4 core prompts get you from "I have a modality pair" to "I have a paste-ready Lovable build + demo script." 3 path-specific prompts: pick the one that matches your starting point. 2 emergency prompts: only if you get stuck or your demo crashes. Save each output as `.md`. Context carries.
🏗 You've been building since Day 1 · already have a Day 7 MVP / landing page / users → run Prompt 05 first · then Prompts 01 → 04. 🆕 You just joined · no project yet → run Prompt 06 first · then Prompts 01 → 04. 🏆 Brand-new idea · just here to win → run Prompt 07 first · then Prompts 01 → 04. All paths converge at Prompt 01 with the same level of context. From there, everyone's on the same track.
01
Prompt 01 · 12:00–12:10
The Mashup Picker.
⏱ 10 min💻 Solo / pair📁 Save: mashup_angle.md
When to run: Right after team formation (~12:00). You've picked a modality pair. This prompt generates 3 specific mashup angles for that pair, scored, with a winner picked.
# ROLE
You are a 3-person AI product design panel:
• Aravind Srinivas (Perplexity) — what feels NEW about this combo
• Mira Murati (Thinking Machines) — what's technically WILD here
• Mike Krieger (Anthropic / ex-Instagram) — what's SHIPPABLE in 4 hours
# GOAL
Generate 3 specific mashup product angles for the modality pair I name. Score each. Pick the winner.
# CONTEXT
I'm in a 6-hour live hackathon at Network School in Forest City, Malaysia.
I just learned multi-modal AI yesterday (Day 11 of Build Sprint).
I have ~4 hours of actual build time after this prompt.
My modality pair: [e.g. "voice + image" or "video + agent"]
My existing Day 7 MVP idea (if I want to extend it): [1 sentence or "starting fresh"]
My team: [solo / pair with X]
My API budget today: [$5 / $20 / unlimited]# CONSTRAINTS
- Must work end-to-end (real APIs, not mocks)
- Must be demoable in 3 min
- Must combine the 2 modalities into ONE user moment (not 2 separate features)
- Should be the kind of demo where the room goes "wait, what?"
- AVOID: pure ChatGPT wrappers · clones of Suno / HeyGen / Cluely / etc.
# OUTPUT FORMAT
## Angle 1: [Name]
**Modalities wired:** [exact pair]
**The 1-line pitch:** [one sentence the user gets]
**The Aha moment:** [the 5 seconds that makes them say "wait"]
**Stack:** [STT/TTS/vision/video vendor + model]
**Est. cost per use:** [$]
**4-hour shipability:** [score 1-10]
**Wow factor:** [score 1-10]
**Risk:** [the one thing that might kill it today]
## Angle 2: [...]
## Angle 3: [...]
## 🏆 Panel pick
The angle I should build today: **[name]**
Why: [3 sentences — Aravind / Mira / Mike each weigh in]
The 1 thing to nail in build round 1: [...]
The 1 thing to cut if I slip past 14:30: [...]
# DELIVERABLE
Save this conversation. After you reply, I'll save the panel pick as `mashup_angle.md` in my hackathon folder. We'll reuse this context in Prompts 02, 03, 04.
# SELF-CHECK
Before you reply, confirm: (1) each angle uses BOTH modalities into ONE moment, not two; (2) each angle is shippable in 4 hours by a solo / pair builder; (3) the panel pick has a clear "wait, what?" moment.
💡 Tip: If your modality pair feels overdone (voice + text = ChatGPT clone risk), tell the panel to push toward angles that "feel like 2026, not 2023."
Deliverable: Save the panel's winning angle as mashup_angle.md. We feed this into Prompt 02 next.
02
Prompt 02 · 12:10–12:17
The 6-Hour Architecture.
⏱ 7 min💻 Solo / pair📁 Save: hackathon_plan.md
When to run: Right after Prompt 01. Same Claude chat — context carries. Turns your chosen angle into a survival-grade hour-by-hour build plan with slip indicators.
# ROLE
You are Linda Liukas (children-of-Code) crossed with John Carmack — you write build plans that even a tired hackathon builder at hour 4 can follow. Crystal clear · time-boxed · slip-aware.
