On mvanemmerik.com, Nov 16: NFL Week 10 chalk talk: Detroit dazzles, Cincinnati takes the North
vanemmerik.aiClaude, ChatGPT and Grok news, twice a day

Sunday, November 1, 2026

ChatGPT and Codex will share one usage pool in December; Claude stops rebuilding existing games

OpenAI says ChatGPT and Codex will draw from a single allowance for Plus, Pro and Business users from December, with Instant chats exempt. Claude has new guidance against recreating specific existing video games, and Anthropic set a Nov. 30 shutoff for Opus 4.8 fast mode. Elon Musk said Grok Bot had cost too much to serve as users hit weekly limits early, and Sam Altman met about 50 Samsung presidents in Korea amid tight memory supply.

Sunday, November 1, 2026

ClaudePolicy

Claude's instructions now tell it not to rebuild specific existing video games, even renamed ones

A user spotted new guidance telling Claude not to recreate a particular existing game, even if it's relabeled as original, and Anthropic staffer Addy Osmani confirmed the change on X, citing the usage policy's ban on IP infringement. Original games, genre games and games built from a user's own designs and assets are still allowed, and Osmani said Anthropic will keep adjusting the balance. Engadget ties it to a wave of browser-playable clones of big-budget games.

Why it matters: It's a visible new limit on what Claude will build, aimed at AI-assisted copies of commercial games.

ClaudeHappened Nov 1, 2026 · Sources: Engadget

GrokBusiness

Musk says Grok Bot cost too much to serve as users report hitting weekly limits earlyReported

Replying on X on Nov. 1 to a user who said many people were burning through their weekly Grok Bot allowance unusually fast, Elon Musk wrote that the agent's serving cost had been unsustainably high relative to revenue. xAI hasn't announced a change to the limits, which it doesn't publish as fixed numbers, or said whether prices or resets will change.

Why it matters: It suggests the always-on agent's usage allowances are being tightened to control compute costs.

Grok BotHappened Nov 1, 2026 · Sources: Tesla North · Startup Fortune

OpenAIPolicy

ChatGPT and Codex will draw from one shared usage allowance starting in December

An OpenAI help-center article updated Oct. 31 says Plus, Pro and Business subscribers will get a single token-based allowance covering both ChatGPT and Codex from December, with no exact start day given. Chats in Instant mode won't count against it. OpenAI hasn't published the allowance sizes or reset cycle, and the page doesn't say how Free, Go or Enterprise plans are handled.

Why it matters: Heavy Codex users and heavy ChatGPT users will be competing for the same pool, which could change how far a plan stretches.

ChatGPTCodexHappened Oct 31, 2026 · Sources: CellCog · XenoSpectrum

OpenAIBusiness

Sam Altman meets about 50 Samsung presidents in Korea as AI memory supply stays tightReported

Samsung said OpenAI CEO Sam Altman spoke with about 50 presidents from its electronics, finance and biotech affiliates at its Yongin training center on Oct. 31, covering where AI is headed and how it will change management. No new agreement was announced; Samsung already supplies memory for OpenAI's Stargate data centers. OpenAI has not posted about the meeting.

Why it matters: Memory chips are one of the tightest constraints on AI build-outs, and Samsung is among OpenAI's key suppliers.

OpenAIHappened Oct 31, 2026 · Sources: The Next Web

ClaudeFeature

Fast mode for Claude Opus 4.8 is deprecated and will stop working on Nov. 30

Anthropic's platform release notes say fast mode for Claude Opus 4.8 is deprecated, with removal set for November 30, 2026. After that, API requests to claude-opus-4-8 with speed: "fast" will return an error; developers who want fast mode are pointed to Claude Opus 5.5.

Why it matters: Apps that rely on faster Opus 4.8 output have a month to switch models or drop the setting before requests start failing.

Try it: Search your code for speed: "fast" on claude-opus-4-8 calls and move them to claude-opus-5-5.

Claude APIHappened Oct 30, 2026 · Sources: Claude Platform release notes

ClaudeBusiness

Nvidia is reportedly in talks to back billions in Nscale debt for the West Virginia data center Anthropic plans to useReported

The Information reported, and Reuters carried, that Nvidia is discussing billions of dollars in credit guarantees to help Nscale fund power and construction at its West Virginia AI campus, where Anthropic agreed in August to rent $45 billion of computing. Nscale is also said to be negotiating a second deal of about 460 MW with Anthropic; nothing is signed, and Nvidia, Nscale and Anthropic did not comment.

Why it matters: It shows how much of Anthropic's compute build-out depends on outside financing, at a time when AI data-center debt is under scrutiny.

AnthropicHappened Oct 29, 2026 · Sources: Business Standard (Reuters) · investingLive

OpenAIFeature

Codex CLI 0.166 adds keyboard transcript selection and GPT-6.1 Terra on Vertex AI and Bedrock

Version 0.166.0 of OpenAI's Codex CLI adds keyboard-driven selection and search in the transcript, clearer explanations when a safety check blocks a request, a codex doctor --network-only diagnostic and codex queue --resume. It also adds GPT-6.1 Terra to the bundled, Vertex AI and Amazon Bedrock model catalogs, and fixes bugs around resuming and forking threads.

Why it matters: Teams running Codex through Google Cloud or AWS can now pick the cheaper GPT-6.1 Terra model from the CLI.

Try it: Update with npm install -g @openai/codex@0.166.0.

CodexHappened Oct 29, 2026 · Sources: ChatGPT & Codex changelog