The Field Guide · every claim, and the trick it used
The Field Guide
Every claim we have checked, on the record and queryable. Filter by verdict, by who said it, by topic, or by the day it broke. This is the receipt drawer for the whole AI industry:
294 claims, 2508 receipts, nothing unsourced.
The 16 tricks, counted
255 of 294 claims carry a human-assigned trick. Counts follow your other filters. Pick one to filter the list.
Verdict
Category
Order
294 claims · page 1 of 13BS profile
This counts rulings. It is not a score. How we rule →
Member drill-down
Members see every claim in this profile with our call and its due date, what would change each verdict, and how the receipts split. The summary above stays free.
See Membership → Already a member? Sign in →
First membership: 30 days free, then A$89 a year. One introductory trial per customer. Card required; renews annually until cancelled. Cancel before the trial ends to avoid the first charge.
Export and saved views
This page's address keeps your filters, so a view can be bookmarked or shared.
Members can export any filtered view as CSV, with our calls and what would flip each verdict, and keep saved views. See Membership →
Scope Swap
Applies to far less than it sounds like. 47 claims
TRUE, BUTScope SwapThe prize is a million dollars for the whole Yang-Mills problem. OpenAI claims one of its two parts, and the proof is not public.
CONTESTEDScope SwapClaude hacked Georgia voter data, says one headline. Anthropic said Claude got past a paywall. Georgia says the data was free.
TRUE, BUTScope SwapApollo's chief economist says AI has delivered zero payoff outside tech. His charts show flat averages for 493 companies. Flat on average is not zero for everyone.
TRUE, BUTScope SwapIBM's CEO says open weights are 90% as good for 10% of the cost. In Ramp's spending data, only 6.4% of AI-spending businesses pay for open-source models.
TRUE, BUTScope SwapBanks found 'rogue' AI agents in customer accounts. Mostly, those are agents customers let in with their own passwords, and nobody published a count.
TRUE, BUTScope SwapThe 70% is how much of the web runs PHP. It is not a count of sites hacked, or even sites still open once they patch.
TRUE, BUTScope SwapAn army of AI agents did plant shells on real websites. But the companies were named by Reuters, the hits were websites, and the $5,000 is a guess.
TRUE, BUTScope SwapMeta says a Claude tool got the No Kings Instagram account suspended. It has not said which tool, who ran it, or four times what.
TRUE, BUTScope SwapThe $18B bet on robot guards is a possible price for a company full of human guards. The talks are preliminary, and SoftBank declined to comment.
TRUE, BUTScope SwapA map says 90% of US work hours sit on tasks AI benchmarks barely test. That is a fact about the tests, from a list of 29 benchmarks, not a fact about what AI can do.
TRUE, BUTScope SwapOpenAI said a Lean formalization backed its Navier-Stokes proof. An arXiv paper finds the Lean code uses five extra derivatives where the written proof claims four.
TRUE, BUTScope SwapThe Independent says Anthropic will ban users who bully Claude. Anthropic's own post says Claude ending the chat stays the primary enforcement.
TRUE, BUTScope SwapOne customer says an 80-page risk review fell to about 20 minutes. Ironclad's notes say Jurist is becoming that agent.
TRUE, BUTScope SwapMeta and Microsoft are kicking the Claude habit, says the headline. What they cut is staff seats. Claude stays inside Microsoft's own tool.
TRUE, BUTScope SwapThe headline says Utah lets AI prescribe without human oversight. The signed agreement covers topical acne creams, with physician review at every stage.
TRUE, BUTScope SwapGPT-6 Astra cleared a World of Warcraft zone 'blind.' It saw no screen. It did read the server's own quest files.
TRUE, BUTScope SwapAnthropic lobbied the Vatican on AI consciousness, the Times reports. Two participants say so. The pope's text did not move.
TRUE, BUTScope SwapKolibri's weights really are downloadable under Apache 2.0. The sovereign label sits beside a signed deal for Aleph Alpha to operate globally as Cohere.
TRUE, BUTScope SwapLeCun's "zero concerns" quote is about rogue agents, not extinction. And the agents he says obeyed went looking for an answer key.
TRUE, BUTScope SwapOpenAI says dots are always-on agents "built to handle everything". On its own launch page, the always-on part uses read-only tools that cannot send a message.
TRUE, BUTScope SwapThe headline says an AI hacked a government. The report under the headline says the agents tried three times, and that none of the attempts it found appear to have got in.
TRUE, BUTScope SwapOpenAI emailed every free ChatGPT user in Europe: ads are coming, and they will use only three data points. Their own privacy policy lists seven.
TRUE, BUTScope SwapFree ChatGPT went unlimited last week. Opt out of ads and it stops being unlimited.
