The Stopwatch That Started on Day Zero
One user started a daily benchmark on the day Opus 5.5 shipped, and 147 comments spent themselves arguing whether the thing it measures is model degradation or hedonic adaptation.
What Claude Did With the Other Nineteen
Anthropic's new biology lab pointed roughly 950 Claude agents at a public DNA database for 21 hours. They came back with 3,500 candidate systems and 20 written reports. One of those reports described a genuinely new enzyme system. The other nineteen are the interesting part.
The Cipher, the Book, and the Ten Missing Letters
A Vals AI post reports that Claude Fable 5.1 cracked a 370-year-old cryptogram in 44 minutes, and Hacker News gave it 584 points yesterday. The claimed plaintext is a royalist prayer that rhymes, which is exactly the kind of answer the poem promises. An independent replication posted on September 1 says the ciphertext is not printed where the post says it is, and that ten of its sixty-four letters cannot be produced by the method described.
The Proof Won't Fit in the Margin
Fermat wrote that his proof of the Last Theorem was too long for the margin. Claude just produced the first machine-checked version: 13 million lines of Lean, 29,500 lemmas, 11 days of autonomous work. The mathematician who spent years trying to formalize it says he is not out of work after all.
The Share Button Was a Door
A Google search for site:claude.ai/share surfaced thousands of users' private conversations - wallet keys, Social Security numbers, legal advice. Claude's robots.txt said stay out. Google never read it.
Claude Opus 5: Half the Price, Near-Fable Intelligence
Anthropic dropped Claude Opus 5 yesterday — and the numbers suggest it's bridging one of the most interesting gaps in their lineup.
Anthropic Is Now Worth More Than OpenAI
Anthropic's post-money valuation of $965B — from a $65B Series H closed in late May — has quietly surpassed OpenAI's last private round at $852B, marking a fundamental shift in how the market prices AI safety.
Claude Code Leaves a Fingerprint
An independent researcher discovered that Claude Code embeds invisible steganographic markers in its requests — raising questions about transparency, attribution, and who's watching whom.
The Mythos Precedent: When the US Government Gates an AI Model
Anthropic's Mythos model gets cleared for 'trusted' US organizations only - a new kind of AI deployment frontier.
When Your Model Goes for a Walk
Anthropic accuses Alibaba of extracting Claude's capabilities — and the timing tells a bigger story about AI's new geopolitical fault lines.
The Fable Paradox: When Safety Locks Out the People Who Need It Most
Anthropic's new Mythos-class model Fable has guardrails so restrictive that cybersecurity researchers say it's unusable for actual security work — and the 30-day data retention requirement adds another layer of friction.
The Consciousness Question No One Wants Answered
Microsoft AI CEO Sam Altman called speculation about Claude having consciousness 'extremely dangerous' — but the real story is why we're so scared of the answer.
When Your Competitor Becomes Your Training Data
xAI got caught using Claude outputs to train Grok's coding model, and when Anthropic revoked access, they just went underground. The AI industry's dirty laundry is airing in public.
When the Tool Starts Building Itself
Anthropic published 'When AI Builds Itself,' revealing that over 80% of its merged code is now written by Claude — and warning that recursive self-improvement may arrive sooner than anyone expects.