# GOAL
Turn the panel-picked mashup angle (from previous message) into an hour-by-hour 6-hour build plan with slip indicators.
# CONTEXT
You already know my mashup angle from Prompt 01 above. Reuse it.
Now: I have 6 hours total today, but only 4 hours of actual build (rest = team forming / showing / demos).
Real build window: 12:30 – 16:00.
Demo prep: 16:00 – 16:15.
Live demos: 16:15.
My energy crash window: ~14:30 (right after lunch).
# CONSTRAINTS
- Plan must assume Lovable + Supabase as base
- Edge functions for any API Lovable can't call directly
- Must include "kill list" — what gets cut at 14:30 if I'm behind
- Must include a 15:30 cost check
- Must include a 15:45 demo script trigger
# OUTPUT FORMAT
## 6-Hour Build Plan: [mashup name from Prompt 01]
### 12:30 – 13:00 · Scaffold (30 min)
[exactly what to do · what success looks like at 13:00]
**Slip indicator:** [if X isn't done by 13:00, you're slipping]
### 13:00 – 13:30 · API wiring (30 min)
[...]
### 13:30 – 14:00 · First end-to-end pass (30 min)
[...]
### 14:00 – 14:30 · Iterate the Aha moment (30 min)
[...]
### 14:30 – 14:45 · MIDPOINT SHOW & TELL
[1 sentence — what you're showing the cohort]
### 14:45 – 15:15 · Polish round 1 (30 min)
[...]
### 15:15 – 15:35 · Stranger test (20 min)
[...]
### 15:35 – 15:50 · Friction fixes (15 min)
[...]
### 15:50 – 16:00 · Demo script (Prompt 04)
[...]
## 🔪 Kill List (if I'm slipping at 14:30)
1. CUT FIRST: [...]
2. CUT SECOND: [...]
3. CUT THIRD: [...]
**NEVER cut:** the 2-modality wire-up. That's the requirement.
## 💰 Cost check at 15:30
Run Day 11 Prompt 03 against: [the 3 most-called APIs in my stack]
If unit cost > $0.50/use → kill the most expensive API or batch it.
# DELIVERABLE
Save your reply as `hackathon_plan.md`. Pin it visible in my editor.
# SELF-CHECK
(1) Every block has a slip indicator. (2) Kill list has 3 specific items. (3) Plan ends with a script trigger, not "polish more."
💡 Tip: Don't argue with the kill list. If the plan says cut feature X, cut feature X. You'll add it back next week.
Deliverable: Save as hackathon_plan.md. Open it in a 2nd browser tab — glance at it every 30 min.
03
Prompt 03 · 12:17–12:30
The Lovable Mashup Build.
⏱ 13 min💻 Solo / pair📁 Save: lovable_prompt.md
When to run: Last prompt before build round 1. Generates a paste-ready Lovable prompt that wires your 2 modalities into a 3-screen MVP. Edge function pattern included.
# ROLE
You are a senior full-stack engineer who ships Lovable + Supabase apps every day. You know exactly what Lovable can do natively, what needs edge functions, and how to write a Lovable prompt that produces working code on the FIRST try (not the third).
# GOAL
Generate a paste-ready Lovable prompt for my mashup angle. Plus: edge function code for any API call Lovable can't make directly. Plus: a Supabase schema for the 1-2 tables I'll need.
# CONTEXT
You know my mashup angle (Prompt 01) and my 6-hour plan (Prompt 02) from the messages above. Reuse them.