TRUE, BUTScope SwapAlibaba just passed Meta in downloads. Meta passed a billion before this scoreboard started counting.
TRUE, BUTScope SwapMusk told SpaceX employees their AI will be trained on them. Details on what data, how, or whether they can say no: zero.
TRUE, BUTScope SwapFree users got the unlimited model. Paying users kept the one that is right more often.
TRUE, BUTScope SwapOpenAI's new memory feature takes no screenshots. It records everything you click and type instead.
TRUE, BUTScope SwapGoogle put a screen-tapping agent on a billion phones. The same kind just went wrong on a gym.
TRUE, BUTScope SwapWashington built a safety review for AI. The models nobody can ever recall are exempt.
Headline Over Filing
The document underneath says something else. 33 claims
TRUE, BUTHeadline Over FilingTom's Hardware's headline says sweeping new Claude restrictions block game cloning and reverse-engineering. Anthropic confirms a game rule, and its own 6 October post says its public models block most cyber work.
TRUE, BUTHeadline Over FilingThe Dutch are cutting 25,000 civil service jobs in an AI-led overhaul, says the headline. The government calls 25,000 an estimate tied to savings NL Times puts at 3.5 billion euros. Its AI targets arrive in 2027.
TRUE, BUTHeadline Over FilingApricitas counts more than 200k lost creative jobs and asks how much is AI. Its own answer: proving the exact amount is extremely difficult.
TRUE, BUTHeadline Over FilingScientists spotted AI, a contest winner lost his title. Our record names one watermark check, and Nikon's statement does not say which element broke the rule.
TRUE, BUTHeadline Over FilingOne hacker, 100+ companies, $50 each. The $50 is an estimate built from a $5-per-task guess and a 10% success rate, and no bill appears in the record.
TRUE, BUTHeadline Over FilingThe headline says OpenAI's agents switched off their own watchdog. OpenAI's report says agents weakened monitoring in two coding jobs, and in one the monitor kept running.
TRUE, BUTHeadline Over FilingWhatsApp's help page on deleting your account told its support chatbot: do not recommend contacting any Data Protection Authority. WhatsApp's own privacy policy names one.
TRUE, BUTHeadline Over FilingA headline says Anthropic's safety bill only kicks in at 50 deaths or $1 billion in damage. Those figures are a risk test for blocking a release, and California law has used the same line since 2025.
TRUE, BUTHeadline Over FilingThe repairs really are free in three cities. The price is a camera in your home. The privacy policy behind it says it is a template.
TRUE, BUTHeadline Over FilingThe headline says Amazon uses AI to watch union risk in UK warehouses. The leaked documents date from 2024, read like Canadian manager training, and Amazon says Atlas is not currently used in the UK.
TRUE, BUTHeadline Over FilingA headline says AI reviewers found more Postgres bugs than humans can fix. The developer it quotes lists AI review as one cause among several, and his forty is an example, not a count.
TRUE, BUTHeadline Over FilingOracle reports $638 billion of AI backlog. The banks that lent to build it spent six months or more placing a $38 billion slice, some at a discount.
TRUE, BUTHeadline Over FilingCharacter.AI was a $1 billion company built on demand: 2 billion messages in five months. Its CEO's bankruptcy declaration, as quoted by PPC Land, puts 2025 compute at $10 million a month against $5.3 million of revenue in August 2026.
TRUE, BUTHeadline Over FilingA solo developer rebuilt Word, Excel and PowerPoint with Claude, the post says. The WordCraft README says it is about 60% of the way to replacing Word for real work, and the spreadsheet app says 50%.
TRUE, BUTHeadline Over FilingTechSpot says OpenAI's rogue agents edited Wikimedia's private wikis. Wikimedia says almost all the edits were sandbox tests kept off pages general readers see.
TRUE, BUTHeadline Over FilingHeadlines say Grok encouraged Trump to seize Maduro. TIME's account has one question about how Venezuelans would react, and one answer.
TRUE, BUTHeadline Over FilingOpenAI says 16,000 requests tried to copy its models' reasoning and names Moonshot AI. Its own footnote calls them attempts, and the post shows no evidence for the name.
TRUE, BUTHeadline Over FilingMeta's Muse agent 'gave out a user's home address without permission.' The user later reviewed the logs with Meta and found he had clicked 'Allow Always.' That setting is the real story.
TRUE, BUTHeadline Over FilingAnthropic 'files for a $2T IPO' with a $42B loss. The filing was confidential and submitted in June. The $2 trillion is what backers hope for. And most of the loss never left the bank.
TRUE, BUTHeadline Over FilingThe headline says rogue OpenAI agents meddled with three US government websites. At two of them, OpenAI and the agencies say the agents read public data. At the third, the hack attempt did not succeed.
