Flash Lab — Build Section: Plan → Prompt → Prototype

Status: model updated 2026-06-04 to match the slide (prep/flash-lab-build-section.html, commit e1cf9a0). Diagram source is the .html; the .png export is from the older “Problem → Plan → Prototype” model and needs re-exporting.

---

The flow (vertical)

Design hands off a PROBLEM (a real need + who it’s for). The Build section is three steps from there — the three P’s:

1. PLAN — name the bet in two blanks: “Our prototype lets them ___ (ACTION) so we find out if ___ (INSIGHT).” Action = the one thing it does; Insight = what you’re still unsure about. ★ boost — human or AI.
   • e.g. lets them snap 5 exit tickets & get a “who’s lost” summary · find out if a quick AI read flags the right kids
2. PROMPT — assemble one instruction from the Plan. Who + action drop down from Plan; you add the type of thing, the magic moment, and what to fake: “Build a ___ that lets ___ do ___. It should ___ back.” Paste it into a tool, or use it as your brief / act it out. ★ boost — human or AI.
   • e.g. Build a one-page web app that lets a teacher upload 5 exit-ticket photos · It should show a short “who’s lost” summary
3. PROTOTYPE — a prop that tests the idea, not a finished product. ★ boost — build the prop one of two ways (both live inside Prototype now, not as an upstream fork):
   • If low-fi → human boost: use the prompt as a brief, act it out / sketch it — you’re the engine. Materials: storyboard · sticky notes · paper · role-play. A first-class outcome, not the consolation prize.
   • If digital → AI boost: paste it into a tool, an AI builds it — pick your altitude on the vibecoding spectrum (below).

The thinking lives in Plan + Prompt (what to prove, then how to ask for it). Prototype is then heads-down doing.

---

Boosts: human or AI, on every step

A boost cue rides every step (Plan, Prompt, Prototype) — 🧑 human · 🤖 AI.

• At Plan, either can pressure-test “what must this prove?” — a facilitator move or a conversational AI.
• At Prompt, either can help draft the instruction.
• At Prototype, the digital boost is AI (the build tools); low-fi leans human, but conversational AI can still generate low-fi content (storyboard beats, role-play scripts, fake data).

So “AI boost” isn’t the only boost — name the human one too.

---

The digital spectrum — 2 modes (shown only on the digital path)

The two modes are the column names — chat-to-make · prompt-to-app. Teach those (not the tools); logos are swappable examples so the framework survives tool churn. (Code-with-AI dropped 2026-06-05 — too much for a 2–3 hour workshop; people who already use those tools will reach for them anyway.)

Zone	Verb	What you get	Examples	Floor / Ceiling
Chat-to-make	chat → artifact	a thing in the chat	Gemini, ChatGPT, Claude artifacts	low floor / low ceiling
Prompt-to-app	describe → real app	a shareable app, no code	Google AI Studio, Base44, Lovable.dev	low floor / higher ceiling

---

Example outputs (gallery on the slide)

What a prototype can look like, across both paths:

1. ‘Vibe-coded’ digital prototype — iterative design
2. Paper storyboard — hand-drawn sketches
3. AI-generated mockup — high-fidelity design
4. Role-play script — user–AI dialogue
5. Interactive flowchart — post-it decision tree
6. Video walkthrough — team demonstration

---

Design principles

• The thinking (Plan + Prompt) precedes the building; that’s where the boost matters most.
• Boost = human or AI — don’t collapse it to “AI boost.”
• Low-fi is a first-class outcome, not the fallback. The spectrum only appears on the digital path.
• Teach the two modes (chat-to-make · prompt-to-app); treat logos as examples (tools churn fast).
• Build-a-Bot is NOT on this spectrum — separate offering, not a vibecoding tool.

Open questions (to decide)

1. Default path: low-fi-leaning, or digital-forward with low-fi named as legit?
2. Hands-on: everyone prototypes in the middle zone, or self-select a zone?
3. Right side (code-with-AI): awareness-only, or a live demo moment? — Resolved 2026-06-05: cut code-with-AI from the workshop entirely.