Claude Code with GLM-5.3: The Setup That Works
Use a provider wrapper, disable the incompatible Artifact tool for Z.ai, and register GLM models with behavesAs. Test an interactive tool call: print mode alone missed the failure.
AI Agents, Growth & Startups
Detailed writeups on multi-agent orchestration, AI engineering patterns, and what actually works in production.
Beside You, Not Between You
The best agent UX was shipped by Google Docs over a decade ago: the anonymous animal cursor. Sunil Pai named the thesis: an agent beside you in a shared canvas, not between you and your work. tldraw, Clicky, and my own diagramming project all point the same way. Even Cursor, the autonomy leader, is converging there.
A Software Factory Is No Substitute for Maturity
Agent fleets don't remove organizational dysfunction, they write it down. Agents still need a clear objective, someone who can make the call, and a rule for when to stop and ask. A shop with its act together gets faster. A shop without one ships its confusion at a much higher rate.
A Deep Research Agent That Survives Its Own Failures
Durable Researcher is a browser-native deep research agent that checkpoints every model step, rebuilds state from the transcript, routes tasks by answer shape, and uses eval failures as the product loop.
An autopsy of Claude Code's deep research
I had Claude Code pry its own deep-research workflow out of its binary, then pointed that workflow at a question about itself. The verdict: it searches wide and never doubles back. The systems it resembles do. The whole game is that second hop.
A Harness for Every Run
Anthropic named the three ways long agent runs fail: agentic laziness, self-preferential bias, goal drift. Building Wire, a browser agent on Steel, I'd fought all three without the words. What the post nails, and three things the browser substrate forces you to add.
The Model Is the Smallest Decision You'll Make
Picking a model feels like an architecture decision. It mostly isn't. Swap a model under a fixed harness and your numbers wiggle; rebuild the harness under a fixed model and they move a lot. The only benchmark that gets a vote is your own workflow. A leaderboard orients you; it can't tell you what survives your production loop.
Serve Markdown to Agents, HTML to Humans
Same source, two channels. One middleware file parses Accept headers, falls back to UA sniffing for known bots, and respects sec-fetch-dest as a browser safeguard. HTML keeps a Link rel=alternate header so agents can discover the markdown variant. Plus llms.txt and llms-full.txt as prerendered static files. About 500 lines total.
Agent-Native Extensibility
The next evolution of extensibility isn't a better plugin system. It's agent-readable recipes that teach coding agents how to adapt external capabilities into your app's contracts. The extension is no longer just code — it's an installation procedure plus judgment.
Trained Qwen to Write Clojure Better Than GPT-5.4 (Kinda)
Fine-tuned Qwen3 on Clojure. 30B SFT hits 83.8% best-of-16, smashing GPT-5.4's 64%. But RLVR with shaped rewards actually lowered the ceiling—the verifier loop matters more than the training method. Built a deployable agent from it. Data quality was the bottleneck all along.
Request an AI summary of this blog
