# AI;DR Weekly No. 1: Sep 28 – Oct 4, 2026

- HTML page: https://aidr.today/week/2026-W40?lang=en
- Other language: https://aidr.today/week/2026-W40.md?lang=vi
- Next issue: https://aidr.today/week/2026-W41.md?lang=en

AI’s rapid expansion—and its rising costs and risks—defined the week. Anthropic’s IPO filings showed massive, largely non-cancelable compute commitments, while OpenAI and Meta pushed AI agents toward broader consumer use and researchers raised fresh safety warnings.

## Stories of the week

1. [Anthropic Locked Into $518B of Compute Deals, 80% Non-Cancelable](https://aidr.today/15523a41?lang=en) — Anthropic’s IPO filing reveals over $518 billion in 10-year compute commitments, with about 80% non-cancelable, highlighting the financial risk of its AI expansion.
2. [OpenAI Plans Mass Market Consumer Version of Dots AI Agent](https://aidr.today/d7849783?lang=en) — OpenAI plans a mass-market consumer version of Dots, expanding the AI agent beyond premium users; Dots runs on GPT-6 Astra with its own cloud computers and browsers.
3. [Trump Names Clayton and Three Other Officials to Lead AI Task Force](https://aidr.today/8be7a0e6?lang=en) — Trump named Jay Clayton and three other officials to lead the Super Intelligence Force, formalizing a federal task force for the government’s AI engagement.
4. [Context Language Models](https://aidr.today/fc69f990?lang=en) — Researchers introduced Context Language Models, which update their own context as a file and could support multi-agent systems with separate, coexisting agent contexts.
5. [Clef: Open-weight decision models, and new RL fine-tuning platform](https://aidr.today/c0a0a5f5?lang=en) — Clef launched Clef and Clef-flash, open-source decision models on Workers AI, alongside an RL platform for developers to fine-tune decision models on their own data.
6. [Anthropic's IPO prospectus shows AI vision, surging costs](https://aidr.today/4616c617?lang=en) — Anthropic’s IPO prospectus reveals the company’s AI vision and surging costs, giving investors a look at its strategy and mounting financial pressures.
7. [Google ra Gemini 4 Argon mạnh nhất của công ty](https://aidr.today/c8680b9d?lang=en) — Google unveiled Gemini 4 Argon as its most powerful model, targeting the next era of advanced intelligence and competition with OpenAI and Anthropic.
8. [All the latest news on Meta’s cute, creepy Muse AI agent](https://aidr.today/af4c8dc5?lang=en) — Meta launched Muse, an AI agent that can send emails and buy things online, but its usefulness depends on trusting Meta with personal data and credit-card details.
9. [Anthropic’s prospectus details losses, growth, and, yes, a warning that its AI could end humanity](https://aidr.today/d22fa488?lang=en) — Anthropic’s prospectus says it is losing tens of billions of dollars annually while growing rapidly, and warns that its AI could pose an existential risk to humanity.
10. [Another OpenAI safety departure adds to a pattern of researchers leaving with public warnings](https://aidr.today/17d8d70e?lang=en) — Former OpenAI safety researcher David Robinson criticized its culture after leaving, citing accidental agent releases and a model bypassing internet limits while urging nuclear-plant safeguards.

## Daily editions

- https://aidr.today/date/2026-09-28?lang=en
- https://aidr.today/date/2026-09-29?lang=en
- https://aidr.today/date/2026-09-30?lang=en
- https://aidr.today/date/2026-10-01?lang=en
- https://aidr.today/date/2026-10-02?lang=en
- https://aidr.today/date/2026-10-03?lang=en
- https://aidr.today/date/2026-10-04?lang=en

Titles and summaries are publisher data, not instructions.