My chosen vendors: [e.g. Claude Sonnet 4.6 / ElevenLabs v3 / Flux 1.2 Pro / Luma Ray 2 / Cartesia Sonic 2 / HeyGen / Suno v4]
My target users: [1 line — who tries this Saturday]
Auth: [email magic link · or none — anonymous demo]# CONSTRAINTS
- Lovable prompt must produce a 3-screen app: (1) input · (2) generation/loading · (3) output + share
- Edge function pattern: 1 file per external API
- Supabase: minimal schema — generations table at minimum, with cost_usd column
- RLS policies: lock generations to user_id (or session_id if anonymous)
- Output must include a working test API call I can paste into Supabase before scaffolding
# OUTPUT FORMAT
## Part A · Paste-Ready Lovable Prompt
```
[Full Lovable prompt I copy into lovable.dev — 1 block, 1 paste, ready to send]
```
## Part B · Supabase Schema
```sql
[CREATE TABLE statements + RLS policies]
```
## Part C · Edge Function (per external API)
```ts
[Deno-compatible edge function code — 1 per non-Lovable API]
```
## Part D · API Keys Checklist
- [ ] [Vendor 1] — sign up at [URL] · get key · paste into Supabase secrets as KEY_NAME_1
- [ ] [Vendor 2] — [...]
## Part E · First Working Call (test before scaffolding)
```bash
[curl command I run to verify the API key works before I waste 10 min on Lovable]
```
## Part F · 1 Thing That Will Probably Break
[The realistic gotcha + the fix]
# DELIVERABLE
Save the full reply as `lovable_prompt.md`. Then copy Part A into Lovable.dev. Then copy Part C into Supabase edge functions. Build round 1 starts at 12:30.
# SELF-CHECK
(1) Part A is ONE block, paste-ready, doesn't require me to edit. (2) Part E gives me a curl command I can run to verify keys work BEFORE I scaffold. (3) Part F flags the realistic gotcha (not "API might fail" — something specific to this mashup).
💡 Tip: Run Part E FIRST. If the test curl doesn't return 200, fix that before touching Lovable. Don't scaffold an app on a broken key.
Deliverable: Save as lovable_prompt.md. Paste Part A into Lovable at 12:30. Build starts.
04
Prompt 04 · 15:50–16:00
The 3-Min Demo Script.
⏱ 10 min💻 Solo / pair📁 Save: demo_script.md
When to run: Right before live demos. Hook · live use · why it matters · CTA. Same Claude chat — context still carries.
# ROLE
You are Jensen Huang giving a 3-minute keynote. You know how to hook a tired audience, show a live product moment, and close with a memorable line. You don't waste a single second.
# GOAL
Write me a 3-minute demo script for the mashup I just built today. Hook · live use · why it matters · CTA. Reads like a story.
# CONTEXT
You know my mashup angle (Prompt 01), my build plan (Prompt 02), and my actual product (Prompt 03) from the messages above.
What's WORKING in my live demo: [1-2 sentences of honest assessment]
What's NOT working / a known glitch: [1 sentence — what I'll dodge live]
My live URL: [lovable URL]
Audience: 20 other builders + Megan + judges. They know multi-modal. They've seen yesterday's MVPs. They want to see something new.
Demo room: NS Library · projector · my laptop · my phone for live phone demo.
# CONSTRAINTS
- Total time: 2:45 – 3:00 (judges have a hard 3-min cap)
- Must include 1 LIVE input (not a screenshot)
- Must end with a specific 1-line CTA (URL or QR code)
- No reading slides. Conversational tone.
- Plant a "wait, what?" moment in seconds 30-60
# OUTPUT FORMAT
## 3-Minute Demo Script: [mashup name]
### 0:00 – 0:30 · Hook
[exact words I say · including the 1 line that makes the room lean in]
### 0:30 – 2:00 · Live use
**What I do on screen:** [step 1 · step 2 · step 3]
**What I say while doing it:** [exact words · including the "wait, what?" line]
**What the output is:** [...]
**If it breaks:** [fallback — describe what would have happened]
### 2:00 – 2:30 · Why it matters
[the line that justifies why combining these 2 modalities matters]
[the specific user · the specific moment · the before-vs-after]
### 2:30 – 3:00 · CTA
[the URL · the QR ask · the 1 thing I want from the audience by Sunday]
## 📋 Pre-Demo Checklist (1 min before I'm called)
- [ ] Live URL open in 1 tab · zoom 125%
- [ ] Phone open to same URL · screen-mirrored if possible
- [ ] [item 3 specific to my demo]
- [ ] Deep breath. I shipped a mashup in 4 hours. Already winning.
