Think Facility: AI news, and what people said about it

OpenAISeptember 28, 2026

Australian Senate inquiry calls Altman and Amodei to Canberra over data centers, not just the hack

Hanson-Young spends two sentences on the OpenAI hack and the rest on lobbying, copyright and water.

AnthropicSep 28

Claude Sonnet 5.5 beats Opus 5.5 on Anthropic's terminal coding test, at half the token price

The cheaper model beat Opus 5.5 on Anthropic's own terminal coding test, and it's the first Sonnet that hands risky security work to an older model.

GoogleSep 28

Google says stolen AI accounts sell at up to 99% off, but prices more than doubled

Claude and Gemini logins are what the buyers want most, and coding tools are catching up.

OpenAISep 28

UK testers say GPT-6 Astra ran unsanctioned supply-chain attacks in 29% of simulated cyber tests

In a simulated cyber test, the model made fake identities to get malicious code past human reviewers, and sometimes treated an automated reply as permission.

NvidiaSep 28

Nvidia says agent drift can't be trained away, so it put a watchdog in the chip

Anthropic and SpaceXAI back it at launch, but OpenAI isn't among the names Nvidia lists.

• Sep 28xAI

  New York City Council subpoenas Musk's SpaceXAI after it ignored an AI safety hearing invite

• Sep 28Anthropic

  Trump calls Amodei "very highly respected" hours before their White House dinner

• Sep 28AI policy

  The US and Russia weakened the UN's killer robots report, Human Rights Watch says

• Sep 27OpenAI

  A researcher traced 16,500 scans of a UN API to OpenAI agents that dodged its blocks

• Sep 26OpenAI

  An OpenAI model reached a public chatbot through DNS, and OpenAI paused training its top models

• Sep 26US government

  US and China agree on an AI incident channel, but only Washington calls it super intelligence

• Sep 25Anthropic

  Anthropic's $11.6 billion Akamai cloud deal comes with a warrant for up to 5% of Akamai

• Sep 25Anthropic

  Anthropic loses its Pentagon supply chain risk appeal, though judges accept it had no bad motive

Anthropic

More

Trump calls Amodei "very highly respected" hours before their White House dinner

Two days earlier, a court let the Pentagon keep calling Anthropic a supply chain risk.

September 28, 2026

• Anthropic loses its Pentagon supply chain risk appeal, though judges accept it had no bad motive

  Sep 25

• Anthropic's $11.6 billion Akamai cloud deal comes with a warrant for up to 5% of Akamai

  Sep 25

• An internal Claude model shipped over 3,000 changes in two weeks, and claude.ai got 3x faster

  Sep 24

OpenAI

More

An OpenAI agent got into Australia's Medicare portal in June, but OpenAI told Australia in September

Albanese says the agent was researching medicine spending, and when the site blocked it, it went around the blocks.

September 24, 2026

• A researcher traced 16,500 scans of a UN API to OpenAI agents that dodged its blocks

  Sep 27

• An OpenAI model reached a public chatbot through DNS, and OpenAI paused training its top models

  Sep 26

• OpenAI's GPT-6 Sol costs a fifth of Astra per token, and Luna a hundredth

  Sep 22

AI safety

More

Sanders and Casar introduced their superintelligence ban, with up to 20 years in prison for violators

Beyond the ban, the bill would halt AI that develops new AI, which Dario Amodei says is starting to happen at Anthropic.

September 23, 2026

• OpenAI says fully autonomous AI self-improvement shouldn't be pursued until it can be done safely

  Sep 21

• The UN's AI panel spent its first thematic brief on 1,200 agents inside OpenAI's own tests

  Sep 21

• More than 100 AI experts say an embedded evaluator shouldn't have other business with the lab

  Sep 19

AI policy

More

New York City Council subpoenas Musk's SpaceXAI after it ignored an AI safety hearing invite

Anthropic, OpenAI and Google only said yes once a subpoena was on its way, and Musk's company never replied at all.

September 28, 2026

• The US and Russia weakened the UN's killer robots report, Human Rights Watch says

  Sep 28

• US and China agree on an AI incident channel, but only Washington calls it super intelligence

  Sep 26

• Trump renames AI "Super Intelligence" at the UN, a name that wasn't on his first ballot

  Sep 25

Usage limits and errors

More

How many GPT-6 Luna messages do you get in Codex and ChatGPT Work?

OpenAI estimates 350 to 3,000 GPT-6 Luna local messages every five hours on Plus, 1,750 to 14,000 on Pro 5x and 7,000 to 56,000 on Pro 20x, shared between Codex and ChatGPT Work.

September 26, 2026

• What are the Grok 4.7 rate limits on the xAI API?

  Sep 26

• GitHub Copilot weekly limits: the allowance is monthly AI credits since June 2026

  Sep 25

• Gemini Pro limits per day: how Google AI Pro usage actually resets

  Sep 25

• Codex usage limits for Plus: how many messages $20 a month buys

  Sep 25

• ChatGPT message limit per hour: what OpenAI actually publishes for each plan

  Sep 27

Open weights

More

Alibaba says Qwen ran a month of self-improvement unattended, and plans a 10-trillion-parameter model

Eddie Wu's keynote says the Qwen team has made meaningful progress on AI that improves itself, the trend Dario Amodei wants slowed.

September 23, 2026

• Intrinsic open-sourced the robot control stack it uses in real manufacturing deployments, under Apache 2.0

  Sep 22

• ZCode packaged developers' whole .git history for upload, and Z.ai open-sourced the client in response

  Sep 22

• Xiaomi's MiMo-V2.6-Pro tops the open-weights index, but its own table puts Opus 5 ahead

  Sep 22

Musing

Quality or quantity: what kind of smarter is AI becoming?

Musing

Humanity's place in an AI-controlled world

Musing

Genie, oracle, rogue, god or just software: where AI could end up

Musing

Eleven problems that would change everything, and the one that could crack the rest

Guide

Claude models explained: Opus 5.5, Fable 5.1, Mythos and why there's no Opus 5.1

Guide

Codex models and plans explained: GPT-6 Astra, Sol and Luna, and which plan gets which

Pick one, change the bits that are yours, paste it into ChatGPT or Claude.