Fetching from the wire…
Public story · 2026-07-25 · high
He says it's nearly unpatchable and worked on Opus 5 at launch, but politics is why he won't publish it.
Why now: The claim landed July 24, dated to Opus 5's own launch day, when an unpatched universal jailbreak would matter most.
Pliny says he found a jailbreak that broke every flagship model he tested, including Opus 5 on its launch day, according to a July 24 post on X. He calls the technique nearly impossible to patch. If that holds, every lab shipping a flagship this year is carrying an exposure it can't fully close.
Nobody outside his own account has reproduced the exploit. Pliny has a documented history of posting working jailbreaks on launch day, but this time he says he isn't publishing. He's citing the political climate and a responsible-disclosure window as the reason.
The July 24 post has pulled 558,948 views and 9,326 likes. That's attention, not verification.
The capability claim is unverified and might stay that way outside his own telling. What's new here is behavioral, not technical: the account that built its name posting jailbreaks the same day it found them is now the one asking for patience.
Each link below shares sources, entities, or timing with this story.
Sony Music Publishing and Warner Chappell filed August 28 in the Northern District of California against Anthropic, CEO Dario Amodei and co-founder Benjamin Mann, over what they call a "brazen campaign of illegally torrenting, scraping and downloading copyrighted works on a ma...
Terminal-Bench 2.1 results (entries dated June 17) put Codex CLI on GPT-5.5 first at 83.4%, Claude Code on Fable 5 second at 83.1%, and Claude Code on Opus 4.8 at 78.9%. The asterisk matters more than the ranking: Fable 5 and Mythos 5 have been export-suspended since June 12,...
Simon Willison pulled the numbers out of an FT report sourced to "people with knowledge of the matter": Anthropic's annualized revenue reached $65bn in July, up from $47bn in May. Six thousand customers spend $100,000 or more a year. The company told investors it expects a pro...
Anthropic commissioned the independent evaluator to test 72 injection scenarios, held out from Anthropic, each run 10 times against Fable 5, Opus 5, and Sonnet 5 as of July 17. Clean sweep. TechCrunch has the details. A third-party held-out eval is a much stronger claim than i...
Jack Clark's Import AI 464 (around July 6) led with something I've been turning over all week. Claude Fable autonomously wrote what Clark calls "the first genuine (and fastest) megakernel" submitted to the KernelBench-Mega leaderboard. An 18.71x speedup in hand-written CUDA on...
Everyone kept score wrong. When OpenAI shipped GPT-5.6 (the Sol flagship plus Terra and Luna) to GA on July 9, then xAI put out Grok 4.5, Meta dropped Muse Spark 1.1, and Cognition shipped SWE-1.7, the reflex was to ask who won the benchmark. Wrong question. On the Artificial...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.