computational-sensors-firstDeterministic checks before LLM judges
Wire linters, test suites and schema validators into the agent's loop before adding any model-based judgment.
Sensors verify output after execution. Computational ones — a linter, a type checker, a unit test, a schema validator — are fast, free, and deterministic. Inferential ones — an LLM-as-judge, an AI code review — add semantic judgment but cost tokens and return non-deterministic verdicts. The rule is to exhaust the deterministic checks first and add an inferential sensor only for a property no rule can express, treating its verdict as an advisory signal rather than a gate until it has been measured against human review.
Requires: An existing test suite, linter or validator that the loop can call. Where none exists, the first sensor has to be written before the technique applies at all.
Contexts: Coding agent, Autonomous agent
Does it work?
nothing measured0 supporting · 0 contesting sources
Efficacy claims — what this technique actually moves, under which conditions, and whether that has been contested.
No efficacy claim filed yet. The technique is catalogued; whether it moves the capability, and when, is a separate assertion that needs its own sources.
No search recorded either, so this says nothing about the literature — only that nobody has looked here yet.
Code
No repository linked yet. Contribute one.