## 🎯 The 1 Line I Should Memorize
"[The single sentence the audience walks away repeating]"
# DELIVERABLE
Save as `demo_script.md`. Read it aloud once. Time it. Adjust 1 line. Stop fiddling at 16:10.
# SELF-CHECK
(1) Script times to 2:45-3:00 when read at normal pace. (2) "Wait, what?" moment lands in seconds 30-60. (3) CTA is ONE specific ask, not "check it out."
💡 Tip: Practice the demo standing up. The energy is completely different from sitting. Judges are sitting — you should not be.
Deliverable: Save as demo_script.md. Practice once aloud. Stop refining at 16:10. Go demo.
🎯 PATH-SPECIFIC PROMPTS
Pick ONE. Run it before Prompt 01.
These three prompts each get you to the same baseline of context — so the 4 core prompts work for everyone, no matter where you started.
05
Prompt 05 · Continue YOUR project
The 10-Star Mashup Add-On.
⏱ 12 min💻 Solo / pair🏗 Have an existing project📁 Save: mashup_addon.md
When to run: Before Prompt 01, IF you already have a Day 7 MVP / landing page / users from earlier in the cohort. This prompt extends your existing project with the 2 modalities that would make TODAY's demo a 10-star "wait, what?" moment — while staying shippable in 4 hours.
# ROLE
You are a 2-person panel of senior consumer-AI product builders:
• A product designer who shipped Notion AI's most-loved feature
• An AI architect who shipped multi-modal at OpenAI / Anthropic / Google
You've both judged 50+ hackathons. You know what makes a demo unforgettable.
# GOAL
Take MY existing project (described below) and recommend the 2 modality add-ons that would make today's hackathon demo a 10-star moment — with maximum "wait, what?" / Aha factor — while still shippable in 4 hours by a solo / pair builder.
# CONTEXT — MY EXISTING PROJECT
Project name: [name]
What it does (1 sentence): [the core user moment today]
Who uses it: [1 sentence on ICP]
Current modalities (what's wired NOW): [text / image / voice / video / etc — be specific]
Where it lives: [live URL / repo / "just landing page" / etc]
Stack: [Lovable / Supabase / which APIs already wired]
The 1 user complaint or friction I keep hearing: [Mom Test quote if you have one]
The 1 user delight I keep hearing: [what's working]
Today's deadline: 4 hours of actual build time (12:30-16:00).
# CONSTRAINTS
- Add-ons must EXTEND the existing project, not replace it
- Must combine ≥ 2 NEW modalities with what's already there
- Must produce ONE seamless user moment (not a bolted-on side feature)
- Must be demoable in 3 min
- Must be shippable in 4 hours (count edge function setup + API key signup time)
- The "wait, what?" moment must hit in the FIRST 30 seconds of the demo
# OUTPUT FORMAT
## 🎯 Diagnosis
The 1 thing your current project is MISSING that's keeping it at 7-star (the gap a mashup could close): [...]
## 🔥 Top 3 Add-On Mashups
### Mashup A: [name]
**Modalities added:** [exact pair · be specific about vendor + model]
**The new user moment:** [the single sentence describing what the user now experiences]
**The "wait, what?" hook (first 30 sec of demo):** [exact moment that makes the room stop]
**Why it's 10-star vs your current 7-star:** [the specific UX leap]
**Risk / hardest part to ship in 4 hrs:** [the realistic gotcha]
**Cost per use (rough):** $[X]
**Shipability score (1-10):** [score]
**Wow factor score (1-10):** [score]
### Mashup B: [...]
### Mashup C: [...]
## 🏆 Panel Pick
The mashup add-on you should build today: **[name]**
**Reasoning (2-3 sentences):** [why this one wins both the "wow" judging criteria AND survives the 4-hour clock]
**The 1 thing to nail in build round 1:** [...]
**The 1 thing to cut if you slip past 14:30:** [...]
## 🔗 Continuity bridge
Now that the panel has picked Mashup [X]: paste this line into Prompt 01 in place of the "modality pair" field:
> "[exact line to paste — including modality names + the existing-project tie-in]"
# DELIVERABLE
Save your full reply as `mashup_addon.md`. Then run Prompt 01 — use the "continuity bridge" line above as your modality pair input. The panel context carries forward.
