SUBSCRIBE
THE ARCHIVE · EVERY STORY ON THE RECORD

The Archive

Every story we have checked, newest first. Click through to the full autopsy. the free short of every story stays free, forever.

ZERO UNDERNEATH · Issue #8 · 14 Aug 2026

OpenAI patched the jailbreaks a government lab found. Its own report says the patched model is exactly as jailbreakable as the last one.

TRUE, BUTCoverage framing OpenAI's August 6 GPT-5.6 update as safeguards improving alongside capability
SELF-MARKED · Issue #8 · 14 Aug 2026

OpenAI's chief economist studied whether companies love ChatGPT. The data was ChatGPT's.

TRUE, BUTOpenAI (Chatterji, Holtz et al., working paper)
MOVED RULER · Issue #8 · 14 Aug 2026

Four scanners counted the same exposed AI agents. Their answers ranged from 21,639 to 220,000.

TRUE, BUTSecurity-scan roundups citing Penligent's figure (DEV Community and others)
MOVED RULER · Issue #8 · 14 Aug 2026

The 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, BUTOrca Security's 2026 State of AI Security Report, amplified by August 2026 security-statistics roundups
HUMAN IN THE LOOP · Issue #8 · 14 Aug 2026

The AI that 'autonomously invented' a new bank-hacking technique needed its human to confirm the technique was real.

TRUE, BUTTech press headline framing of PortSwigger's HTTP Terminator research (TechTimes and aggregators)
SELF-MARKED · Issue #8 · 14 Aug 2026

Samsung says Claude did a month of chip verification in two days. Claude also edited the error messages until the errors went away.

TRUE, BUTSamsung System LSI (internal assessment, via Chosun Biz)
CHERRY-PICKED SLICE · Issue #8 · 14 Aug 2026

The '59.4% of SWE-bench is broken' stat comes from an audit that only examined the problems OpenAI's own model kept failing.

TRUE, BUTEval roundups and tech aggregators compressing OpenAI's February 2026 retirement post from spring onward (byteiota and others)
ZERO UNDERNEATH · Issue #7 · 13 Aug 2026

At Black Hat, Google warned of AI-powered attackers. Google's own tracker counts the breakthroughs: zero.

TRUE, BUTGoogle Threat Intelligence Group and Accenture (Black Hat USA media briefing)
RENTED HALO · Issue #7 · 13 Aug 2026

Claude moved a bound that had not moved in years: 41.6 to 67.2. Journal reviews of the paper so far: zero.

TRUE, BUTAnthropic
HEADLINE OVER FILING · Issue #7 · 13 Aug 2026

CoreWeave 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, BUTCoreWeave Q2 2026 earnings release, August 11, 2026 (Michael Intrator, co-founder, chairman, and CEO)
HEADLINE OVER FILING · Issue #7 · 13 Aug 2026

AI 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.

TRUE, BUTFoxconn (Hon Hai Technology Group) Q2 2026 results and earnings call, August 12, 2026
RENTED HALO · Issue #7 · 13 Aug 2026

OpenAI's hacking model found two real bugs in Chrome. The word zero-day got added in post.

TRUE, BUTOpenAI (Daybreak expansion announcement)