the first half of 2026 while losing billions.
No horizon Evidence as of
Limitation: The financial statements became public after the post but before the evaluation cutoff.
Assessment: UnresolvedSaid
Observation
“To be clear, they’re doing so because you can’t really justify anything about this industry based on what it does today. It costs too much, none of the businesses built on top of it are profitable”
Our reading: The AI industry cannot be justified by what it does today; it costs too much and none of the businesses built on it are profitable.
No horizon Evidence as of
Limitation: This hold does not endorse the claim that every qualifying business loses money.
Assessment: MixedSaid
Causal claim
“Its software lies because it doesn’t actually know anything and generates answers probabilistically”
Our reading: AI software lies because it doesn't actually know anything and generates answers probabilistically.
No horizon Evidence as of
Limitation: The result depends on reading knowledge as accessible factual information.
Assessment: Can’t be checkedSaid
Causal claim
“It deliberately subsidized its subscription products because it knew its customers wouldn’t pay the actual cost of AI”
Our reading: The AI industry deliberately subsidized subscriptions because it knew customers would not pay the true cost.
No horizon
Assessment: Can’t be checkedSaid
Observation
“It refuses to explain how it becomes profitable, because nobody knows how or has a tangible plan to do so.”
Our reading: Nobody in the AI industry knows how it becomes profitable or has a tangible plan to do so.
No horizon
Assessment: Can’t be checkedSaid
Causal claim
“And it’s losing because, at its core, AI was never built on very much.”
Our reading: The AI industry is losing because, at its core, AI was never built on very much.
No horizon
Assessment: Can’t be checkedSaid
Rhetoric
“It’s so much weirder, and in the end will be so much worse.”
Our reading: This time really is different: the AI bubble is so much weirder and, in the end, will be so much worse.
No horizon
Held from this post (25)
These records did not pass every check, so nothing about them is published. Held is not a finding.
- The two reviews read the claim differently
- The record does not meet the current rules. Our reading, not assessed: CoreWeave and Nebius will raise more billions of dollars in 2026.
- The two reviews read the claim differently
- The two reviews reached different assessments. Our reading, not assessed: Once investment stops flowing, LLM improvements will stop.
- The two reviews read the claim differently
- The two reviews read the claim differently
- The two reviews read the claim differently
- The two reviews read the claim differently
- The two reviews read the claim differently
- The record does not meet the current rules. Our reading, not assessed: Once the financial mania recedes, current LLM uses will not be enough to justify further investment, and a true LLM cannot be trained on the cheap.
- The criteria are ambiguous. Our reading, not assessed: Attempts by AI labs to sell premium, higher-priced models appear to be failing.
- The two reviews read the claim differently
- The two reviews reached different assessments. Our reading, not assessed: Once the AI bubble bursts, there will be no justification to train models at their current scale, and models to distill from won't exist if Anthropic and OpenAI don't train them.
- The two reviews read the claim differently
- The two reviews reached different assessments. Our reading, not assessed: Zitron estimates Anthropic and OpenAI will spend $200 billion in 2027, and lists their lacking the actual compute demand to justify that spending as one of several points at which the mismatch could happen.
- The two reviews reached different assessments. Our reading, not assessed: If OpenAI and Anthropic collapse, hyperscalers (except Oracle) would not go bankrupt but would stop growing and miss analyst expectations.
- The criteria are ambiguous. Our reading, not assessed: While 'doing AI' is still rewarded, enterprise AI spending will continue until everybody realizes they can likely cut a large amount of costs, first by moving to open-source models and then by not using them at all.
- The two reviews read the claim differently
- The two reviews read the claim differently
- The two reviews read the claim differently
- The two reviews read the claim differently
- The criteria are ambiguous. Our reading, not assessed: Because AI companies have not worked out their business models, current AI service pricing will change.
- The criteria are ambiguous. Our reading, not assessed: Paying users of LLM-powered services are guaranteed at some point to face service degradation where models misbehave, shifts in rate limits, or changes in product functionality driven by the services' shifting economics.
- The two reviews read the claim differently
- The two reviews read the claim differently