Harness engineering debate: LangChain fires back at Matt Shumer's 'harness is stupid' take
hwchase17 · x · 2026-11-01
Matt Shumer sparked debate arguing harness engineering is 'usually stupid' — just use the harness the model was trained for and write great behavior prompts. Vtrivedy10 at LangChain took the exact opposite view: building a great harness is deeply core to building a great product.
He cites examples from Grok Bot videos, Factory blogs, LangChain and OpenAI Codex blogs, framing the harness as the user experience layer for agents, including:
- Tuning prompts (which are part of the harness)
- Building bespoke tools/skills that compress common workflows into single tool calls to save context
- Aggressively tuning compaction thresholds per use case
- Offloading work to cheaper models via subagents/routing
- Building connectors so agents can efficiently access and search context
- Supporting background work so users get immediate feedback
The clash hits a core tension in agent development: lean into the model's native shape, or invest in engineering around it for differentiated experiences.
Related event: Shumer Sparks Debate: Is Harness Engineering a Mistake?(3 posts)→
More from coding & agent
- Blogger lets Grok spend up to $100 autonomously and receives a LEGO Harry Potter gift — venturetwins · 2026-11-01
- TensorAgent Adds Local Multimodal Indexing for Documents, Images, Audio and Video — fuzhongkai · 2026-11-01
- Theo claims new model hallucinates 10x worse in real codebases; bugs traced to ACP integration — cgarciae88 · 2026-11-01
- One prompt to an AI agent adds note-taking and search to X bookmarks — HaijunXia · 2026-11-01
- Running 16 desktop AI agents at once on a single Mac, clicking via accessibility refs — cramforce · 2026-11-01
- Grok helps user find 74GB Cursor chat backup, doubles free disk space — prasenx · 2026-11-01