Fingerprint
A hash over everything that determined a result, used to detect when the result has gone stale.
A fingerprint is a hash computed over all the inputs that produced an artifact, stored alongside it, and compared later to decide whether the artifact is still valid. It is the general mechanism behind build caches, memoisation, and stale-recording detection.
Its value is that it makes staleness mechanical rather than a review step. If the fingerprint covers the prompt, the model identity, and every tool definition, then any drift in any of them turns into a failing check rather than a subtle behavioural difference nobody notices.
The cost is that the coverage is deliberately broad, so unrelated changes invalidate everything. A change that only rewords one tool's description stales every recording in the system, because the model saw that description in every request.
Two mechanical hazards recur. Fingerprints usually hash an ordered list, so anything that changes ordering without changing membership — for instance, a bundler that sorts module exports where another preserves source order — produces a different hash for identical content, across runtimes rather than across revisions. And a fingerprint computed against a merged preview differs from one computed against your local base, so a check can fail in one place while passing in the other, correctly.
See also8
Hand-picked in the note itself — the neighbours worth reading next.
Record and Replay Testing
Capturing a real interaction once and replaying it deterministically in later test runs.
Agents & Language Models12 connections
Hash Function
A function mapping arbitrary input to a fixed-size digest, used for identity, integrity, and addressing.
Systems & Tooling7 connections
Determinism
The property that identical inputs produce identical outputs, and the foundation of caching and verification.
Systems & Tooling27 connections
Cache Invalidation
Deciding when a cached value has stopped being correct — famously one of the hard problems.
Web Platform11 connections
Provenance
The recorded origin of a change or a claim — who made it, when, and on what evidence.
Method18 connections
Evaluation Harness
A repeatable test suite for model behaviour, since prompts and tool descriptions cannot be type-checked.
Agents & Language Models11 connections
LLM-as-Judge
Using a language model to score another model's output against a rubric.
Agents & Language Models10 connections
Tool Use
Giving a model a set of callable functions so it can act on the world rather than only describe it.
Agents & Language Models24 connections
Related2
Nearby in the graph rather than deliberately chosen. Looser, sometimes surprising.
Linked from9
Notes elsewhere in the wiki that reach for this one.
- Cache InvalidationWeb Platform
Deciding when a cached value has stopped being correct — famously one of the hard problems.
- DeterminismSystems & Tooling
The property that identical inputs produce identical outputs, and the foundation of caching and verification.
- Evaluation HarnessAgents & Language Models
A repeatable test suite for model behaviour, since prompts and tool descriptions cannot be type-checked.
- Hash FunctionSystems & Tooling
A function mapping arbitrary input to a fixed-size digest, used for identity, integrity, and addressing.
- LLM-as-JudgeAgents & Language Models
Using a language model to score another model's output against a rubric.
- ProvenanceMethod
The recorded origin of a change or a claim — who made it, when, and on what evidence.
- Record and Replay TestingAgents & Language Models
Capturing a real interaction once and replaying it deterministically in later test runs.
- Test DoubleTesting & Verification
A stand-in for a real dependency during a test — and the ways a stand-in can lie.
- Tool UseAgents & Language Models
Giving a model a set of callable functions so it can act on the world rather than only describe it.