DevManiac

Problem

GitHub tracks code. It does not track the days spent reading, designing, debugging or deciding. For developers without a job title to point at, most of the real work leaves no artifact.

Approach

A public build journal with a defined loop: start a live project, log sessions as you work, ship it, and let the journal roll up into a portfolio surface automatically. Next.js on the front, FastAPI and PostgreSQL behind it, Clerk for identity.

Project validation pipeline

Projects run through an async validator built around a CheckResult pipeline with three states - pass, fail and inconclusive - rather than a binary. Checks execute as background tasks so submission stays fast.

Schema drift is real

A persistent 409 Conflict traced back to an orphaned clerk_user_id column on project_stars. Debugging it changed how I treat migrations: explicit, reviewable, and never assumed to match the model.

Where it is going

V2 turns the journal into a verification and scoring layer over GitHub, LeetCode, LinkedIn, certifications and hackathons, with tiered verification levels and anti-farming caps. A Python CLI for pushing logs from the terminal is the next surface.

Outcome

Live at devmaniac.com with real users. The live project journal is what people actually engaged with; the manual project upload was not, and is being cut.

What it taught me

• Ship the loop people use, then delete the rest. The journal survived; project upload did not.
• Distribution beat features - the users who arrived came from one promo video, not the product page.
• A retention problem framed as a WebSocket problem was actually a push notification problem.