Hey HN,

I built Doculearn after watching my team (and myself) ship faster than ever with Claude, Cursor, and Copilot—but understand less and less of what we were actually deploying.

The vibe coding problem:

We'd accept a 200-line AI suggestion, tests pass, PR approved, merged. Two weeks later: "Wait, how does this authentication flow work again?" No one knew. The person who approved it just trusted the AI. The person who merged it moved on to the next feature.

Sound familiar?

What Doculearn does:

It watches your GitHub activity and automatically generates flashcards from YOUR code. Not generic "what is a closure?" cards—actual flashcards about the authentication middleware you merged yesterday, the API endpoint you refactored, the algorithm you copy-pasted from Claude. When you push code, Doculearn:

Generates spaced-repetition flashcards from your commits

Updates your team board automatically (no more manual Jira)

Creates "Context Cards" that surface when you're working on related code

Sends daily/weekly LogLetters showing what everyone shipped

Why this matters in the AI era:

You can ship 10x faster with AI. But if you don't understand what you shipped, you can't:

Debug it when it breaks at 2am Extend it for the next feature Explain it in code review Onboard new teammates

We're optimizing for velocity at the cost of understanding.

Doculearn tries to bridge that gap.

How it works:

Connect your GitHub repo Doculearn analyzes commits, PRs, code changes Azure AI agents generate personalized flashcards Review cards in your workflow (web, mobile, CLI coming) Team stays synced on what everyone actually knows

The stack:

Next.js + Django on Azure Container Apps Azure AI Foundry for flashcard generation GitHub Apps for real-time monitoring PostgreSQL for persistence

Current features:

AI-generated flashcards from commits Context Cards (study while you build) Auto-updating team/work boards Bug tracker with AI-suggested fixes LogLetters (changelogs from GitHub) Social login (GitHub, LinkedIn, X, Microsoft)

Things I'm wondering:

Do you find yourself shipping code you don't fully understand?

How do you currently retain knowledge about your codebase? Would flashcards feel like homework or helpful? Is "team knowledge sync" a problem you experience?

Try it: doculearnapp.com – Live now with 7-day free trial

I've been testing this with early teams for the past month.

The most common feedback: "I didn't realize how much I forgot until the flashcards reminded me."

Would love HN's feedback. Is this solving a real problem or am I overthinking the vibe coding phenomenon?

Show HN: Doculearn – How much of your Gen-AI code do you understand?