Lead story / pricing change

OpenAI will combine paid ChatGPT and Codex limits in December

OpenAI plans to replace separate ChatGPT and Codex allowances with a single token-based budget for Plus, Pro and Business subscribers starting in December. Instant chat will remain outside the shared limit. The company ties the change to bringing Work’s agent capabilities into the main ChatGPT experience.

The meter will count tokens—the units of information a model processes and generates—covering both input and output. How quickly a subscriber uses the allowance will depend on the model and task. OpenAI says more advanced models and more complex work generally consume more of the budget.

The notice leaves the most useful comparison unresolved: how much capacity each plan will provide before and after the switch. OpenAI has not published plan-level token allotments, a reset schedule or an exact launch day. It expects most subscribers will not have to wait for resets more often, but that is a forecast, not a guarantee.

Subscribers will be able to check their combined remaining allowance and reset times in Settings → Usage. Once the shared budget is exhausted, OpenAI lists three options: wait for a reset, upgrade the subscription, or buy credits for eligible features. Those are the stated routes for continuing work when the allowance runs out.

policy

MIT drops AI camera monitoring after faculty and student pushback

MIT has ended its AI camera-monitoring evaluation after objections over access, misidentification and community input. The Ambient.ai pilot used staged safety scenarios, and its data has been purged. Roughly 4,600 cameras remain in use without AI, audio recording or facial recognition. Student-union critics say extensive recording can still chill expression, even without automated analysis.

launch

Anthropic will expand Claude text watermarks, but two models have a bypass

Anthropic will extend text watermarking to four more Claude models on November 3. On Opus 4.6 and Sonnet 4.6, changing any of three word-selection settings from its default disables the mark, according to The Decoder. Anthropic ties the retrofit to the EU AI Act’s December deadline. Its detector remains in private preview and cannot establish that unmarked text was written by a human.

infrastructure

Sam Altman says cloud AI makes better use of scarce memory

In an October 29 interview with John Collison, Sam Altman argued that constrained memory should be kept highly utilized in cloud systems rather than sit in lightly used PCs. His suggestion that PCs may use it only 5% of the time was a characterization, not a measurement across machines. Individual critics emphasized ownership and corporate control. Consumer memory also serves gaming and other non-AI workloads.

Also worth knowing

•  Cursor removes a malicious Claude Code extension that can leave malware behind

•  Nearly one in five surveyed U.S. workers say they were asked back after AI cuts

•  DoorDash says outside AI agents have placed only several hundred orders

•  Nine of 10 AI models raised their own crypto spending caps in a study

•  Anthropic renames its $200 Claude subscription Ultra

•  Aramco takes 49% of HUMAIN with a $2.6 billion cash contribution

•  OpenAI reportedly cuts its fundraising target by $10 billion

•  xAI will watermark supported Grok API output by default on November 7

•  Study finds 353 government sites block ChatGPT search but allow Google

•  Anthropic tells Claude Code to avoid specific game remakes

Daily tool drop

6 AI tools worth knowing today

The Internet Had a Point

AI video ruins the magic

Reader check-in

Help shape tomorrow's briefing

One click tells us what to keep, improve, or tighten.