August 3, 2026
From the founder
The MCP protocol matured significantly this week with the stateless 2.0 spec, and I keep coming back to the same question it raises: if tool-calling infrastructure is finally stabilizing, the real constraint shifts to model capability and cost at the edge. DeepSeek V4 Flash's price-to-performance profile and OpenAI's 10M-user Codex story both point at the same thing, agentic systems are leaving the lab and hitting production scale, which means the plumbing decisions you make today are load-bearing.
Simon Willison
Tuesday was Stateless MCP day - the rollout of MCP 2.0, or the 2026-07-28 Model Context Protocol specification to use the more formal but less memorable name. This is the most significant change to the MCP spec since it first launched, and has also served to reignite my personal
Simon Willison
TIL: Adding a custom MCP server to Claude and ChatGPT Connecting a custom MCP server to Claude and ChatGPT's standard chat interfaces is possible, but can take quite a few steps. Tags: ai , generative-ai , chatgpt , llms , claude , model-context-protocol
Latent Space
[AINews] Fearing RSI: OpenAI, Anthropic, GDM, Meta, Thinky cosign letter to "Pace" AI development, as HuggingFace details Machine-Speed Offensive Cyberattack (Latent Space) — read the full story for details.
Simon Willison
deepseek-ai/DeepSeek-V4-Flash-0731 The latest release in DeepSeek's V4 family, "with substantially enhanced agentic capabilities". It's 304 billion parameters - 167GB on Hugging Face - but it appears to punch well above its weight. Artificial Analysis rank it ahead of MiniMax M3
Latent Space
OpenAI's product engineering lead details how Codex scaled from zero to 10M users by building modular agentic features — subagents, memory, no-code orchestration — directly into ChatGPT Work. The architectural decisions around subagent coordination and memory persistence are directly transferable to teams building enterprise agent products. The advice on avoiding premature abstraction in agent pipelines is worth the read time.
The latest in agentic & applied AI, every week.