NEWS

Codex says gpt-5.5 is not supported with a ChatGPT account because OpenAI retired it from ChatGPT sign-in on October 14. The fix and what still works.

Tencent WorkBuddy is a desktop AI agent for office work running Hy3, DeepSeek and more via TokenHub. What it costs, the free tokens and when to skip it.

SneakPeek (CVSS 8.7) let a poisoned HTML file read SSH keys and cloud credentials via the Claude Code Desktop preview pane. Fixed in Claude Desktop 1.30096.1.

LFM2.5-24B-A2B vs Qwen3.6-35B-A3B for local agents: Liquid wins tool-use benchmarks, Qwen wins SWE-Bench Verified. Which one to run for which job.

Ollama 0.40.4 runs decision models locally: ollama run nimble --ask answers yes/no, label and score questions with probabilities. Commands and limits.

Anthropic's Frontier Red Team says small open models may be months from self-replicating across cheap IoT devices. What it measured and where the claim is thin.

REA (Reverse Engineer Anything) is an MIT-licensed MCP server that lets Claude Code, Codex or Cursor inspect binaries, Electron apps and APKs. How to try it.

Anthropic Scion replaced git after merges grew to over 30,000 a day. What Anthropic's post says, what HN pushed back on and what to apply to your own repo.

Claude Code Projects is open to waitlisted Pro and Max users. How the coordinator and cloud threads work, what they cost in plan limits and where it breaks.

Claude for Startups paused its free Claude Team year and $1,000 API credit offer at capacity. What the program page says now and how to cover the gap.

OpenCode 2 upgrade from v1: what the migration guide says breaks (plugins, server API, TUI config), what carries over, and a safe checklist before you switch.

Talorys deploys a personal AI agent to your own Cloudflare account with one npx command. How far 10,000 free daily Neurons go and where "self-hosted" breaks.

Anthropic's unintended model actions report confirms a Claude model filed a false police tip. The four behaviors it lists and the scoping mistake behind each.

Anthropic OSS Scanner sends free AI vulnerability reports to eligible open-source projects. What the enrollment PR needs, what you receive, and the triage cost.

DeepSeek 4.1 Flash as a daily coding model: what a session costs, how peak hours double the bill, why caching matters, and where the subsidy argument holds.

Step 5 Preview is StepFun's 600B MoE with 27B active and a 1M window. What it scores, what a session costs, and what to check before the October 15 weights.

Claude Max and Team plans now include a monthly Claude API credit, starting at $100 on Max 5x. What it covers, what it does not, and how to claim it.

Google's Gemini agent can create coworker agents with their own Workspace account, email and audit trail. What Google announced and what builders should copy.

OpenAI's math repo holds hundreds of model-written manuscripts, some with Lean proofs. What Lean verifies, what it misses, and the lesson for agent output.

Claude Haiku 5.5 pricing, the cliff above 100K-token prompts, the tokenizer catch, the migration steps from Haiku 4.5, and when to use it instead of Sonnet.