Launching today

Perfstaq
Are you a top 1% developer? Prove it in one command
16 followers
Are you a top 1% developer? Prove it in one command
16 followers
Anyone can claim 5 years of TypeScript. Perfstaq lets you prove it. One command reads your own git history and your Claude Code, Cursor and Codex sessions, on your laptop. Your code never leaves it. Then GitHub spot-checks your commits at random to prove they're really yours. You get your exact standing (Top 1%, 2%, 3%…), cards made to share, and a link any client can verify in 5 seconds. Free, with an open-source scanner.








Free
Launch Team / Built With

Intellemo Cinematic StoriesApprove your storyboard before rendering your AI video
Promoted



Hey Product Hunt 👋 I'm Sathvik.
Quick question: are you a top 1% developer? Most of us think we're above average, and none of us can prove it. A résumé line like "5+ years of TypeScript" looks exactly the same whether it's true or not. On Upwork and Fiverr, honest developers who are new to the platform lose gigs to people who simply claim more.
So I built Perfstaq. Run one command:
npx perfstaq scan
It reads your real work, on your laptop. Your own commits, plus how you use Claude Code, Cursor and Codex. Your code, file names and AI chats never leave your machine. The scanner is open source, so you can check that yourself.
GitHub confirms it's yours. We pick your public commits at random, after you upload, and check each one on GitHub: it exists, you wrote it, and the size matches. Fake one, and the ✓ Verified seal doesn't appear.
You get your number. Your exact standing (Top 1%, 2%, 3%…), plus cards for your best stack, years per language, test discipline, how much of your AI code survives review, and more. Post them on LinkedIn, X or Instagram. Anyone hiring you can paste your link into perfstaq.com/verify.
Why percentiles, not ranks: you're measured against every verified developer plus a fixed benchmark of 1,000 anonymous reference profiles across web, AI, systems, mobile and data. Being early doesn't make you #1. You land where your work does.
What it can't prove, said up front: private repos can't be checked on GitHub, so those profiles say "self-reported" and aren't ranked.
It takes about a minute and it's free. Run it and drop your % below. I'll reply to every comment today, especially if you think your score is wrong.
— Sathvik
Just ran it and it took about a minute. My profile came out self-reported, since most of my commits are in private repos, so no percentile for me. It still picked up 795 AI agent sessions and 154 commits of my own, which was fun to see.
I like that you say up front what it can't prove. Have you thought about a way to verify private repos, maybe through a GitHub app with read-only access? Congrats on the launch, upvoted.
@akarsh_hegde thanks for running it! A read-only GitHub app is one option. The part I’d want to get right is verifying that the commits are yours without pulling in private source code, especially when the repo belongs to an employer or client. For now, those scans stay self-reported and don’t get a percentile.
whats stopping ppl from padding their history with a ton of tiny commits to bump their number
@igorblink A commit isn’t a unit of work. Two developers can make the same change and split it into a different number of commits. Perfstaq looks at Git history alongside other signals, like test discipline and how much AI-written code survives review. GitHub verification is separate from scoring: it checks that the sampled work is yours, not that the score is immune to padding.
Congrats on the launch, Sathvik, upvoted. Proving it with random spot-checks of public commits is a smart way to make a claim like "5 years of TypeScript" mean something. Measuring against a fixed benchmark of 1,000 reference profiles, not ranks, also means early users don't get an unfair edge. Does the scanner count work done through Claude Code, Cursor and Codex differently from hand-written commits when it scores test discipline?
@adityaharish2002 thanks for the upvote! Perfstaq reads both Git history and AI sessions, and the card for how much AI-written code survives review is separate from test discipline.