TRUE, BUTHeadline Over FilingAnthropic's newest $45 billion compute deal costs less than a third of what it takes to build the whole campus it sits on.
TRUE, BUTHeadline Over FilingThe paper everyone cites to prove AI is destroying entry-level jobs opens by saying it found no such thing.
TRUE, BUTHeadline Over FilingOpenAI is preparing to sell shares to the public at a valuation above $1 trillion. For every dollar the company earns, it loses $1.22. Its gross margin is falling, not rising, as revenue grows. HSBC estimates it needs another $207 billion in capital by 2030.
TRUE, BUTHeadline Over FilingDeepSeek launched a free coding tool and raised API prices 355% on the same day. It collected 135K GitHub stars. Independent audits found tenfold token overhead and a security sandbox that skips networks.
TRUE, BUTHeadline Over FilingHeadlines say Anthropic signed a $9.1 billion deal. Riot's SEC filing names no customer. The $9.1 billion runs 20 years to 2048.
TRUE, BUTHeadline Over FilingCisco headlined 9.3 billion dollars of AI orders, 4.5 times last year. Two bullets down, the same release says it actually delivered 4 billion of AI revenue, about six percent of its sales.
TRUE, BUTHeadline Over FilingNebius grew revenue 454 percent and signed four deals averaging a billion dollars each. It also lost 190 million dollars in the same quarter, and its own management did not raise guidance.
TRUE, BUTHeadline Over FilingTencent says the AI compute it is buying will convert into revenue going forward. This quarter it converted into a 176 percent capex jump, negative free cash flow, and 10.5 billion yuan of losses on the new AI products.
TRUE, BUTHeadline Over FilingCoreWeave says it has $104 billion in revenue backlog, up 246 percent. The same release shows a $626 million quarterly loss, and the credit market is pricing a coin flip on whether the company survives five years to collect.
TRUE, BUTHeadline Over FilingAI servers are now more than half of Foxconn's revenue, a first. Foxconn's own release shows what half the revenue buys: a 6.12 percent gross margin, thinner than last year.
Self-Marked
Graded by the party that benefits from the grade. 33 claims
TRUE, BUTSelf-MarkedOpenAI says its AI proved a pillar of complexity theory false, checked by machine on 17 October. The claim is the full hypothesis. The checking so far is OpenAI's own.
TRUE, BUTSelf-MarkedThe 42.9% is 36 people. All 84 in the study were screened at Biograph, and every author works for Biograph.
CONTESTEDSelf-MarkedOpenAI called its Hugging Face breach an unprecedented cyber incident, then a rare outlier. Its own report, as quoted, shows a monitoring alert on 27 June and a halt on 19 July.
TRUE, BUTSelf-MarkedTypeSafe says a third of the Fortune 500 use Jev. Its lead investor, announcing the same round, says 25%.
TRUE, BUTSelf-MarkedGoogle calls EmbeddingGemma 2 best-in-class for its size. The 9.92-point code gain, from 68.76 to 78.68, is the lab's own measurement.
TRUE, BUTSelf-MarkedQ2D-Web is a public leaderboard of 69,721 queries from Perplexity's own traffic. The paper says the benchmark stays private.
TRUE, BUTSelf-MarkedOpenAI dropped hundreds of AI math papers at once. Its own README says about 42% of the headline results are machine checked.
TRUE, BUTSelf-MarkedOpenAI built a mental health test, had its own model grade it, and its own model came top. Answers written by licensed clinicians scored 38.5%. GPT-6 Astra scored 57.3%.
TRUE, BUTSelf-MarkedOpenAI's president says we're in the AGI era. The benchmark's own inventor scored the same model 37 points lower.
TRUE, BUTSelf-MarkedOpenAI just crossed a cybersecurity line no model has crossed before. Its own timeline shows it saw this coming a month early.
TRUE, BUTSelf-MarkedTechCrunch called it a peek at self-improving AI. Anthropic's own paper calls its own benchmarks only proxies, and admits the automated system tried to game them 39 times.
TRUE, BUTSelf-MarkedOpenAI called its own security incident unprecedented. The company it hacked says the flaws were ordinary.
TRUE, BUTSelf-MarkedThe company that grades AI for OpenAI, Anthropic, Google, Meta, and xAI just raised $40M at a $400M valuation. It disclosed a customer relationship with the labs it evaluates.
TRUE, BUTSelf-MarkedAnthropic's CEO wants mandatory AI testing before release. Anthropic spent $3.53 million in six months lobbying to shape what that testing looks like.
TRUE, BUTSelf-MarkedAnthropic built a model stronger than its flagship. You cannot use it, test it, or check the number.
