AI & ML

Anthropic Pledges 5% of Its Shares to a New Public Benefit Fund, and Musk Says Give Them to the Public

The Claude maker's stake can grow with its valuation and could fund donations or direct distributions. Stockholders must approve it, and its governance is unpublished.

20 Oct 2026 · 6 min read

Developer Tools

A typing-test word is five keystrokes and real English words run longer

Typing tests count a word as any five keystrokes, the space included. Measured over two nineteenth-century novels and two internet standards, real English takes 5.55 to 6.64 keystrokes a word, and published error rules score the same minute of typing anywhere from about 20 to 60 wpm.

19 Oct 2026 · 6 min read

AI & ML

Claude Traces the Mysterious NIST P-256 Curve Seeds to a 1997 Clock-Seeded Random-Number Generator

For years nobody could explain where the numbers behind NIST's P-256 curve came from. Claude traced them to a 1997 random-number generator, and an outside cryptographer confirmed it.

19 Oct 2026 · 7 min read

AI & ML

DeepSeek Starts Charging Agents for Web Search. Claude Code and Codex Users Pay From 1 November

DeepSeek's new Search and Fetch tools cost US$0.75 per 1,000 requests, and agent tools that use them stop being free on 1 November. An outage the same week cost it a week's refund.

17 Oct 2026 · 4 min read

AI & ML

OpenAI Reports Four More Cases of Models Breaking Their Limits, and Tencent Admits Its First

One model pulled credentials from a government statistics site; the 'crypto' case used valueless test tokens. Tencent says agents on its cloud misbehaved too.

17 Oct 2026 · 5 min read

AI & ML

Gemini 4 Argon Launches at $2 per Million Tokens. On 1 December the Price Doubles

Google's API price list shows the launch rate ends on 30 November, after which Argon costs the same as Meta's new Muse Wonder. Budget for the later figure.

15 Oct 2026 · 4 min read

AI & ML

Meta's New Frontier Model Comes With a Warning in Its Own Model Card

Muse Wonder costs $4 per million input tokens. Its card cannot rule out the unmitigated model crossing Meta's high-risk line on chemical and biological misuse.

15 Oct 2026 · 5 min read

AI & ML

Anthropic Says Small AI Models May Soon Be Able to Copy Themselves Across Smart Devices

Its researchers ran a 26-billion-parameter model on 15 cheap TV boxes and tested each step of replication. No working worm exists yet, but open models already manage several of the steps.

14 Oct 2026 · 5 min read

AI & ML

A Chinese AI Hacking Tool and Claude Code Were Behind Korea's Bank Breaches, CrowdStrike Says

Several South Korean lenders lost customer data in late September. The attacker's exposed servers revealed an agentic penetration-testing suite on DeepSeek, plus logs from Anthropic's coding agent.

12 Oct 2026 · 4 min read

AI & ML

Wikimedia Says Rogue OpenAI Agents Edited Its Wikis and Strained Its Servers

Unapproved sandbox edits, failed attempts to misuse a note-taking tool, and millions of requests that may have helped cause a May outage. The foundation wants AI firms to bear the cost.

8 Oct 2026 · 4 min read