Vision & Strategy¶
Honest Assessment (June 2026)¶
What exists¶
The original platform (8 stores) is production-grade. FAS, FGS, PAS have real SDKs (published to npm), real CLIs, real backend Workers, real auth, real CI/CD, real apps/games shipping. 344 repos, 170+ live items. Stripe payments wired on PAS. This is a real product.
The new stores (5 free + 1 pro) are content-rich prototypes. FRS, FQS, FBS, FCS, FSS have 72 working interactive simulations and 34 educational pages — more content than most edtech startups ship in a year. But they're static HTML on Cloudflare Pages. No auth, no CI/CD, no npm-published SDKs, no GitHub orgs, no user data. The "creator profiles" are hardcoded. The "challenges" have no backend scoring. They're impressive demos, not production products.
The SDKs are written but not published. Four TypeScript packages (quantum circuit, molecule analyzer, orbital mechanics, logic simulator) exist as source code in the monorepos. None are on npm. No GitHub orgs exist to push them to. The API surface is real; the distribution isn't.
What's the gap¶
| Layer | Original stores | New stores |
|---|---|---|
| Content/demos | Yes (170+) | Yes (72) |
| SDK (published npm) | Yes | No (source only) |
| CLI | Yes | No |
| Auth (GitHub OAuth) | Yes | No |
| Backend (D1/R2/DOs) | Yes | FRS only |
| CI/CD (GitHub Actions) | Yes | No |
| GitHub orgs | Yes (8 orgs) | No |
| Custom domains | Yes (all .online) | FRS only |
| Compliance checks | Yes (20-42) | No |
| Analytics | Partial | No |
| Payments (Stripe) | PAS only | No |
| Users | Pre-launch | Zero |
What this means¶
The new stores are 6-12 months behind the original stores in infrastructure maturity. But they have something the original stores took months to build: content. 72 simulations, each one a self-contained interactive experience. The content is the hard part. Infrastructure is a proven vendor-copy.
The vision¶
What we're building¶
Open Frontier — a constellation of free interactive learning platforms covering the hardest subjects in science and engineering. Each platform is:
- A store of simulations — click and learn, no install, no signup, no cost
- An educational institute — curated courses, papers, tools, community
- A component library — importable SDK engines for developers and AI agents
- A creator platform — build simulations, get a profile, join a community
The thesis¶
Every subject that requires expensive software, specialized hardware, or PhD-level access to understand can be democratized through browser-based interactive simulation.
Quantum computing? You don't need IBM hardware — simulate circuits in your browser. Chip design? You don't need Cadence ($100K/yr) — build a CPU from logic gates. Molecular biology? You don't need a wet lab — fold proteins and analyze molecules. Orbital mechanics? You don't need STK ($10K/yr) — plan missions interactively.
Why this matters now¶
Three converging trends:
- WebGPU + WASM — browsers can now run physics engines, neural networks, and scientific simulations that required native apps 2 years ago
- AI models in browser — Transformers.js runs DETR, SAM, ESM-2, SmolLM entirely client-side. No API keys, no server costs, no privacy concerns
- AI-assisted development — a solo founder can build 72 interactive simulations in one session with Claude Code. The production cost of educational content has collapsed
The model¶
FREE TIER (browser-only)
├── Interactive simulations (72 and growing)
├── Educational institute (34 deep-dive pages)
├── SDK packages (importable engines)
├── Challenges & learning paths
└── Creator profiles & community
PRO TIER ($9/mo — future)
├── Server-powered simulation (30+ qubits, GPU training)
├── Real hardware access (IBM Quantum, lab equipment)
├── API endpoints (programmatic access)
├── MCP tools (AI agent integration)
├── Persistent state (save projects, share)
└── Export to professional formats (Qiskit, Verilog, GCODE)
Free is genuinely free — the simulations run on the user's hardware. Pro adds server compute for workloads that exceed browser capabilities.
Revenue¶
Pre-revenue. Building density before monetization. Target: 200+ simulations across all stores, then launch Pro tier at $9/mo.
Revenue comes from Pro subscriptions. Free stores are acquisition channels, not cost centers (Cloudflare free tier covers hosting).
Who we serve¶
Today: Students, hobbyists, self-learners who want to understand hard subjects by doing, not reading.
Tomorrow: Professionals who prototype on free, then pay for production (quantum circuit optimization, drug screening, mission planning).
Long-term: AI agents that call our simulation engines as tools to solve problems.
Goals¶
3-month goals (by September 2026)¶
- [ ] Register domains for all new stores
- [ ] Create GitHub orgs, push all repos
- [ ] Publish all 4 SDKs to npm
- [ ] Add auth (GitHub OAuth) to all new stores
- [ ] Set up CI/CD for all new stores
- [ ] Reach 100+ simulations across all stores
- [ ] Launch Pro tier on quantum store (PQS)
- [ ] Register openfrontier.dev as umbrella domain
- [ ] First external creator contributes a simulation
6-month goals (by December 2026)¶
- [ ] 200+ simulations
- [ ] MCP tools live (AI agents can call simulators)
- [ ] SDKs used by 3+ external projects
- [ ] Pro tier revenue >$0
- [ ] 5+ external creators with published profiles
- [ ] Institute content cited by 1+ university course
12-month goals (by June 2027)¶
- [ ] 500+ simulations
- [ ] 10+ store domains
- [ ] Pro revenue covers infrastructure costs
- [ ] SDK packages have 1000+ npm downloads
- [ ] AI agents routinely use our simulation tools
- [ ] Partnership with 1+ educational institution
North star¶
The Khan Academy of interactive simulation. Free, comprehensive, covering every hard technical subject. But instead of videos, our content is interactive — you don't watch someone explain quantum mechanics, you build and run quantum circuits yourself.
What we are NOT¶
- Not a MOOC — no courses, no enrollment, no completion certificates (yet). Just simulations and reference material.
- Not a professional tool — our simulators are educational, not production-grade. A real chip designer uses Cadence, not FreeChipStore.
- Not an AI company — we use AI models, but we're an education platform. The AI makes the simulations better, not the other way around.
- Not a marketplace (yet) — creators can't sell simulations. Everything is free and MIT-licensed.
What makes this defensible¶
- Content density — 72 simulations built in one session. The marginal cost of adding more is near-zero with AI assistance. Competitors need teams of developers.
- SDK composability — simulations aren't just demos, they're importable libraries. This creates an ecosystem moat.
- Cross-domain coverage — no one else covers quantum + bio + chips + space + robots under one umbrella. Each store is niche; the platform is broad.
- Zero marginal cost — everything runs on the user's browser. No server costs per user. Scales infinitely.
- AI-native creation — every simulation was built with AI assistance. As AI gets better, we can build faster. The creation cost decreases over time.