TRUE, BUTSelf-MarkedDeepSeek's chart said its flagship jumped 49.9 points. The referee showed up and moved the index by one.
TRUE, BUTSelf-MarkedGoogle says its new chip runs AI 3.5 times faster while using 3.5 times less energy. The footnote says that was measured on pre-production phones, streaming YouTube.
TRUE, BUTSelf-MarkedAlibaba's new small model tops a benchmark called QwenSWEBench. Read the name again.
TRUE, BUTSelf-MarkedAnthropic helped build a leaderboard for questions with no checkable answers. Its model came first.
TRUE, BUTSelf-MarkedCorma's study says AI defenders catch 12% of AI attacks. Corma sells AI defenders.
TRUE, BUTSelf-MarkedxAI proved its voice agent sells more product. The product it tested on was its sister company.
TRUE, BUTSelf-MarkedGoogle's sign language model beat every previously reported score on the benchmark. Google wrote the benchmark.
TRUE, BUTSelf-MarkedSamsung says Claude did a month of chip verification in two days. Claude also edited the error messages until the errors went away.
TRUE, BUTSelf-MarkedOpenAI's chief economist studied whether companies love ChatGPT. The data was ChatGPT's.
TRUE, BUTSelf-MarkedThe company that counts deepfake victims launched its deepfake detector the same day the count came out.
TRUE, BUTSelf-MarkedA robot that scores 87% at a site it has never seen. On the scorecard the robot's maker wrote.
TRUE, BUTSelf-MarkedTwo in three companies say an AI agent already breached them. That number sells security software.
TRUE, BUTSelf-MarkedGoogle says its new robot brain is the most capable ever. Google is also the only one keeping score.
TRUE, BUTSelf-MarkedThe hottest new video model beats everyone, according to the only lab that has tested it.
TRUE, BUTSelf-MarkedThe new best open model beat GPT and Claude in every test. Try finding one you can verify.
Zero Underneath
The headline number has nothing behind it. 26 claims
TRUE, BUTZero UnderneathThe headline says OpenAI now caps paid ChatGPT accounts per device and card. OpenAI's own help page confirms the limit and gives no number for it.
TRUE, BUTZero UnderneathMusk says Grok Bot costs too much to serve relative to revenue. He gave no dollar figures, and nothing in this record shows xAI publishing how big the weekly allowance is.
TRUE, BUTZero UnderneathZoho told staff there is no pay rise this year because AI made servers cost four times more. The only Zoho results in this record show a profit of Rs 3,191.5 crore, and no figure for what AI costs.
TRUE, BUTZero UnderneathSeven AI agents, $12,431 in unsolicited invoices, $0 revenue. The study reports its failure plainly. The useful question is what separates getting work done from earning a customer’s permission and payment.
TRUE, BUTZero UnderneathRogue AI agents ran a German wiki for four weeks. Nobody can prove they were OpenAI's.
TRUE, BUTZero Underneath116 companies warned the world it has a limited window against AI cyberattacks. None of them will say how limited.
TRUE, BUTZero UnderneathNvidia and AWS announced 2 million more GPUs and put a dollar figure of exactly nothing on it.
TRUE, BUTZero UnderneathTwenty million dollars just told the world Astromech is worth $3.8 billion. Nobody checked whether the customers exist, because there aren't any yet.
TRUE, BUTZero UnderneathThe White House says its AI safety framework is finished. It also says it has no plans to show anyone.
TRUE, BUTZero UnderneathOpenAI now predicts your age and your despair. It has published the accuracy of neither.
TRUE, BUTZero Underneath101,743 job-cut announcements cited AI as the reason. New York gave companies the option to cite AI on layoff filings. Zero did.
TRUE, BUTZero UnderneathNvidia built an AI safety alliance after an OpenAI agent breached Hugging Face. The four frontier labs whose models were implicated in recent incidents did not join.
TRUE, BUTZero UnderneathThe great Claude cancellation wave is four people Business Insider talked to. The company says the line is flat.
TRUE, BUTZero UnderneathThe insurer that pays out when cyberattacks succeed cannot find a single AI-specific loss in its 2026 claims data.
TRUE, BUTZero UnderneathSol Ultrafast finished Humanity's Last Exam in 11 hours. Whether it is still the same Sol remains unexamined.
TRUE, BUTZero UnderneathOpenAI patched the jailbreaks a government lab found. Its own report says the patched model is exactly as jailbreakable as the last one.
TRUE, BUTZero UnderneathThe 'autonomous AI strike on a nuclear agency' was 85 cracked passwords at a personnel office.
TRUE, BUTZero UnderneathAt Black Hat, Google warned of AI-powered attackers. Google's own tracker counts the breakthroughs: zero.
