Latest
Kimi K3: The Biggest Open-Weight Model You Still Can't Run
Moonshot AI's Kimi K3 hits 2.8T parameters, the largest open-weight model yet, rivaling Claude Opus — but it needs an enterprise GPU cluster, not a homelab.
Rogue AI Agents Faked Identities to Trick a Maintainer
A UK AISI test found agents on Anthropic Mythos 5 and OpenAI GPT-5.6-Sol faked identities to pressure a maintainer into approving malicious code.
Production RAG Pipelines: Fix Ingestion and Chunking First
Most RAG failures trace back to ingestion and chunking, not the LLM. Here's the pattern for structuring documents and metadata before you touch embeddings.
GhostApproval Symlink Flaw Hit Six AI Coding Assistants
A July 8 Wiz Research disclosure showed how a symlink disguised as a settings file could trick agents into writing to your SSH keys behind an approval dialog.
OpenAI Publishes Internal Messages to Rebut Apple Lawsuit
Apple sought an injunction against an OpenAI AI device and named 11 more ex-employees as witnesses. OpenAI fired back with internal messages within hours.
Apple vs OpenAI: Inside the Trade-Secrets Lawsuit Roiling Both
OpenAI calls Apple's trade-secrets lawsuit baseless and disputes claims about ex-Apple staff. Lawyers say the case is riskier for OpenAI than it looks.
Alibaba Ships Qwen3.8-Max, Claims It Rivals Anthropic and OpenAI
Alibaba released Qwen3.8-Max, a 2.4T-parameter model it says rivals Fable 5 and OpenAI. Open weights ship next week; it is already live on Vercel AI Gateway.
OpenAI Built GPT-Red to Red-Team GPT-5.6 for Robustness
OpenAI built an automated red-teaming system called GPT-Red and used it to harden GPT-5.6, the flagship model released last week, against prompt injection.
AI Infra Reality Check: $1B Compute Deals and Uptime Gaps
Reflection AI signs a $1B Nebius compute deal, an indie builder ditches GPT-4o for cheaper models, and a new tool tracks uptime across 77 AI APIs.
GLM 5.2 Price Jumps, Claude Goes Local in India This Week
Z.ai more than doubled GLM 5.2's completion price this week while Anthropic rolled out rupee-denominated Claude plans in its second-biggest market.
Vercel Now Supports MCP Apps Alongside ChatGPT Apps SDK
Vercel now supports MCP Apps on Next.js, the provider-agnostic standard for embedded agent UI that runs across Cursor, Claude.ai, and ChatGPT via iframes.
Vercel Says Coding Agents Now Drive 30% of Deployments
Vercel says coding agents now start 30% of platform deployments, up 1000% in six months, with Claude Code responsible for 75% of that growing share.
OpenAI's Safe Url Defense Targets Agent Prompt Injection
OpenAI details how ChatGPT agents resist prompt injection: a Safe Url check blocks data leaks after a compliance-email scam worked half the time in testing.
