LESSON 7.4 · Theory · 120 min
Planning, reflection, and multi-agent systems
Planning, reflection, and multi-agent systems. Learn decomposition, search, critics, and coordination cost and complete: Run ablations on a verifiable task. Part of the “From Model
Learning objectives
- Explain what problem “Planning, reflection, and multi-agent systems” solves without hiding behind terminology.
- Trace the variables and causal links across decomposition, search, critics.
- Complete “Run ablations on a verifiable task” and judge the result with evidence rather than intuition.
Core concepts
decomposition
decomposition is part of the lesson’s causal model. State its inputs, outputs, invariants, and failure mode; then verify it with a hand-check or a minimal experiment before moving to an optimized implementation.
search
search is part of the lesson’s causal model. State its inputs, outputs, invariants, and failure mode; then verify it with a hand-check or a minimal experiment.
critics
critics is part of the lesson’s causal model. State its inputs, outputs, invariants, and failure mode; then verify it with a hand-check or a minimal experiment.
and coordination cost
and coordination cost is part of the lesson’s causal model. State its inputs, outputs, invariants, and failure mode; then verify it with a hand-check or a minimal experiment.
Build and verify
Run ablations on a verifiable task
- Predict: write the expected output, trend, or failure before running code.
- Build: implement only the minimum components needed to answer the question.
- Verify: compare with a baseline or trusted implementation; save seeds, parameters, and raw outputs.
- Transfer: change one shape, dataset, scale, or workload condition and explain whether the conclusion still holds.