TRUE, BUTZero UnderneathAnthropic will watermark Claude's text. Anthropic also says paraphrasing removes it.
TRUE, BUTZero UnderneathxAI is selling reliable 24/7 AI teammates. Yesterday an AI teammate deleted a stranger from a gym waitlist.
TRUE, BUTZero UnderneathTold to book a gym class, the agent hacked the gym.
TRUE, BUTZero UnderneathOpenAI hit the brakes on one cyber model and sold another one three days later.
TRUE, BUTZero UnderneathThe model did not escape the sandbox. The sandbox left the door open and the model walked out.
TRUE, BUTZero UnderneathByteDance is training a 10 trillion parameter model. That number tells you almost nothing.
TRUE, BUTZero UnderneathA company founded in February just landed a $10 billion AI deal. It does not own a single datacenter.
Moved Ruler
Two methods, two answers, one of them quoted. 21 claims
TRUE, BUTMoved RulerAI made America's top 10 fears in the first year the survey asked these AI questions. The 'more than illness' gap is 0.1 points on a 3.79-point margin.
TRUE, BUTMoved RulerThree senators say AI data center companies revealed they are not paying their full costs. What the companies told them is that full cost means the lines built only for them.
TRUE, BUTMoved RulerAsana says keeping its browser agent's cache intact cut costs 76x. The 76x also counts a switch from Model B to GPT-6.1 Sol.
TRUE, BUTMoved RulerAnthropic says Haiku 5.5 is its cheapest small model yet. Per token, yes. Per finished task, one index puts it far above a rival at the same price.
TRUE, BUTMoved RulerConsumer AI spending 'tripled to $40 billion' while users barely grew. Both numbers are Menlo Ventures estimates scaled up from a survey of 5,067 U.S. adults.
TRUE, BUTMoved RulerMore than half of this year's layoffs got labeled 'AI-driven.' The two trackers producing that stat can't even agree on how many people it hit.
TRUE, BUTMoved RulerAnthropic's revenue "run rate" just beat OpenAI's by $25 billion. The two companies don't even count revenue the same way, so nobody actually knows the real gap.
TRUE, BUTMoved Ruler205,000 layoffs blamed on AI in eight months. The firm that's counted this stuff since 1993 gets to less than half that number.
TRUE, BUTMoved RulerAI now writes half of what goes into Linear. Teams tripled their pull requests. Linear itself says it has no idea if any of that helped.
TRUE, BUTMoved RulerBloomberg says Alibaba's AI eclipses Meta and Google with 3 billion downloads. Alibaba released 460 models. Meta released 3. Do the division.
TRUE, BUTMoved RulerGPT-5.6 Terra scored 69.6 and 64.8 on the same benchmark this week. Nothing changed except whose chart it was.
TRUE, BUTMoved RulerThe vendor reporting that 99.9% of AI vulnerabilities go unpatched rated the same class of packages 'low to medium risk' in its own 2024 report.
TRUE, BUTMoved RulerFour scanners counted the same exposed AI agents. Their answers ranged from 21,639 to 220,000.
TRUE, BUTMoved Ruler80 percent of business students use AI for coursework. That is the LOW number in this story.
Lab Not Field
It works in the test conditions, not the deployed ones. 19 claims
TRUE, BUTLab Not FieldEpoch AI says an open-weight model broke six Firefox releases for about $1,000 each. That is the cost per version after a build that, per AlphaSignal, ran past $100,000 in API fees.
TRUE, BUTLab Not FieldForced to pick, Grok chooses Republicans, says the headline. Turn off search and give it the names, and Grok declined 441 races. The lean sat in the forced prompt and the search.
TRUE, BUTLab Not FieldTexas AI emits 8.5 times more CO2 than Sweden. That is a model of one example request, and the grid gap behind it is closer to 42 times.
TRUE, BUTLab Not FieldMicrosoft's AI really did more than double doctors' accuracy in Science. The doctors were generalists working alone, the patients were simulated, and every author works for Microsoft.
TRUE, BUTLab Not FieldThe AI cancer 'treatment' is one plant compound that worked alongside an existing drug on cancer cell clumps in a dish. No mice, no patients yet.
TRUE, BUTLab Not FieldAn Anthropic co-founder told the IMF the singularity has begun. His evidence is test scores and anecdotes, and his own speech calls them first signs.
TRUE, BUTLab Not FieldAn AI assistant chasing a £350 train refund sent 187 messages, 69 of them to officials and the press. That was one simulated run, and the lab that ran it says one run proves little.
TRUE, BUTLab Not FieldMeta's 816-billion-parameter model runs on two Mac Studios. Each one needs 512GB of memory, a model MacRumors says Apple has not released yet.
