Why Eval++ Is the Next Great Compute Primitive
summary
Claims that Durable Objects are the right compute unit for agents — addressable, persistent, hibernating, stateful — and that the Agents SDK adds resumable streaming, multi‑tab sync and background scheduling without userland distributed systems work. Measured London latency of 15ms London latency fits inside an animation frame. Demonstrations include real-time collaborative apps and MCP servers; a technique called Dynamic Workers runs LLM‑generated code in sandboxed isolates with explicit capability grants. They report collapsing 2,600 APIs into a 1,000‑token MCP tool and say parts are already shipping.
AI-written summary · about
chapters (6)
- 0:00Introduction to Cloudflare AI
- 2:17Durable objects architecture
- 6:54Dynamic workers and code mode
- 11:36Real-world agent use cases
- 15:29Future vision and agent tooling
- 19:00Q&A and closing remarks
related talks
21mBuilding Agents (the hard parts!)
29mFrom Copy-Paste to Composition: Building Agents Like Real Software
32mProgramming language evolution in the AI era
01h 03mAI Coding Agents and how to code them
36mPutting AI Agents to Work: A New Era of Open Source Connectivity with MCP
12mMCP and the Dawn of Agentic Intelligence
36mComposable Architecture for AI Systems: APIs, MCP, Skills, Sub-Agents, and Wor…
49mThe Headless Web