Author: Léo Date: 2026-02-24 Status: Draft v1
The Vibe Coding Platform lets anyone turn a text prompt into a playable, visually impressive mobile game in under 30 seconds — and publish it to the iOS App Store without writing a single line of code.
The core insight: Every competitor generates game code from scratch via LLM, producing broken, unplayable output 80%+ of the time. This platform inverts the model: pre-built, playtested Phaser.js game backbones provide a proven quality floor; AI generates only the themed assets. The result is a game that always works, always looks good, on the first try.
Users: Two segments share the same platform. Casual creators (Maya) generate and share games for fun — they're the viral engine. Serious publishers (Jake) iterate and export to the App Store — they're the revenue. Both follow a progressive ownership arc: Preview → Share link → GitHub → App Store, with each step a natural conversion moment.
Business model: Free to generate and share. Pay to own — watermark removal, GitHub export, iOS App Store export. All payments via web (Stripe); no Apple IAP.
Platform: Next.js web app for creation (desktop), React Native/Expo iOS companion app for on-device testing (QR scan → play), Phaser.js + Capacitor for the generated game exports (.ipa files).
12-month goal: $30K+ MRR or 50K+ MAU → fundraise or acquisition.
3-month targets (validation): - 1,000+ games generated/week - >15% of sessions result in a shared link - First paying exports — any revenue validates willingness to pay - At least one piece of organic social content showing the graphics quality goes viral
12-month targets (fundraise/acquisition readiness): - $30K+ MRR or 50K+ MAU with >20% week-over-week growth for 8+ consecutive weeks - Free → paid export conversion: ≥5% - Cohort retention: users who export come back and export again
| Metric | 3-Month Target | 12-Month Target |
|---|---|---|
| Games generated | 3,000+/month | 50,000+/month |
| Share rate | >15% | >20% |
| Free → paid conversion | — (learning) | ≥5% |
| MRR | First $ | $30K+ |
| Returning users (7-day) | >25% | >35% |
Persona: Maya, 24, content creator, 80K TikTok followers. Posts about Japanese street culture. Not a developer, has never shipped an app.
Opening Scene: Maya watches a competitor's AI game tool on Twitter. The output looks terrible — pixelated, misaligned, ugly. She comments "lol this looks broken." Someone replies with a link to Vibe Coding Platform. She clicks it from her phone at 11pm.
Rising Action: She types: "endless runner through a neon Tokyo street, anime style." She hits Generate. While waiting (~20s), a progress indicator shows assets being created by name. The game appears. It looks good. She scans the QR code and plays it on her iPhone. Her character is an anime girl running through a glowing Shibuya crossing.
Climax: She copies the share link and posts it to her TikTok story with "I made a game in 30 seconds." 400 responses in an hour. She didn't pay anything. She didn't sign up.
Resolution: She comes back two days later. Makes five more games. On the third session she notices the "Export to App Store" button.
Requirements revealed: Zero-friction creation (no signup to generate), share link works on iOS Safari, QR code on desktop, visual quality clears the "screenshot-worthy" bar on first try, prompt handles style keywords naturally.
Persona: Jake, 32, indie entrepreneur, day job in product. Has had a game idea for 18 months. Has tried Unity twice. Quit both times.
Opening Scene: Jake finds the platform from a Product Hunt launch. Skeptical — burned by "no-code game tools" before. Tries it expecting to be disappointed.
Rising Action (iteration path): First result is close but not quite right. He edits the prompt, adds art style detail. Third iteration: happy. Plays it for ten minutes. Enables meta-modules — adds a lives system and daily rewards. Shares with friends, gets feedback. Tweaks difficulty.
Rising Action (export path): Two weeks later, clicks Export → iOS. Enters Apple Developer credentials. Platform generates app icon, screenshots, App Store metadata. He reviews, approves, submits. Apple approves in 48 hours. It's live.
Climax: He opens the App Store on his phone and sees his game. His name in the developer field.
Resolution: $340 in first month from IAP. Starts building a second game. Upgrades to Pro subscription.
Requirements revealed: Prompt editing + regeneration without losing meta-module config, difficulty controls, meta-module UI, iOS export with credential input, AI-generated App Store assets, App Store submission flow, Pro subscription tier.
Persona: Sam, 27, frontend developer. Knows React, TypeScript. Never touched Phaser.js or Swift. Wants to ship a mobile game without learning a new framework.
Opening Scene: Finds the platform via a dev Twitter thread. Not interested in a "no-code toy" — wants real output she can own and modify.
Rising Action: Generates a Match-3 game. Wants to change the scoring system. Clicks "Export to GitHub." Authorises GitHub. A repo appears: clean Phaser source, assets, Capacitor config, README with build instructions.
Climax: Opens in VS Code. Changes two values in config.js. Runs npx cap run ios. It builds. Opens in Simulator.
Resolution: Ships a modified version that feels distinctly hers. Saved 3-4 days of setup. Tells other devs about it.
Requirements revealed: GitHub OAuth export (not zip), clean readable backbone code, documented config values, Capacitor config included, README with local build instructions.
Persona: Léo (and eventually a small ops team). Keeps platform healthy, quality high, liability low.
Key tasks: - Quality monitoring: Flag games where asset generation failed or produced low-quality output. Review and trigger regeneration. - IP violation review: Filter flagged prompts with trademarked terms. Review borderline cases, delete violations. - Backbone management: Add new backbone variants to staging, generate test games, approve for production. - Cost monitoring: Generation cost per game, CDN egress, AI API cost trends.
