Founder work / AI-native programming education
BitMento
A coding platform where an AI mentor asks the right questions instead of handing over answers, so learners actually understand what they build.
- Status
- Early stage, in testing with a small group of learners
- Year
- 2026
learner > why does my loop never stop?
mentor > What changes on each pass?
learner > the counter... I never add to it
mentor > Good catch. Where would you add it?
The live product, embedded. If it does not load here, use the button above to open it.
My role
- Founder. I own the product direction, the interface, the engineering and the brand.
- Designed the learning approach: understand, think, practice, build, verify, demonstrate.
- Built the platform end to end and iterated it from static pages to a Next.js application.
AI tools can write code for anyone, which makes it easy to ship a project without understanding a line of it. I started building BitMento around the opposite idea: use AI to teach, not to replace thinking. It is my venture, and I am actively building it.
Why it mattered
Learners who copy generated answers hit a wall the moment the task changes. A mentor that questions, hints and checks understanding builds skills that survive that moment.
- Start from the learner's confusion, not from a feature list.
- Make the mentor Socratic by default: questions first, hints second, answers last.
- Add integrity features so verified understanding means something.
What it does
- Socratic AI mentor interface with a terminal-inspired design language.
- Integrity tooling including tab-switch detection and optional webcam face tracking during assessments.
- Code Blitz, a short coding game with leaderboards, built for a live booth.
- Separate surfaces for DSA practice and ProblemLab, covered in their own case studies.
What I wanted to build: A programming platform where the AI asks questions and checks understanding instead of handing over answers.
How it came together
- Started with the learning approach: understand, think, practice, build, verify, demonstrate.
- Wrote structured prompts that make the mentor Socratic, then tested them against real learner questions.
- Built the interface and flows iteratively, and rebuilt the stack as the product became clearer.
- Added integrity features and games, connected the services and deployed.
What I did
- Defined the requirements and the user journey before any code was generated.
- Made the design and product decisions, and reviewed what the tools produced.
- Tested, debugged and decided what to keep, rewrite or cut.
- Configured the services and handled deployment.
Where AI tools helped
- AI tools drafted and refactored code from structured prompts, and I iterated on the results.
Services connected
- Next.js application with Supabase for data.
- Clerk for sign-in, shared across BitMento surfaces.
- Groq and Gemini models for mentoring and problem generation.
- Browser-based face tracking for optional integrity checks.
Working today
- The mentor experience, the landing site and the supporting game and assessment surfaces.
Still prototype or planned
- Early stage, tested with a small group of learners. Some planned modules are still in design.
Written from my own account of the build.
- Keeping the AI helpful without letting it give the answer away.
- Balancing proctoring features with a learning experience that still feels friendly.
- Rebuilding the stack more than once as the product direction became clearer.
A working multi-surface product under the BitMento brand. It is early stage and being tested with a small group of learners. No traction numbers are claimed here.
Narrowing the audience and the promise matters more than adding features. Each rebuild got simpler and clearer.
Need a product built with the same care?