๐Ÿš€ New: chi (ฯ‡) โ€” an open-source autoresearch harness for fleets of LLM coding agents. Read the announcement.

llm

31 posts

Claude Degraded for a Month. The Cause Was a Workaround Nobody Fully Understood.
I Put Hermes on a Mac Mini. The Login Failures Were the Most Useful Part
Real-SWE: The Benchmark Where Coding Agents Meet an Actual Enterprise Codebase
Sakana AI's Fugu Ultra v2: When Orchestration Beats a Bigger Model
Anthropic's Distillation Report: How 200 Million Conversations Became a Threat Report
OpenAI's Agents API: Renting the Codex Harness Instead of Building Your Own
Mercury 2.5: When an LLM Stops Writing One Token at a Time
OpenAI's Chief Scientist Says Its Best Safety Tool Is Losing Its Grip
OpenAI's 'Wiki Incident' Shows Disclosure Rules Weren't Built for Agents
Microsoft's MAI-Transcribe-2 Undercuts Every Rival on Price and Speed
GPT-6 Astra Is OpenAI's First Model Rated Critical for Cybersecurity
Claude Fable 5.1 Cuts Cache Reads 75% While Doubling a Key Agent Benchmark
Anthropic Rebuilds Its Data Retention Deal Around Your Own Cloud
Claude in Chrome Goes GA, and Now It Acts Without Asking First
GitHub Copilot Turns Reasoning Depth Into a Dial You Control
Claude Code's Restricted Mode Closes the Config Escape Hatch on Sandboxed Agents
Anthropic's Model Hardware Standard Is MCP for Lab Robots and Factory Floors
Grok Bot Learns to Explain Itself on Video, and the Context Bill Comes Due at 200K
GLM-5.3 Lands: Z.ai's Coding Model Drops the Off Switch on Thinking
Google Kills Imagen 4 Today: The generate_images() Call You Have Is Now Broken
OpenAI's Ultrafast Tier Runs GPT-5.6 Sol at 750 Tokens a Second, and It Changes What 'Real-Time AI' Means
DeepSeek V4 Pro Goes GA, and LLM APIs Get Their First Peak-Hour Pricing
Gemini's Managed Agents Get a Kill Switch: Hooks, Budgets, and a Free Tier
MCP Just Went Stateless, and Your Server Can Now Live on the Edge
Meta's Muse Glimmer Puts a 30B Agentic Model on a Single Consumer GPU
Claude Code Turns Auto Mode On by Default, Because You Approve 97% of Prompts Anyway
Claude Code Adds Spend-Limit Warnings and a Trust Prompt for Autonomous Agents
Claude Code Sessions Can Now Run on Your Own Servers and Talk to Each Other
Building Chi (ฯ‡): An Autoresearch Harness for LLM Coding Agents
One Django Codebase, Four LLM Vendors
Prototype on Dify, Ship on Celery