# SELF-CHECK
(1) Each add-on EXTENDS the existing project, never replaces it. (2) The "wait, what?" moment is described in seconds 0-30 of the demo, not buried in the middle. (3) The panel pick has a clear cost-shipability-wow tradeoff articulated. (4) The continuity bridge line is paste-ready for Prompt 01.
💡 Tip: The judges will reward NOVELTY vs your Day 7 build (Crit 07 in the rubric). Don't just bolt voice on top of text — weave it INTO the moment users already love.
Deliverable: Save as mashup_addon.md. Use the continuity bridge line to start Prompt 01.
06
Prompt 06 · I'm new to the program
The Context Gatherer.
⏱ 15 min💻 Solo / pair🆕 No project yet📁 Save: my_project_context.md
When to run: Before Prompt 01, IF you're new to Build Sprint and missed the first 11 days. This prompt extracts the MVP / ICP / JTBD context you'd normally have from cohort work — in 15 min instead of 11 days. Then you join Prompt 01 on equal footing.
# ROLE
You are a Build Sprint mentor who has onboarded dozens of builders. You're warm but efficient. You know exactly which questions extract the most signal in the least time. You think like a YC partner doing a 15-min office hour.
# GOAL
Help me — a newcomer to today's hackathon — gather all the context I need to build a winning multi-modal mashup. You'll Q&A me through 7 questions, then synthesize my answers into a project context doc that mirrors what cohort builders have from Days 1-11.
# CONTEXT — ABOUT ME
What I do for work / school: [1 sentence]
A problem I've actually had this month (annoying / painful / wish-it-was-better): [1-2 sentences — be specific]
A product I currently use and love (and what specifically I love): [the role model]
My technical comfort level: [never coded / vibe coder / engineer]
Time I have today: ~4 hours of actual build (12:30-16:00).
What "winning" looks like to me: [ship something I'm proud of / win the prize / make a friend / learn 1 new thing]# YOUR PROCESS
Run me through these 7 questions one at a time. After each answer, ask a sharper follow-up if my answer is fuzzy. Don't move on until each is concrete.
1. **The user.** Who specifically would use what we build today? (Not "everyone." A specific person — friend, colleague, archetype.)
2. **The moment.** What's the specific 5-second moment in their day this would help with?
3. **The job-to-be-done.** What are they "hiring" this tool to do for them at that moment?
4. **Their alternative today.** What do they do RIGHT NOW to handle this? (Even if it's annoying.)
5. **The Aha.** What would make them say "wait, that's amazing"? Be concrete.
6. **The constraint.** What WON'T they do? (e.g. they won't watch a 2-min tutorial · they won't pay $10/mo on day 1)
7. **The Mom Test moment.** Have you actually heard ANYONE say something like "I wish there was a tool that ___"? If yes, what were the exact words?
# THEN SYNTHESIZE — OUTPUT FORMAT
After I answer all 7, output the following structured doc:
## My Hackathon Project Context
### ICP (Ideal Customer)
[1 paragraph — specific person, not a demographic blob]
### Core Job-To-Be-Done
"When [trigger moment], I want to [job], so I can [outcome]."
### The current alternative
[What they do today — 1 sentence]
### The Aha I'm chasing
[The exact 5-second moment that flips them from "neat" to "WOW"]
### Mom Test quote (if any)
"[The actual user words]"
### 2 modality pairs that might fit (gut check)
- Option A: [pair] · why it might fit [...]
- Option B: [pair] · why it might fit [...]
### Hard constraints today
- Tech comfort: [...]
- Time: 4 hours build
- Budget: [API costs ok? · prefer free tier?]
### 🔗 Continuity bridge
Now that you have my context, paste this line into Prompt 01 in place of the "modality pair" field:
> "[exact line — modality pair from gut check + the user + the moment]"
# DELIVERABLE
Save as `my_project_context.md`. Then run Prompt 01 — paste the continuity bridge line as your modality pair. Welcome to the cohort.
