Rapid Iteration Methodology
Purpose: keep Forge cycles short, demoable, and decision-oriented.
Contents
- Stage timing
- Speed tactics
- Progressive build pattern
- Feedback cadence
- Pivot rules
Stage Timing
| Prototype type | SCAFFOLD |
STRIKE |
COOL |
PRESENT |
Total |
|---|---|---|---|---|---|
| Quick check | 30 min |
1-2h |
30 min |
30 min |
2-4h |
| UI component | 1h |
4-6h |
1h |
1h |
7-9h |
| Page / flow | 2h |
8-12h |
2h |
1h |
1-2 days |
| Full-stack PoC | 4h |
12-16h |
4h |
2h |
2-3 days |
Speed Tactics
| Tactic | Why it helps | Use when |
|---|---|---|
| Hardcode first | Removes API waiting time | Data-dependent UI |
| Write Tailwind utilities inline | Avoids context switching | Fast UI work |
| Single-file first | Minimizes navigation | Quick checks |
| Copy-paste is allowed | Speed beats DRY in prototypes | Throwaway work |
| Reuse existing UI primitives | Skip styling from zero | Component prototypes |
| Fixed-seed Faker data | Stable demos and screenshots | Repeated demos |
Template MSW handlers |
Instant CRUD-like mocks | API-dependent UI |
| Build one vertical slice | Keeps each iteration demoable | Pages and flows |
| Use AI for boilerplate | Speeds structure, not judgment | Repetitive scaffolding |
Progressive Build Pattern
Build in this order:
- Static layout with hardcoded data
- Basic interaction with state
- Mock data flow with inline mocks or
MSW - Loading / error / empty states
Run a COOL check after each step:
- Does it compile?
- Does it render?
- Does the main interaction work?
- Is the concept obvious to another person?
Feedback Cadence
| Feedback loop | Frequency | Goal |
|---|---|---|
| Self-check | Every 30 minutes |
Catch breakage early |
| Pair check | Every 2-4 hours |
Direction check |
| Stakeholder demo | PRESENT phase |
Force ADOPT / ITERATE / DISCARD |
PRESENT structure:
- Hypothesis in one sentence
- Demo the happy path
- Show 1-2 meaningful variations
- State what was learned
- Ask for
ADOPT / ITERATE / DISCARD
MVP Slicing Rules
Prefer vertical slices:
- Good: one user-visible slice that can be demoed end-to-end
- Bad: all UI first, all logic later, all APIs last
Include:
- Core interaction
- Minimum context to understand the idea
- One happy path
Exclude unless the hypothesis depends on them:
- Auth and permissions
- Full accessibility work
- Performance optimization
- Animation polish
- Full responsive coverage
Pivot Rules
Iterate when:
- Direction looks right but needs refinement
- New evidence suggests a narrower next test
- Stakeholders request specific, bounded follow-up
Discard when:
- The hypothesis is clearly disproven
- The approach is technically infeasible for the intended path
- Business priority changed
- The prototype exceeds the time-box twice without clearer learning
Forge Integration Rules
- Set the time-box before coding.
- Use progressive build order by default.
- Require a
COOLcheck at least every30 minutes. - Force scope review after
3iterations.