Requirements revealed: Ops dashboard (generation stats, cost tracking, quality flags), IP filter review queue, manual game deletion, backbone registry management.
| Capability Area | Driven By |
|---|---|
| Zero-friction creation (no signup to generate) | Maya |
| Share link + QR code for iOS Safari | Maya |
| Prompt editing + multi-iteration without losing state | Jake |
| Meta-module UI (lives, daily reward, currency) | Jake |
| iOS export with App Store submission flow | Jake |
| GitHub OAuth export with clean, buildable code | Sam |
| Pro subscription tier | Jake + Sam |
| Ops dashboard (quality, cost, IP flags, backbone mgmt) | Internal Ops |
1. Backbone-Augmented Generation (structural quality innovation) Every competitor (Rosebud, Gambo, Chaotix) generates game code from scratch via LLM — producing broken, unplayable output 80%+ of the time. This platform inverts the model: pre-built, playtested Phaser.js backbones provide a proven quality floor; AI generates only the themed assets. Same insight that made Canva beat early AI design tools: templates + AI > pure AI generation.
2. Mobile-Native Output (distribution innovation)
Every AI game tool outputs web or desktop. Nobody produces .ipa files. The gap from "playable in browser" to "on the App Store under your name" is the monetisation unlock nobody else fills.
3. The Creator Ownership Arc (UX innovation) Progressive ownership: Preview → Share link → GitHub → App Store. The game starts as a toy (free, no account) and becomes a real product (your repo, your App Store listing) as the user's intent escalates. Each step is a natural conversion moment — a new category of creator journey.
| What exists | What's missing |
|---|---|
| AI game generators (Rosebud, Gambo) — web only, low quality | High-quality output on first try |
| Unity AI (beta) — professional devs only | Zero-code path to App Store |
| No-code game builders (GameByte) — guided but slow | Speed + quality simultaneously |
| Vibe coding tools (Lovable, Bolt) — web apps, not games | Same paradigm applied to mobile games |
AI game generation market: $1.94B in 2026 → $32.48B by 2035 (31% CAGR). First mover with iOS-native output.
| Surface | Tech | Phase |
|---|---|---|
| Creation UI | Next.js web (desktop) | Phase 1 |
| iOS companion app | React Native / Expo | Phase 1 — QR scan → game in WebView → play on device |
| Game preview (web) | Phaser.js in iframe | Phase 1 — play in browser without installing |
| Game exports (.ipa) | Phaser.js + Capacitor → EAS Build | Phase 3 |
| Full iOS creation app | React Native / Expo | Phase 4+ (only if data shows mobile creation demand) |
Key decisions: - iOS only for game exports — stronger monetisation (~2x Android), simpler pipeline. Android not in roadmap unless explicitly revisited. - No Apple IAP — all payments via web (Stripe). App checks subscription status via backend API. Avoids Apple's 30% cut; revisit if web conversion data justifies adding IAP later. - AI image generation vendor TBD — Google Vertex AI / Imagen 3 and Flux Kontext (fal.ai) are primary candidates. Evaluated on style consistency, reliability, and cost in Phase 1 technical spike. Architecture is vendor-agnostic. - Auth: Sign in with Apple + Google. No email/password.
Approach: Experience MVP — ship the smallest thing that produces a genuinely impressive game and gets shared. No monetisation in Phase 1. Pure quality + virality validation.
Resource requirements: Solo founder viable for Phase 1. Phase 2+ benefits from contracted backbone development ($3-5K/backbone).
Supports Maya (Casual Creator) only. Jake and Sam enter in Phase 2-3.
Must-Have: - Prompt input + art style picker (cartoon, pixel, flat vector) - 1 backbone: Endless Runner (playtested, genuinely fun with placeholder art) - Asset pipeline: AI image generation (vendor TBD) + background removal + post-processing - Prompt parser: Claude Sonnet → structured JSON → backbone selection - Theme injector: assets + colour palette + text into backbone - Game preview: Phaser.js in iframe, play/restart/regenerate - Share link: persistent URL, iOS Safari optimised - QR code on desktop → opens game in iOS companion app - iOS companion app (React Native/Expo): QR scanner + WebView + share button - Sign in with Apple + Google (optional for generation, required to save) - Basic ops dashboard: generation logs, cost tracking, IP flag detection
Not in Phase 1: no monetisation, no meta-modules, no exports, no additional backbones, no game library.
Validation gate (before Phase 2): - [ ] 10 diverse prompts → 10 visually distinct, impressive games - [ ] Share rate >15% of sessions - [ ] At least one organic social post showing the quality - [ ] Generation <30s cold
| Risk | Likelihood | Mitigation |
|---|---|---|
| Asset quality doesn't impress | Medium | Curate MVP launch assets manually; 50 pre-generated fallback sets |
| Backbone not fun with placeholder art | Low | Quality bar: must be genuinely fun before any AI theming |
| Apple rejects companion app | Low | Lightweight scope (QR + WebView), clean, no IAP steering |
| AI image API reliability | Medium | Fallback library; retry logic; vendor benchmarked in Phase 1 spike |
| Solo founder bandwidth | High | Strict Phase 1 scope — no new backbones until validation gate hit |
.ipa fileGeneration speed is a core product experience — it determines whether users share or abandon.
| Integration | SLA | Failure Behaviour |
|---|---|---|
| AI image generation API | <15s per batch | Fallback to pre-generated theme set; user notified |
| Claude API (prompt parser) | <3s | Retry once; if fails, ask user to rephrase |
| GitHub OAuth | Auth in <30s | Graceful error; user can retry |
| EAS Build (.ipa) | Async, 5-15 min | Status polling; notify on complete or failure |
| Supabase Auth | 99.9% uptime | Anonymous session fallback; sync on reconnect |
| Cloudflare R2 / CDN | 99.95% uptime | CDN redundancy handles this |
| Stripe | Handled by Stripe | Platform does not process card data directly |
react-native-webview for game playback.ipa exports (separate from companion app).ipa generation; GitHub API for repo creation