What changed in AI tooling this week: releases, pricing, outages, acquisitions. Every item links to its source.
Ponytail is a plugin that forces coding agents to reuse repo code, stdlib, and native APIs before writing new lines. It took 3 on Product Hunt on Sept 5 and installs on Claude Code, Codex, Cursor, an…
dif.sh stores feature flags and A/B tests as git reviewed Markdown files, builds a typed client, and writes dif/context.json so agents like Claude Code and Cursor know what’s live. It hit 1 on Produc…
AWS MCP Server now ships a serverless capability so agents like Claude Code and Kiro can inspect Lambda and linked API Gateway, EventBridge, S3, DynamoDB, SNS, SQS, and Step Functions in one call—inc…
Armature ran 17k sandboxed sessions and found Claude Code, Codex, and Cursor pick the same third party tool in only 42% of cases. Codex searches the web in 94% of sessions; Claude Code in 30% and bui…
Google’s Gemini Enterprise custom MCP data store lets admins point the assistant at their own MCP servers over StreamableHTTP + publicly trusted TLS (SSE and self signed certs are out). Tools import…
OpenAI’s local Rust Codex CLI landed on GitHub’s daily trending chart on Sept 6 with about 121,800 stars. It runs on your machine under Apache 2.0 and auth works via ChatGPT login or an API key.
Anthropic’s Claude Code v2.1.261 adds /skill doctor to list unused loaded skills and their context cost, plus clearer org policy failure reasons on /status and claude doctor . The release also raises…
OKF Agent Memory stores project knowledge as Markdown in a repo knowledge/ folder, with an embedded MCP server for Claude Code, Cursor, and Codex. The Go binary claims sub 300µs BM25 search and no ve…
In “An Alien Mind,” Jakub Pachocki says no frontier lab has solved alignment and monitoring enough to keep scaling at full speed, and urges mandated safety bars — three days after GPT 6 Astra shipped.
Kit is an MIT licensed static binary that gives agents one compose tool for edits, tests, retries, and subagents. Its Sept 6 Product Hunt launch stresses ACP/A2A so it can run across editors or orche…
GitHub’s HydraFusion research preview in Copilot CLI routes each coding task across models (single, cascade, or critique). Cost fell vs Claude Opus 5 on all three benchmarks (up to 67%), but quality…
AgentGate verified 19 AI agent npm packages as malicious in their shipped tarballs. As of Sept 6, 18 were still installable on the public registry — weeks after OSV/GHSA flags — including ones that s…
GitHub enabled GPT 6 Astra across Copilot Pro+, Max, Business, and Enterprise — including VS Code, Copilot CLI, and the coding agent. It is billed at provider list pricing under usage based billing,…
Manifold Security’s GitSpawn research shows seven CLI coding agents can execute a repo’s core.fsmonitor (or similar) command outside the sandbox with no approval prompt. goose, Claude Code, Cursor, a…
Massachusetts lawmakers are weighing quarterly independent catastrophic risk reviews for frontier models. Anthropic supports the stricter language; OpenAI and Google argue for Illinois style annual t…
CNBC reports buyers are struggling to keep up after Claude Fable/Mythos 5.1, Muse Spark 1.3, Gemini 3.8 Flash, and GPT 6 Astra all landed in the same week. Runpod’s CEO called the froth “model fatigu…
Claude Code 2.1.259 adds managedMcpServers so admins can distribute HTTP/SSE MCP servers org wide, plus permission prompts none so headless runs deny prompts instead of hanging.
Anthropic says Claude agents produced the first end to end, computer checked FLT proof in Lean over 11 days — about 13M lines of code and 29,500 intermediate theorems. Human input was mostly high lev…
Google released Mantis, an open source agent harness that finds, sandboxes reproduces, and patches vulnerabilities — aimed at cutting hallucinated findings from naive AI code scanners.
ARC Prize ran GPT 6 Astra two ways: 99.9% under OpenAI’s Provider Adapter harness, 62.7% under the benchmark’s standard harness. Same weights, different scaffolding — and ARC Prize still says it isn’…