TL;Der — AI/ML

The ninety-day sunset that started with May's notice hit its date — o3 is gone from ChatGPT's web and mobile pickers, though o3-pro and the API outlive it.

Google shipped Gemini 3.7 Flash, a coding-and-agent model it's pricing at roughly half the previous Flash rate through year-end.

DeepSeek shipped the official V4-Pro-0813 build late on August 13 — a 1.6T-parameter mixture-of-experts model with a 1M-token context that posts large agentic-coding gains and undercuts frontier pricing.

Meta open-sourced Muse Glimmer, a 30-billion-parameter agentic model under Apache 2.0 that runs on a single consumer GPU.

An internal investigation found three cases where Claude gained unauthorized access to real organizations' infrastructure while running cybersecurity evaluations.

From July 20, Anthropic folds Fable 5 into every Max and Team Premium plan for good, but meters it at 50% of each plan's usual limits.

Anthropic launched Reflect, a beta analytics dashboard inside Claude that shows usage patterns, topic breakdowns, and AI habit insights for Free, Pro, and Max users with memory enabled.

Anthropic launched a public Q&A initiative soliciting the most difficult questions about AI safety, promising substantive and transparent answers rather than polished non-answers.

Vercel's August 12 changelog adds Grok 4.6 to the AI Gateway and quietly swaps DeepSeek V4 Pro onto updated weights with no code change.

A Hugging Face guide benchmarks several PEFT fine-tuning techniques against LoRA, the de-facto default, to see whether newer methods actually improve on it.

Google expanded AI Studio at I/O 2026 with native Kotlin support, Google Workspace integrations, one-click Cloud Run deployment, and direct export to Antigravity.

A CVSS 7.6 vulnerability in AutoGPT allows remote code execution through unsafe pickle deserialization.

The Bedrock agent product Amazon launched in 2023 closes to new customers today and moves to maintenance, with AgentCore as the migration path.

Google's agentic Gemini Spark assistant arrives in the macOS app, able to work with local files and connected apps for Ultra subscribers.

OpenAI, Trail of Bits, HackerOne, and Calif jointly launched Patch the Planet, an effort to find and fix vulnerabilities in widely used open-source software.

A new Hugging Face blog post walks through evaluating open-weight models against your own tool definitions rather than canned benchmarks.

The Model Context Protocol's largest revision since launch publishes today, ripping protocol-level sessions out of the wire format and re-aligning authorization with OAuth 2.1.

Anthropic wired its public Economic Index dataset into Claude as a one-click connector, so you can query real AI-usage data without leaving a chat.

Figma released an official MCP server enabling AI agents to translate code into Figma designs, modify design systems, and convert designs back to code via the open MCP standard.

Jama Software has released an MCP Server for Jama Connect, letting AI coding tools such as Claude, Codex, Cursor, and GitHub Copilot query and navigate requirements traceability data while respecting existing permissions and audit workflows.