The open-source AI literacy curriculum for people who ship products.
Ascent is a structured AI upskilling platform built for practitioners, not researchers. Product managers, engineering managers, and individual contributors work through 36 modules, 18 decision simulations, and 11 real-work exercises — writing their reasoning and receiving AI-generated or expert feedback. No videos. No passive slides.
Most AI training tells you what AI is. Ascent trains you to decide what to do with it.
- Key features
- What's inside
- Who it's for
- Screenshots
- AI feedback
- Stack
- Quick start
- Points and levels
- Adding content
- Contributing
- License
- Decision simulations — Realistic scenarios with rubric-based evaluation: a vendor demo with hallucinations, an AI-generated PR nobody reviewed, a headcount cut justified by automation
- Applied missions — Real-work exercises grounded in actual job tasks: draft an AI usage policy, audit a workflow for AI opportunity, write an AI feature brief, run a bias check on model output
- Multi-provider AI feedback — NVIDIA NIM, OpenAI, Anthropic, or Ollama out of the box; fully functional without any key via expert-written static feedback
- Role-filtered content — Modules, scenarios, and missions tagged for PMs, EMs, and ICs; each user sees what's relevant to their role
- Progress and levels — Points, levels, and completion tracking per authenticated user (Aware → Informed → Practitioner → Leader)
- Guest access — A curated selection of content is freely accessible without signup
- Self-hostable — One
docker compose upgets you a full stack with PostgreSQL, app, and optional Ollama - Admin panel — Content management, bug reports, and user feedback — no external tool required
| Pillar | Count | What it is |
|---|---|---|
| Foundation | 36 modules | Focused reading on AI concepts, LLM mechanics, security, ethics, evaluation, team dynamics, vendor selection, and economics. Readable in under 15 minutes each. |
| Scenarios | 18 decision simulations | Realistic PM/EM/IC situations — a vendor demo with hallucinations, an AI-generated PR that nobody reviewed, a headcount cut justified by automation. Users write their reasoning; the platform evaluates it. |
| Missions | 11 real-work exercises | Applied tasks on the user's actual job: draft an AI usage policy, audit a workflow for AI opportunity, write an AI feature brief, run a bias check on a model output. |
Progress is tracked per authenticated user. Guests can access a curated selection of content without signing up.
| Role | What they train |
|---|---|
| Product Managers | Speccing AI features, evaluating vendors, measuring ROI, communicating tradeoffs, leading AI roadmap decisions |
| Engineering Managers | Governing AI-generated code, managing team adoption, handling AI incidents, growing junior engineers in an AI-assisted environment |
| Individual Contributors | Using AI tools without losing judgment, evaluating benchmarks, owning AI feature implementation, staying technically sharp |
Landing page
Dashboard — progress overview with points, level, and completion tracking
Mission — applied real-work exercise with checklist and AI feedback
Profile — user level, earned points, and role
Users can opt into AI feedback at submission time — toggle it on before submitting a scenario or mission. When enabled, an AI model evaluates the response. When off, or if no provider is configured, the platform falls back to expert-written static feedback. Four providers are supported out of the box:
| Provider | Default model | Key required |
|---|---|---|
| NVIDIA NIM (default) | meta/llama-3.1-70b-instruct |
Yes — build.nvidia.com (free tier available) |
| OpenAI | gpt-4o-mini |
Yes |
| Anthropic | claude-haiku-4-5-20251001 |
Yes |
| Ollama | llama3.2 |
No — runs fully locally |
Set AI_PROVIDER and the corresponding key in .env.local. Without a key, the platform falls back to expert-written static feedback — the app is fully functional either way.
| Layer | Technology |
|---|---|
| Framework | Next.js 14 (App Router), TypeScript |
| Styling | Tailwind CSS, Radix UI primitives |
| Database | PostgreSQL via Prisma ORM |
| Auth | NextAuth v4 — credentials (email + bcrypt), JWT sessions |
| AI | NVIDIA NIM / OpenAI / Anthropic / Ollama — all optional |
| Container | Docker + Docker Compose |
| Deployment | Vercel-ready |
git clone https://github.com/divarun/ascent.git
cd ascent
cp .env .env.local # then edit .env.local — set NEXTAUTH_SECRET at minimum
docker compose up # starts app + postgres
docker compose exec app npm run db:seedOpen http://localhost:3000.
git clone https://github.com/divarun/ascent.git
cd ascent
npm install
cp .env .env.local # set DATABASE_URL and NEXTAUTH_SECRET
npm run db:setup # migrate + seed
npm run devdocker compose --profile ollama up
docker compose exec ollama ollama pull llama3.2
# add to .env.local: AI_PROVIDER=ollamaRequired environment variables
| Variable | Description |
|---|---|
DATABASE_URL |
PostgreSQL connection string |
NEXTAUTH_SECRET |
JWT secret — generate with openssl rand -base64 32 |
NEXTAUTH_URL |
App base URL (e.g. http://localhost:3000) |
Full environment variable reference and setup guide: detailed.md
Users earn points for completing content. Points are difficulty-weighted and configurable in src/config/scoring.ts — no values are hardcoded elsewhere.
| Beginner | Intermediate | Advanced | |
|---|---|---|---|
| Module | 10 pts | 25 pts | 40 pts |
| Scenario | 30 pts | 50 pts | 75 pts |
| Mission | 25 pts | 40 pts | 60 pts |
| Level | Name | Points required |
|---|---|---|
| 1 | Aware | 0 |
| 2 | Informed | 100 |
| 3 | Practitioner | 300 |
| 4 | Leader | 600 |
All content lives in src/data/ as TypeScript files — one file per module, scenario, or mission. Register it in the corresponding index file, then re-seed.
# After adding a file:
npm run db:seed
# or in Docker:
docker compose exec app npm run db:seedSee detailed.md for the full content schema and authoring guide.
Contributions are welcome — new scenarios, modules, missions, bug fixes, and platform improvements.
The highest-value contributions are new content: realistic decision scenarios, focused concept modules, and applied missions grounded in actual PM/EM/IC work.
See CONTRIBUTING.md to get started.