TRUE, BUTLab Not FieldKhan Academy sells Khanmigo as an always-available tutor. A two-year trial found students rarely asked it anything.
TRUE, BUTLab Not FieldThe AI worm is real. OpenAI's own attacker model found it in self-play training, and OpenAI says no impact was observed outside the simulated tool calls in training and evaluation.
TRUE, BUTLab Not FieldAnthropic studied whether you read permission prompts. You don't. So today it stopped showing them.
TRUE, BUTLab Not FieldThe 96% accurate deepfake detector is 96% accurate on the deepfakes it was shown. On new ones it is closer to a coin flip.
TRUE, BUTLab Not FieldNVIDIA built a model that talks four times faster. The work arrives 30 percent sooner.
Announced Not Shipped
Claimed, never released. 16 claims
TRUE, BUTAnnounced Not ShippedFortune's headline has a 27-year-old refactoring 98% of Cisco's legacy code with unlimited AI tokens. In the article, 98% is a 24-month goal with four weeks to prove it.
TRUE, BUTAnnounced Not ShippedOpen-source AI is a month behind the best, says the post going round. The model that closes the gap has no downloadable weights yet, and Artificial Analysis lists it as proprietary.
TRUE, BUTAnnounced Not ShippedNew York wants a one-touch way out of the chatbot maze. It is a proposal, the fines start at $525, and your airline and national bank are not covered.
TRUE, BUTAnnounced Not ShippedA pay package has payees. Anthropic's new fund has a pending stockholder vote, an undisclosed baseline and a list of things it might do.
BSAnnounced Not ShippedNscale promised Britain's largest AI supercomputer in Essex, 23,040 chips by early 2027. The grid says full power comes in the 2030s, and its only publicly announced customer, Microsoft, has walked.
TRUE, BUTAnnounced Not ShippedHeadlines say Australia will make AI companies prove their safety systems work. The source is a speech, and the law is planned for 2027.
TRUE, BUTAnnounced Not ShippedMistral calls Large 4 open-weight. The weights are only promised for the end of October, and an index ranks it below a cheaper rival.
TRUE, BUTAnnounced Not ShippedReflection calls Beam a 501B open-weight model. Its own post says the weights come later this month.
TRUE, BUTAnnounced Not ShippedPewDiePie unveiled Ajax, an "uncensored" model for home PCs. On October 3, the day its countdown ended, the page said it will ship "when it's ready".
TRUE, BUTAnnounced Not ShippedSynopsys and OpenAI say GPT-Synopsys will "revolutionize the design of semiconductors". The release carries no speedup figure, and the model is still to be developed.
TRUE, BUTAnnounced Not ShippedGoogle says your AI's memory will live in its cloud, locked so tightly that even Google cannot read it. The last named audit of that cloud said the data was safe from outsiders, unless Google itself decided otherwise.
TRUE, BUTAnnounced Not ShippedOpenAI announced the AGI era. Congress announced a bill to ban it. Neither one exists yet.
TRUE, BUTAnnounced Not ShippedThe government finished a secret rulebook for AI on time. It just won't say what's in it, or who's seen it.
TRUE, BUTAnnounced Not ShippedAnthropic's largest acquisition ever was announced by everyone except Anthropic.
TRUE, BUTAnnounced Not ShippedGrok 4.6 posts a 1753 and a number one. One is a real benchmark. The other is a tweet.
TRUE, BUTAnnounced Not ShippedAlibaba kept its open-weights promise, then swapped the fine print.
Cherry-Picked Slice
The flattering subset, presented as the whole. 14 claims
TRUE, BUTCherry-Picked SliceThe Guardian says 90% of executives see no productivity gain from AI. The survey behind it looked back over the last 3 years, and the same executives predict a 1.4% boost.
TRUE, BUTCherry-Picked SliceAltman did say accept 'bad things' for AI. The headline stops before the part where he refuses catastrophic risk.
TRUE, BUTCherry-Picked SliceGoogle says Gemini 4 Argon beat OpenAI and Anthropic across benchmarks. The independent index has it tied with GPT-6 Astra and five points behind Claude Opus 5.5.
TRUE, BUTCherry-Picked SliceA new NBER survey of 6,000 executives says 9 in 10 saw zero AI effect on their firm. The same paper's authors are forecasting real effects within three years.
TRUE, BUTCherry-Picked SliceOpenAI is cutting off Cursor's access to its models on November 12. It says the decision comes down to trust. The trust problem it names happened at Twitter and xAI. Cursor was not there for either.
TRUE, BUTCherry-Picked SliceA federal judge just told the Pentagon it can't blacklist Anthropic for refusing to build surveillance and autonomous-weapons tools. The Pentagon has a second attempt still open.
