← back
Why (Senior) Engineers Struggle to Build AI Agents — Philipp Schmid, Google DeepMind
Takeaway
Design agents for contextual decision-making, recovery, measurable outcomes, and clearly documented tools.
Summary
- Agent development uses an iterative run-observe-adjust loop and specifies goals more often than exact execution paths.
- Textual context can preserve nuanced preferences and changing user intent that rigid state flags fail to capture.
- Treat errors as actionable input so agents recover without repeating expensive long-running work.
- Evaluate repeated outcome reliability, trace behavior, and expose self-documenting tool interfaces rather than assuming the model knows the surrounding codebase.
agent-designerror-recoveryevals
Original description
A `deleteItem` endpoint is obvious to the developer who built it. An agent only sees the function schema and docstring. Philipp Schmid from Google DeepMind argues this is why senior engineers struggle most: they carry years of implicit context that agents do not, and design tools assuming it. He names four other shifts: text replaces structured state, errors are inputs not restart triggers (especially costly when an agent has been running for 15 minutes), evals replace unit tests because the right question is how often it works not whether a fixed input always produces a fixed output, and build to delete because you will rebuild the same agent with a better model anyway. Speaker info: https://x.com/_philschmid / philipp-schmid-a6a2bb196 https://github.com/philschmid