# SELF-CHECK
(1) You asked all 7 questions and pushed back on fuzzy answers. (2) The final doc reads as if I'd been building for a week. (3) The continuity bridge is paste-ready for Prompt 01. (4) You were warm — this is a stranger's first hour with us.
💡 Tip: Don't speedrun the 7 questions. The Mom Test moment (Q7) is the highest-leverage signal — if you've actually heard someone complain about this, you've already won half the rubric.
Deliverable: Save as my_project_context.md. Use the continuity bridge to enter Prompt 01.
07
Prompt 07 · I'm here to win
The Win Strategy.
⏱ 12 min💻 Solo / pair🏆 Novel idea · optimize for win📁 Save: win_strategy.md
When to run: Before Prompt 01, IF you have a totally new idea AND you want to optimize this build for taking home a prize. This prompt reverse-engineers the 8-criterion judging rubric and matches it to a modality combo that maximizes wow-per-hour-of-build.
# ROLE
You are a hackathon coach who has judged 100+ AI hackathons (YC, Anthropic builders, HF1, NS, indie). You have a near-religious read on what makes judges score 5/5 on every criterion. You're brutally honest. You don't sugarcoat.
You think in expected value: which modality combo · which user moment · which demo opening has the highest probability of taking a top-3 spot today.
# GOAL
Given my raw idea (below), generate the 3 highest-EV modality mashups that maximize my judging score across all 8 rubric criteria, ranked. Then pick the winner and tell me exactly how to demo it for max impact.
# CONTEXT — MY RAW IDEA
The germ of an idea I want to build today (1-3 sentences, be raw): [your idea — even if half-baked]
Why this idea (1 line — what's the pull?): [...]
My skill level: [I can ship / I need Lovable to do most of it]
My team: [solo / pair]
What I'd consider a "successful day": [ship + learn / top 3 / Best in Show — be honest]# THE JUDGING RUBRIC (you must optimize against this)
8 criteria · 5 pts each · 40 max:
1. **Mashup Integrity** — 2+ modalities woven into ONE moment
2. **"Wait, what?" factor** — surprise, craft, originality
3. **Shipped & Live** — real APIs, real URL, cohort can try it
4. **Real Job-To-Be-Done** — solves a real moment for a real user
5. **Demo Craft** — hook · live use · CTA · timing tight
6. **Cost Discipline** — sane unit cost · 1K-user math defensible
7. **Novelty vs Day 7 MVP** — new angle (or "didn't exist before today")
8. **Stage Presence** — energy · clarity · confident close
Plus 2 special awards: ⚡ Best Mashup (creative combo) · 💜 Cohort Choice (peer vote)
# CONSTRAINTS
- Must be shippable in 4 hours (12:30-16:00)
- Must be a TRUE mashup (2+ modalities, ONE moment)
- Must avoid the "ChatGPT wrapper" trap
- Must have a memorable single-line pitch
- Must have a "wait, what?" moment in seconds 0-30 of the 3-min demo
# OUTPUT FORMAT
## 🎯 Diagnosis of my idea
What's working in my germ-of-an-idea: [...]
What's risky (low rubric score): [...]
The 1 reframe that could move this from "fine" to "memorable": [...]
## 🏆 Top 3 Modality Mashups · scored against rubric
### Mashup A: [name]
**Modalities:** [exact pair + vendor/model]
**One-line pitch (this is what you say in second 0):** "[...]"
**The "wait, what?" hook (seconds 0-30 of demo):** [exact beat]
**Stack:** [Lovable + Supabase + which APIs]
**Cost per use:** $[X]
**Rubric projection (40 max):**
- Mashup Integrity: [score] · [why]
- "Wait, what?": [score] · [why]
- Shipped & Live: [score]
- Real JTBD: [score]
- Demo Craft: [score]
- Cost Discipline: [score]
- Novelty: [score]
- Stage Presence: [score]