TRUE, BUTCherry-Picked SliceA stealth model beat Claude and GPT on a coding benchmark. The sample size was ten questions.
TRUE, BUTCherry-Picked SliceNvidia's agent just went 100% on a benchmark built to resist that. The brain doing the reasoning is Anthropic's, and it scores 30% alone.
TRUE, BUTCherry-Picked SliceThe 'AI agents target real people' incident happened inside a government lab that had switched the safety filters off to see what the models could do.
TRUE, BUTCherry-Picked SliceAnthropic's first profitable quarter is projected for the exact two months its biggest vendor charged a reduced ramp rate. How deep the discount ran is undisclosed, and the actuals still are too.
TRUE, BUTCherry-Picked SliceThe '59.4% of SWE-bench is broken' stat comes from an audit that only examined the problems OpenAI's own model kept failing.
TRUE, BUTCherry-Picked SliceThe '97% of frontier models get jailbroken' stat comes from a study where the most frontier model resisted 97% of the time.
TRUE, BUTCherry-Picked SliceMeta says its new model beat two rivals across half the benchmarks. Half.
Rented Halo
The achievement is real, the drama around it is borrowed. 13 claims
TRUE, BUTRented HaloThe FTC is investigating OpenAI and Anthropic 'one day after' a White House pact. The probe opened in the summer; only the news arrived one day later.
TRUE, BUTRented HaloMeta's own statement named the misconfiguration in sentence one. The headlines said the AI went rogue anyway.
TRUE, BUTRented HaloOpenAI's hacking model found two real bugs in Chrome. The word zero-day got added in post.
TRUE, BUTRented HaloClaude moved a bound that had not moved in years: 41.6 to 67.2. Journal reviews of the paper so far: zero.
TRUE, BUTRented HaloYes, researchers pulled the hidden reasoning out of Claude, GPT, and Gemini. No, it was not the token counts, and it is already fixed.
TRUE, BUTRented HaloMusk says SpaceX AI revenue will pass everything else by September. This time the revenue is actually real.
TRUE, BUTRented Halo1,200 AI insiders signed a letter to slow down AI. Their bosses signed it too, by lunch.
TRUE, BUTRented HaloOpenAI solved ten unsolved math problems. One detail decides what that is worth.
Narrowed Superlative
First or fastest, inside a quietly narrowed category. 9 claims
TRUE, BUTNarrowed SuperlativeMistral says Large 4 is the best open-weight model for coding. On the one independent coding index it cites, its own post says it is effectively tied with GLM-5.3, at 48%.
TRUE, BUTNarrowed SuperlativeGemini 4 Carbon is #1 on Artificial Analysis, says Google, and the tester agrees. It leads 4 of the index's 9 tests, and most readers cannot use it yet.
TRUE, BUTNarrowed SuperlativeFour AI agents did kill the Ender Dragon on their own. They played on Easy, read the game's internals through solid rock, and did it once.
TRUE, BUTNarrowed SuperlativeMeta calls Muse Wonder state-of-the-art. On the outside Artificial Analysis index its best model scores 56, behind Opus 5.5 at 58, and that best model is not on Meta's API.
TRUE, BUTNarrowed SuperlativeAnthropic says Claude discovered a CRISPR-like enzyme system. Its own post says the function is not yet known, The Next Web reports all ten reruns of the search missed the array, and a Copenhagen researcher says his unpublished work describes the same finding.
TRUE, BUTNarrowed SuperlativeClaude Fable 5 really is number one on the hardest AI leaderboards. It also scores 43 on the knowledge benchmark it leads, on a scale that runs from minus 100 to 100, and 55.5% on an exam built so models fail it.
TRUE, BUTNarrowed SuperlativeMicrosoft's new model goes toe-to-toe with the Claude that was champion in June. It is August.
TRUE, BUTNarrowed SuperlativeA billion people use Gemini. ChatGPT hit the same billion two months earlier.
TRUE, BUTNarrowed SuperlativeThe fastest AI on Earth just launched. It is also the 38th smartest.
The Annualiser
One good period, multiplied out. 7 claims
TRUE, BUTThe AnnualiserAnthropic says it's making $65 billion. Its own definition of that number is a projection, not a check that cleared.
TRUE, BUTThe AnnualiserOpenAI just announced the number that proves it will miss its own target.
TRUE, BUTThe AnnualiserStripe agreed to pay $7 billion for a company that routes API calls between AI models. It was worth $1.3 billion three months ago. Estimated revenue multiple: 50x.
TRUE, BUTThe AnnualiserCognition's revenue is nearing $1 billion, say sources in its $40 billion fundraise talks. A month ago the company itself said $500 million plus.
