Kelly Mears

KnowledgeDomain

Agents & Language Models

How systems built on language models actually work, and how they fail.

Notes
30
Within this domain
137
To other domains
90

This domain, and what it touches

Knowledge graph: 69 notes, 227 links.Each dot is a note, sized by how many links it has and coloured by topic. Focus or hover a dot to isolate it and its neighbours; activate it to open the note. The graph is a single tab stop: use the arrow keys to move between notes, Home and End for the most and least linked, Enter to open the focused note, and Escape to leave the graph.Silent FailureGround TruthDeterminismVacuous TruthTool UseLarge Language ModelDocumentation RotPlausible MechanismPrompt EngineeringProvenanceNondeterminismFalsifiabilityCode ReviewSecret ManagementContext WindowSystem PromptExhaustive ClaimGoodhart's LawGuardrailKnowledge GraphPull RequestTokenAgentic LoopFail FastFlaky TestHallucinationPromptPrompt InjectionRecord and Replay TestingTest FixtureCache InvalidationEvaluation HarnessFingerprintLinguistic RelativityRace ConditionSchema ValidationStructured OutputSubagentToken BudgetWikiAgent MemoryBacklinkCharacter EncodingGit WorktreeLLM-as-JudgeModel Context ProtocolMulti-Agent OrchestrationResource StarvationStructured DataTest DoubleUnreachable CodeZettelkastenAdversarial ReviewAgent SkillAnchoring EffectAssertionCode SigningConstrained DecodingDefensive DefaultFan-Out and Fan-InHuman in the LoopJSON SchemaLeast PrivilegeModel RoutingSchema DriftStreaming ResponseSupply Chain SecurityHash FunctionDraft and Published

Drag to pan · hold /ctrl and scroll — or pinch — to zoom.

The notes

Most connected first