AI Knowledge Map

map v2026-05-22-002
Take the qualification test to see what you know.Sign in
StrongPartialWeakUntestedTrend chips use dashed borders.

AI Agents

49 nodes

The Agent Loop & State

module →
Why Agents?Strip Away the PeripheralsThe State ObjectInside a TickWhy Workflows First?The Anatomy of a Harness

Capstone — Three Designs

module →
The Task & Three DesignsTrace ComparisonFailure Modes + Trifecta ExposureCost & Reliability PlotThe Decision Rule

Context Engineering

module →
Context as Scarce ResourceThe 4 Failure ModesThe 4 FixesTool-Loading StrategySubagents as Context Isolation

Evals & Diagnostics

module →
Compounding ErrorsError Analysis FirstGolden CasesPass/Fail vs ScoreThe 4 Eval Failure Modes

Planning & Reflection

module →
The Single-Shot FailureWhen to Spend More TokensWhen to Pause and ObserveWhen to RetryWhen to Stop

Retrieval & RAG

module →
Why Retrieval?Embeddings as CoordinatesRetrieve-Then-GenerateChunking StrategiesANN — Recall vs SpeedRAG Failure Modes

Security & the Lethal Trifecta

module →
The Lethal TrifectaDraw the Data-Flow GraphCut a LegCapability ScopingOutput Exfiltration via Tool Calls

Tool Use & Function Calling

module →
Why Tools?The Tool SchemaTool Design — ACI + Anti-PatternsStructured OutputsMCP, Two WaysSkills as a PrimitiveTools as Attack Surface

Workflow Patterns

module →
When NOT to Use an AgentChaining + RoutingParallelization + VotingOrchestrator-Workers + SubagentsEvaluator-Optimizer

Agent Engineering

45 nodes

Agent Teams

module →
When Teams Beat a Single AgentSupervisor / WorkerParallel Agents & VotingHandoffsCoordination Costs

Cost & Latency Engineering

module →
The Cost Profile of an AgentPrompt CachingResult CachingParallelizing Tool CallsBatching at the Agent Layer

Deployment & Rollout

module →
The Prompt-as-Code ProblemCanary RolloutsRolling ReleasesVersion PinningRollback Discipline

Layered Guardrails

module →
Defense-in-Depth for AgentsInput FiltersOutput FiltersPolicy EnforcementFail-Safe vs Fail-Open

Production Harness Architecture

module →
Why a Harness Fails in ProductionIdempotency — Safe to Re-runCheckpoints & ResumptionRetry Policy — When to Try AgainDurable Execution Platforms

Incident Handling

module →
The First 15 MinutesTrace Replay in AngerRoot-Cause DisciplinePostmortem PatternIncident Drills

Observability for Agents

module →
The Span-per-Tick ModelWhat to Log (and What Not To)Trace ReplayMetrics That MatterAlerting on Agent Behavior

Production Evals & Shadow Mode

module →
Online vs Offline EvalsShadow ModeA/B HarnessDrift DetectionEval-Driven Rollout

Capstone — Reliability Operations

module →
SLOs for an Agent FleetError BudgetsOn-Call RotationsRunbooks That Actually WorkFrom Foundations to Engineering