Path
A learning system that turns a broad goal into a prerequisite graph, maps resources onto the nodes, and answers one question: what now?
- Type
- Web App
- Role
- Creative Technologist
- Built
- -
- Updated
- -
- Visit
- path.chiradeep.me
- Source
- GitHub
Tech stack
- Next.js
- React
- TypeScript
- AI
- Graph data
- Supabase
The internet doesn’t have a shortage of learning resources. It has a shortage of direction.
One next action, always named.
How do I learn this?
- Courses.
- Videos.
- Books.
- Roadmaps.
- Bootcamps.
Search “how do I learn machine learning” and the page answers a different question: what exists. Path answers the one underneath it.
Given what is already known, one node is open. The rest stay shut until the prerequisites are real. Python and the mathematics both have to be done before that node unlocks.
A graph, not a playlist
A course is a list. Path is a graph of concepts, prerequisites, skills, and projects. A node stays closed until the ones under it are done.
Python sits beside mathematics. Both have to be real before machine learning opens.
type LearningNode = { id: string title: string type: 'concept' | 'skill' | 'project' difficulty: number prerequisites: string[] resources: string[]} // A node with an open prerequisite cannot be the next step.What now
The engine reads the goal, what is already known, completed nodes, prerequisites, the project in progress, and recent work. It returns one action.
Not “study algorithms.” Implement binary search from scratch and test it against five edge cases.
Goal ↓ Graph ↓ Prerequisites ↓ Progress ↓ One next action
Resources stay on the node
A course, a book, a paper, an exercise, and a project can all hang on the same concept. They support the node. They are not allowed to become the path.
The hard part
Structure without becoming a syllabus
Too much structure is another rigid curriculum. Too little is another search engine. Either failure answers a question the learner did not ask.
Solution
Prerequisites are strict. The resource used to cross a node is not. What now returns one action, chosen from the graph, not a pile of links.
Trade-offs
| Decision | Choice | Why |
|---|---|---|
| Shape | A prerequisite graph | A playlist hides what is missing. |
| Next step | One action | Forty-seven courses is not a decision. |
| Materials | Attached to a node | The resource must not replace the path. |
What I’d change
- A way to verify a skill, not only to mark a node done.
- Project evaluation, and stronger inference of prerequisites.
- Adaptation that still makes sense after months, not only after a week.
What now, exactly
I already have a roadmap. What is different?
A roadmap is a list you can skip around. Here a node stays closed until the ones under it are done. Machine learning does not open because you bookmarked it.
What do I actually get when I ask what now?
One sentence you can start. “Implement binary search from scratch and test it against five edge cases.” If the answer is a pile of courses, the engine failed.
Where did all the courses go?
Onto the node they belong to, next to the book, the paper, and the exercise. You pick one to cross the concept. None of them is the path.