AI Engineer World's Fair Harness track: when agents break in production, the model is rarely the failure
Roger_M_Taylor · x · 2026-09-15
The Harness Engineering Track from AI Engineer World's Fair 2026 is live, with the thesis: when an agent breaks in production, the model is rarely the part that failed.
Talks include:
- 2026 State of AI Engineering (Amplify Partners)
- DSPy on the unreasonable effectiveness of separating the task from the model
- Two Anthropic sessions: "How Anthropic Builds" and "Tokens Should Have Jobs"
- AWS on building production cages for powerful domain agents (ports, proofs, kill switches)
- Morgan Stanley's "Loophole", Restate on every call, .txt's "Agent Frameworks Considered Harmful"
- PostHog's war story: letting an AI agent execute Bash and living to talk about it
Overall focus: the infrastructure layer around the model, not the model itself.
More from coding & agent
- Models don't have agency, systems do: the one-token trick that seeds tool use — sethjuarez · 2026-09-16
- GPT-6 Astra demos: playable games, robot demos and 3D pipelines built in hours — socialwithaayan · 2026-09-16
- Proposal: hand predictable UI workflows to a local runner to cut computer-use agent costs — Iinganouo · 2026-09-16
- Tutorial: Tracing agent invocations and LLM calls with OpenTelemetry — dl_weekly · 2026-09-16
- LangChain makes every Managed Deep Agent an MCP server callable by coding assistants — LangChain · 2026-09-16
- LlamaIndex on Building SDKs with Stainless, Whose Team Joined Anthropic — llama_index · 2026-09-16