Decides what
needs to happen.
Opus / reasoningA presentation for humans who ship
AI for
dummies.
You do not need an army of models. You need a system: one mind to frame the work, several hands to execute it, and a clean boundary between every task.
01strong model
for ambiguity
for ambiguity
03bounded workers
in parallel
in parallel
∞iterations
with evidence
with evidence
Read this as a route map, not a prompt recipe.
RULE ZEROStrong model for ambiguity.
Light model for bounded work.
Light model for bounded work.
A small fleetcoordination before parallelism
Decides what
→
The orchestrator preserves intent, writes small contracts, and gathers results that can be verified. It does not need to type every line.
Why the boundary mattersone vague task / three predictable failures
Every worker reads everything. Nobody knows which facts are load-bearing.
Two agents touch the same checkout. The fastest path becomes conflict resolution.
The diff is polished, but no one checks whether it solved the original problem.
The subagent loop
Click a phase.
See the handoff.
Delegation means moving one bounded task into a smaller context—not giving away responsibility.
What crosses contextsbrief → diff → evidence
| Package | Contains | Why it matters |
|---|---|---|
| Brief | goal, files, boundaries | stops the worker inventing the problem |
| Worktree | branch and isolated checkout | parallel edits do not collide |
| Checks | tests, build, criteria | turns “looks good” into evidence |
| Diff | small, reviewable change | integration and discard stay cheap |
Git worktrees
One branch
per hand.
A worktree is another directory linked to the same repository. Each agent gets its own checkout and index; history remains shared.
Select a node to inspect its checkout, owner, and next action.
Model routing
Do not pay for
reasoning where
you need rhythm.
Choose a job to see why the model profile changes.
WorkProfilePrompt shape
Model gearboxcapability tier × thinking effort
Two separate knobs
Choose the engine.
Then choose the gear.
A stronger model changes the capability ceiling. Higher reasoning effort gives that model more room to work. Start with the lightest combination that passes your real checks, then move one knob at a time.
REASONING / THINKING
ROUTING RULEUse strong models for ambiguity and judgment. Use lighter models for bounded execution. Raise effort only when evaluation shows a gain.
Skills
Write the right way
once.
A skill is a reusable procedure. It can carry instructions, references, scripts, and assets. It is not magical memory, and it does not replace acceptance criteria.
01 / trigger clearly02 / load detail on demand03 / return evidence
SKILL PACKAGE
name: review-ui · check focus, mobile, reduced motion · run verification · return evidenceCreate a skillrepeatable pain → reusable judgment
The skill forge
Teach the decision.
Keep the context light.
Do not package everything you know. Capture the non-obvious choices that repeatedly improve an outcome, then prove the skill changes behavior.
AFTER REAL USE
observe failure→sharpen one rule→retest behavior→keep it narrow
Common skillschoose behavior before model
The field kit
Different jobs.
Different instincts.
A skill changes how an agent approaches work. Some shape communication. Others enforce research, debugging, review, or completion discipline. Select one to inspect its operating rule and verified source.
ONE PRACTICAL LOADOUT
PLAN unlazy → BUILD ponytail-lite → DEBUG diagnosing-bugs → REPORT caveman
INSTALL PACKAsk your coding agent to verify, install, and validate the skills.
Hands-on10 minutes / one missing feature
Tiny Tasks lab
Same task.
Better operating system.
Start with a deliberately incomplete static task board. Run one prompt as written, reset, then run the skill-enabled version. Compare diff size, verification evidence, and unnecessary complexity.
THE MISSING FEATUREAdd All / Open / Done filters that survive reload and browser navigation.
STACK HTML · CSS · JavaScript DEPENDENCIES none FILES 3
RUN AGood prompt
RUN BGood prompt + skills
COMPARE THE RUNS
01 Files changed
02 New dependencies
03 Checks actually run
04 Evidence returned
Verificationevidence over confidence
From confident drift to decisive proof
Three layers.
Run each separately.
Every gate needs its own line, its own exit code, and its own evidence. The matrix below is the deliverable — not opinion, not prose, not "looks done".
Shape and types
Format, lint, type-check. Fast and scope-bounded. Run on every save.
pnpm lint; echo "lint=$?"
pnpm typecheck; echo "typecheck=$?"Deterministic proof
Unit and contract tests. Repeatable. Run before claiming done.
pnpm test; echo "test=$?"
cd services/api && go test ./...Real surface, real load
Drive the actual UI, API, browser. Slow, fragile, decisive. Only this catches mobile overflow and missing 404.
pnpm check:ui; echo "ui=$?"
TURBO_FORCE=true pnpm e2eFOUR WAYS GREEN IS FALSE1 2 3 4
Pipe a gate
tail, grep, or head hide the real exit code — a pipeline returns the last command's status.
Swallow a rejection
Silent .catch(() => {}) hides a panic, an upstream limit, or a partial failure.
Trust the cache
Turbo caches results. A gate that "passes" may not have run — use TURBO_FORCE=true.
Skip the third layer
Static + unit can both be green while the page breaks on mobile or the API never returns 404.
Path A · verification lab
Run the four-row comparison strip on the starter. Run A naively, Run B with $gate-discipline + $webapp-testing.
Toggle rule sources off and on. Compare the diff size, the count of gate commands invoked, and the names of checks cited in the final reply.
Open the rules lab →Keep learning12 new readings + primary docs
Go deeper with official documentation, production case studies, Medium, and practitioner workflows. Rules and enforcement case study → Primary references → 12-part reading path →