Web and mobile apps
- React 19
- TypeScript
- Vite
- TanStack
- Tailwind CSS
- PWA
- Capacitor
- iOS app
Fitness appPublished by jCode
Fitness appPublished by jCode
WorkoutGen: personalised training that adapts after every session
WorkoutGen generates your training program from your profile, and its coach adjusts it after every session. It is designed to learn from its members' training, with their consent.
- Personalised programs
- Adaptive AI coach
- 350+ coach videos
- Members
- 51,000+ (October 2026)
- Platforms
- Web · iOS · Android · Windows
- Languages
- 9
- My role
- Co-founder, sole technical owner
001
001·The product
A program built around your profile
WorkoutGen writes your program, guides each workout and keeps adjusting the plan as you train, in the browser or in its store apps, in nine languages, with a free plan and an optional Max subscription.
How it works
Your profile
Goal, experience, training days, equipment and place go through nine deterministic steps and five bounded policy layers: the same profile always gives the same eight-week cycles.
Your workouts
Follow the exercise videos filmed by coach Jeremy Tellier and log your sets and loads, even offline, from one React codebase on the web and in the store apps.
Your coach
After every finished session, a capacity score is recomputed from completion, frequency, reps in reserve and records; the coach adjusts volume and rest and suggests your loads, with no generative model.
Your progress
Your history, progress analysis and muscle recovery show what each week of training brings.
002
002·Screenshots
WorkoutGen on the App Store and in the app
003
003·What sets it apart
An engine, a coach and a learning loop that work together
No prewritten program and no chatbot: verifiable training rules, which the coach tunes after every session.
A program engine you can trace and test
Nine pure steps and five bounded policy layers turn a member's profile into eight-week cycles. A layer that would break a training rule falls back to the reference program, and every scientific reference comes from one registry checked against PubMed and Crossref.
- 9 steps · 5 policy layers
- 8-week cycles
- 419 audited exercises
- 350+ coach videos
A coach that adapts after every session
After each finished workout, a capacity score is recomputed from objective signals only: completion, frequency over 84 days, reps in reserve and personal records. The coach then adjusts secondary volume and rest, never the intensity or the main exercises.
- Objective signals only
- Volume ±10% · rest
- For every member
Learning from members, with consent
Since August 2026, members who opt in share pseudonymised, objective training data. Experiments will test one coach-validated change at a time, and only a result confirmed on a later group and reviewed like code will reach the engine: no model runs in the app today.
- Opt-in
- Pseudonymised
- Human review before release
One codebase on four platforms, offline
One React 19 codebase, packaged with Capacitor for the app stores and installable as a PWA. The interface also reads right to left in Arabic, and the API runs on Cloudflare Workers with an isolated database per member.
- Web · iOS · Android · Windows
- Offline sessions
- 9 languages
004
004·Technologies
The stack behind WorkoutGen
A TypeScript product on Cloudflare's network, from the web app to the store apps, which I developed end to end.
API and data
- Hono
- Cloudflare Workers
- Cloudflare D1
- Drizzle ORM
- KV
- R2
- Durable Objects
- Better Auth
Training engine and AI
- Deterministic program engine
- Adaptive coach
- Weight suggestions
- Progress analysis
- Opt-in learning experiments
- Scientific reference registry
Product and operations
- Astro website
- Strapi
- PostHog
- Nine languages
- GitHub Actions
- Observability
Development method
- Claude Code
- Cursor
- Project instructions for agents (AGENTS.md)
- Ragmir knowledge base of the repository, queried through MCP
- 547 automated test files (October 2026)
- Human review of sensitive changes
005
005·Explore more
Related products and services
006
006·Contact
Let's talk about your project.
Tell me what you want to build or improve, your budget and your target date.
- Fully remote
- English or French
- France · UK · US
Describe your project on Malt
Ask a question on LinkedIn007
007·FAQ
Questions about WorkoutGen
How do I access WorkoutGen?
In a browser, or through its apps on the App Store, Google Play and the Microsoft Store. The WorkoutGen website lists the current installation links.
How are the programs created?
By a deterministic engine of nine steps and five bounded policy layers, which I designed and developed. Jeremy Tellier, a French state-certified coach and the app's co-founder, validated it against training methodology.
Is there a free plan?
Yes. The program and workout tracking are free; Max is an optional paid plan, described with its prices on the product website.
How does WorkoutGen use AI?
Its coach suggests loads and adjusts volume and rest after every session from objective signals, without any generative model. I built it with Claude Code and Cursor; on client code, I follow each client's rules, as set out on the agentic development page.
Where can I read about personal data?
On the WorkoutGen website, which publishes the current privacy policy.
Who develops WorkoutGen?
I co-founded WorkoutGen and handled its entire technical side from June 2025 to September 2026, as founder of jCode: architecture, frontend, backend, infrastructure and deployment. My partner Jeremy Tellier, a French state-certified coach, provides the training content and validated the program engine.