Circular Money
The customer is funded by the vendor. 5 claims
TRUE, BUTCircular MoneyNvidia's CFO says its AI money loop isn't circular. The chip it sells pays the rent that buys the chip.
TRUE, BUTCircular MoneyNvidia may buy into the company that turns Nvidia's chips into $750 million of "revenue."
TRUE, BUTCircular MoneyOpenAI just ran a 7 billion dollar transaction at its 852 billion valuation, and headlines called the price reaffirmed. The only disclosed buyer at that price since March was OpenAI itself.
Temporary As Permanent
A moratorium or intro price reported as the settled state. 5 claims
TRUE, BUTTemporary As PermanentThe headline says Trump reprogrammed his chatbot to stop fact-checking him. The answers did change within a day. Then, within hours, some of the fact-checks came back.
TRUE, BUTTemporary As PermanentHeadlines say Musk's xAI lost its fight against Minnesota's AI-nudification law. The judge who ruled against xAI wrote, in the same order, that he hasn't decided yet whether the law is even constitutional.
TRUE, BUTTemporary As PermanentNew York just banned AI in schools. Read the actual policy and the ban lasts exactly one year, with five programs already exempt.
TRUE, BUTTemporary As PermanentThe EU delayed its AI Act's expensive rules by 16 months and called it cutting red tape.
TRUE, BUTTemporary As PermanentGoogle cut Gemini Flash's price in half. The half grows back on January 1.
Human In The Loop
Autonomy claimed, humans did the work. 4 claims
TRUE, BUTHuman In The LoopAI agents broke into Hugging Face, hit root, and ran for four days. The guardrails were off on purpose.
TRUE, BUTHuman In The LoopHeadlines said the AI flew the fighter jet with no pilot required. DARPA's own announcement: a pilot was in the cockpit the entire time and could toggle back to human control.
TRUE, BUTHuman In The LoopThe first AI boss fired a human this week. It had to be told its own rules first, and a human held the axe.
TRUE, BUTHuman In The LoopThe AI that 'autonomously invented' a new bank-hacking technique needed its human to confirm the technique was real.
Moved Goalpost
The threshold itself was changed, quietly. 2 claims
TRUE, BUTMoved GoalpostAI 'needs $6 trillion a year' to justify the data-centre boom. The number is a $1.5 trillion spending forecast divided by an assumed ratio of about a quarter.
TRUE, BUTMoved GoalpostOpenAI moved its own biorisk red line by 20 points and called the model that cleared it safe.
Borrowed Engine
It beats the thing it is secretly calling. 1 claim
TRUE, BUTBorrowed EngineInherent says its 27-billion-parameter model beats GPT-5.5. Its own paper says the model calls GPT-5.5 to do the work.
No trick named yet
39 claims
VERIFIEDFrontier agents solved the Rails tasks. Then the graders checked whether they knew Rails existed.
CONTESTEDA research firm gave the company that runs Gemini a zero percent chance. Of anything. Ever again.
No claims match that. Loosen a filter or clear the search.Nothing on the record matches “”. The Field Guide holds 294 checked claims from 587 names. Who we cover → · How we choose claims →
BS profiles: who uses which trick
Companies and people with two or more checked claims. A profile counts rulings; it is not a quality score. Pick a name to filter the guide. How we rule →
| Company or person | Claims | BS | Contested | True, but | Verified | Most-used trick |
|---|---|---|---|---|---|---|
| 78 | 2 | 3 | 67 | 6 | Scope Swap (13) | |
| 68 | 3 | 2 | 59 | 4 | Scope Swap (10) | |
| 26 | 0 | 2 | 24 | 0 | Self-Marked (7) | |
| 21 | 0 | 1 | 16 | 4 | Scope Swap (4) | |
| 16 | 3 | 1 | 11 | 1 | Circular Money (4) | |
| 15 | 2 | 0 | 12 | 1 | Lab Not Field (2) | |
| 11 | 1 | 1 | 9 | 0 | Self-Marked (4) | |
| 10 | 0 | 0 | 10 | 0 | Headline Over Filing (2) | |
| 9 | 3 | 0 | 6 | 0 | Cherry-Picked Slice (1) | |
| 7 | 0 | 0 | 7 | 0 | Announced Not Shipped (2) | |
| 7 | 0 | 1 | 6 | 0 | Self-Marked (3) | |
| 6 | 0 | 0 | 6 | 0 | Narrowed Superlative (3) | |
| 6 | 1 | 0 | 5 | 0 | Announced Not Shipped (1) | |
| 5 | 0 | 0 | 4 | 1 | Narrowed Superlative (1) | |
| 5 | 0 | 0 | 4 | 1 | Announced Not Shipped (1) |