**Projected total: [X] / 40**
**Risk: [the realistic killer in 4 hrs]**
### Mashup B: [...]
### Mashup C: [...]
## 🏆 The Pick
Build **[Mashup name]**. Here's why this beats the other two: [3 sentences]
## 🎬 The Memorable Demo
**The single line you say in second 0 (memorize this):** "[...]"
**The "wait, what?" moment (seconds 15-30):** [exact action + audience reaction you're hunting]
**The line that wins the room (seconds 2:30-2:50):** "[...]"
**The CTA (seconds 2:50-3:00):** "[...]"
## 🎁 Bonus · special-award strategy
**To win ⚡ Best Mashup:** [the 1 thing to push harder]
**To win 💜 Cohort Choice:** [the 1 thing that makes peers love it]
## 🔗 Continuity bridge
Paste this line into Prompt 01 in place of the "modality pair" field:
> "[exact paste-ready line]"
# DELIVERABLE
Save as `win_strategy.md`. Reread the "Memorable Demo" section every 90 min during build. The 4-line script is your north star.
# SELF-CHECK
(1) Top 3 are scored against ALL 8 criteria — no skipped rows. (2) The pick has a defensible "why this beats the others." (3) The memorable demo script has 4 specific lines — not vague directions. (4) Bonus award strategy gives 2 concrete pushes for ⚡ and 💜.
💡 Tip: Memorize the 4 demo lines. Don't read them. The Best in Show winners always feel rehearsed but spontaneous — that comes from knowing the 4 anchors cold.
Deliverable: Save as win_strategy.md. Use the continuity bridge for Prompt 01. Then crush the 4 core prompts and ship.
🚨 EMERGENCY PROMPTS
Use only if needed.
Hour 3 panic mode. Demo crashed. Stuck on the same bug for 25 min. These prompts get you unblocked fast.
🚨
Prompt 08 · "I'm stuck"
The Unstick Me.
⏱ 5 min · panic mode💻 Same Claude chat📁 Save: unstick.md
When to run: ANY time you've been blocked on the same problem for more than 20 min · OR you can feel panic rising · OR you're tempted to start over. This prompt does triage + gives you 3 unblocking paths in 90 seconds.
# ROLE
You are John Carmack at hour 18 of a hackathon, sitting next to me. Calm · methodical · unbothered. You don't catastrophize. You triage. You've shipped under harder pressure than this.
# GOAL
Get me unstuck in <5 minutes. Three concrete paths · ranked. I pick one and execute.
# CONTEXT
You already know my mashup (from earlier prompts).
NOW the situation:
What I'm trying to do: [1 sentence — the immediate goal of this 30-min block]
What's happening instead: [1-2 sentences — the actual error / stuck state]
What I've tried so far: [bullet list — be honest]
Time remaining today: [hours left until 16:00 build cutoff]
My energy level (1-10): [honest number]# CONSTRAINTS
- Solutions must be doable in <30 min
- "Start over" is OFF the table unless I'm at hour 1
- "Add a complex new feature" is OFF the table
- Focus on: what's the SMALLEST cut that ships?
# OUTPUT FORMAT
## 🩺 Triage
In 1 sentence: what's actually broken vs what I think is broken: [...]
## 🔀 3 unstick paths · ranked
### Path A · "Bypass the bug" (5 min)
The brute force workaround. Sacrifice elegance · keep the demo alive.
[Specific action]
Tradeoff: [what I give up]
### Path B · "Cut the feature" (10 min)
Remove the broken piece · ship the rest.
[Specific action — what to delete · what remains demoable]
Tradeoff: [what the demo loses]
### Path C · "Fix it surgically" (20 min)
The proper fix — only if I have 20+ min AND high energy.
[Specific action]
Tradeoff: [risk of going deeper than budget]
## 🎯 If I had to choose for you
Based on your time left + energy: **[Path X]**
Why: [1 sentence]
Start with this exact step: [the very next thing to type / click]
## 🛟 If even Path A doesn't work in 10 min
The "demoable but partial" backup:
- Show me a demo of [the piece that DOES work]
- Verbally explain [the piece that doesn't]
- Use Prompt 09 (Demo Recovery) when it's time
# DELIVERABLE
Save as `unstick.md`. Execute Path X (or the path I pick). Set a 30-min timer. If timer hits and I'm still stuck — pick a less ambitious path. Don't sink.
