GitHub Repository

🧠 Local-first CLI that gives your repo durable, AI-ready memory — features, ADRs, modules, architecture, standards & tests — and a 🩺 deterministic doctor that validates health, evidence & drift. Git tracks what changed; Persist OS tracks why.

6 starsTypeScript

AI Is Too Good at Coding, Let's Make It Better

Name: AI Is Too Good at Coding, Let's Make It Better
Availability: InStock
Author: karthickrmchn

by karthickrmchn·Jun 26, 2026·1 point·0 comments

Visit Project View on HN

AI Analysis

●●SolidBig BrainSolve My Problem

Git tracks what changed; this tracks why — with a CI-gateable doctor command.

Strengths

•Decisions live in repo files, surviving session compaction and context loss
•Deterministic doctor returns exit codes for CI hooks and automation
•ADR supersession tracking keeps decision trails auditable instead of contradicted

Weaknesses

•AI project memory is a crowded category with supermemory, mem0, and others
•No network means no collaborative features for team-wide decision sync

Post Description

In the past few months, I’ve been building and shipping things extremely fast. Some of the MVPs that I built for my friends were completely vibe-coded and shipped within a day to show a demo. While it was a great experience working with AI tools, I wanted to make sure some basics were followed, decisions of what and why written down, obvious things not skipped and that it could scale.

That’s why I built persist-os. It’s a CLI that runs locally. No network required. It records the architecture decisions and standards in plain files in the repo and 'persist doctor' checks them - it flags if anything contradicts or is missing, or if the task is done with no tests. Because the rules live in your repo files, it doesn't matter if you start a new session or the session is compacted - the standards and rules survive.

MIT licensed. Here’s how you can try it: npx persist-os@latest init