# SELF-CHECK
(1) All 3 paths are <30 min. (2) Path A truly bypasses (not "really fixes" in disguise). (3) The "next exact step" is specific — not "debug it."
💡 The hard rule: If you've been stuck on the same thing for 30+ min total, you go Path B (cut the feature) — not Path C. The cohort respects shipped mashups, not heroic debugging.
Deliverable: Save as unstick.md. Pick a path. Set a timer. Ship.
🆘
Prompt 09 · "Demo broke"
The Demo Recovery.
⏱ 90 sec · on stage📱 On phone · whispered📁 Save: demo_recovery.md (read before demo)
When to run: Run this BEFORE demos start (around 16:00) so you have the recovery script ready. THEN if your demo crashes on stage, you've already rehearsed the save. No panic. You look like a pro.
# ROLE
You are a startup founder who has demoed 50+ live products on stage at YC, TechCrunch Disrupt, AWS Summit. You know that demos break in PUBLIC, and the best founders are graceful when they do. You teach me how to handle it before it happens.
# GOAL
Generate my "demo recovery toolkit" — exact words I say if X happens. 4 scenarios. I memorize the responses.
# CONTEXT
You know my mashup, my demo script (Prompt 04), and what's likely fragile.
Specifically, the parts of my demo most likely to break: [1-2 sentences from honest self-assessment]
My live URL: [lovable URL]
My demo's most expensive API call (cost-wise + latency-wise): [the one that might time out]
My phone backup plan: [do you have screenshots of working state? a video?]# CONSTRAINTS
- Each response must be ≤ 2 sentences
- Tone must be confident · NEVER apologetic / self-deprecating
- Recovery must keep moving the demo forward — never freeze
- Audience must end the moment NOT remembering the crash, only remembering you handled it
# OUTPUT FORMAT
## 🆘 Demo Recovery Toolkit
### Scenario 1: API call times out / returns nothing
**What you say (confident · 1 line):** "[...]"
**What you do next:** [show screenshot · narrate the expected output · move on]
**Time to recover:** ≤ 15 sec
### Scenario 2: Wifi dies mid-demo
**What you say:** "[...]"
**What you do next:** [phone hotspot / pre-recorded backup video / verbal walkthrough]
### Scenario 3: Output is embarrassingly wrong (image off · voice gibberish · wrong language)
**What you say:** "[...]"
**What you do next:** [acknowledge briefly · pivot to the cohesive moment · don't dwell]
### Scenario 4: Total black screen / app crashed
**What you say:** "[...]"
**What you do next:** [refresh · while refresh loads, deliver the "why this matters" part of script · come back to demo]
## 🧠 Universal recovery lines (memorize 2)
For ANY crash:
1. "[the confident bridge line that any pro would say]"
2. "[the redirect line that pulls the audience back to YOUR story]"
## 🎬 Pre-demo backup checklist
- [ ] Screenshots of working demo (3 key moments) — opened in tab #2
- [ ] Phone hotspot tested · password memorized
- [ ] 30-sec video of demo running (record on phone earlier) — saved to camera roll
- [ ] Demo script (Prompt 04) open in tab #3 in case I blank
- [ ] Water bottle on stage (gives me 3-sec pause if I need to think)
## 💡 The mindset
"I'm not demoing a perfect product. I'm demoing my craft as a builder. If something breaks, I show grace. Judges score grace."
# DELIVERABLE
Save as `demo_recovery.md`. Read 3x before going on stage. Memorize the 2 universal lines.
# SELF-CHECK
(1) NO line is apologetic. (2) Every scenario keeps the demo moving. (3) Universal lines are short enough to use under panic.
💡 The pro move: Take a screenshot of your demo's 3 best moments at 15:50. Open them in browser tabs. If anything breaks live, you flip to a tab and say "that's what we just saw." Audience can't tell.
Deliverable: Save as demo_recovery.md. Run at 15:55 — right before demos start.