Last weekend I let an AI roast 50 random GitHub profiles — staff engineers, bootcamp grads, indie hackers, recruiters who somehow have hello-world as their most-starred repo. After roast #12 I noticed it was repeating itself. By roast #38 I realized the AI wasn't really insulting people — it was naming patterns the rest of us politely ignore.
Here are the seven that came up over and over again. If you've shipped anything to GitHub in the last two years, at least three of these are about you. I scored myself a 5/7. Brutal.
The single most common roast was some version of: "You have 23 repos with one commit and a README that says 'WIP'. We need to talk."
Across the 50 profiles, the median ratio of repos that were touched exactly once to repos that survived more than 30 days was 3.2 to 1. In other words, for every project you actually built, there are three half-corpses with optimistic names like better-todo-app and game-engine-from-scratch.
The AI's verdict on this pattern was the same every time: it's not a productivity problem, it's a commitment problem. You like the dopamine of git init more than you like the slog of issue #47.
"Your repos look like a self-help bookshelf: lots of confident first chapters, very few finished books."
Pick the five repos you actually care about. Archive everything else. The archive button is right there — it doesn't delete history, it just stops the embarrassment from being the first thing recruiters see.
32 of the 50 profiles listed 4+ languages in their bio. When the AI looked at actual commit volume, 78% of their code was in one language. The other "languages" were a setup.py here, a Dockerfile there, and one heroic attempt at Rust that ended in a panic!.
This roast hit harder than any of the others, because it's the most honest one. There's a difference between having written Go and being a Go developer. The AI doesn't care about your bio — it looks at your file extensions.
14 profiles had a very specific commit graph signature: an explosive 18-month run, three to five "talk slides" repos, and then… silence. Sometimes for years.
The AI nicknamed this "the Medium post phase". The pattern is real: somewhere between your second and third year as a developer, the urge to write about coding overtakes the urge to actually code. The AI flagged it for what it is — a transition into a different career (DevRel, management, content) that the profile owner hadn't admitted to themselves yet.
This is the only roast that made one of my friends genuinely upset. He'd been telling himself he was "just busy". The AI was kinder than his therapist.
You know the profile. 312 public repos, 0 original. Forks of awesome-rust, awesome-react, awesome-awesome-lists. The contribution graph is a tundra.
The AI's roast for this pattern was the meanest one of the experiment: "You don't have a GitHub profile, you have a bookmark folder pretending to be a portfolio."
This is also the easiest pattern to fix. Forks don't show up on your contribution graph unless you commit to them. You can delete or hide forks in bulk from your profile settings. If you've been padding numbers, the AI sees it. So does every senior who's hiring.
11 profiles had contribution graphs that were essentially empty Monday through Thursday and lit up like a Christmas tree on Friday at 10pm and Saturday at 2am. The AI's diagnosis: "Your day job is killing the part of you that loved this."
This is the only roast I felt protective about. There's a difference between a hobby and a coping mechanism. If your only joyful coding happens at 1am on a weekend, the problem isn't your discipline — it's your job.
The AI doesn't know that. It just sees the pattern.
Six profiles had the same signature: one repo, 4,000+ commits, no users. The AI called it "the cathedral problem" — you've spent 18 months building a perfect framework for a product nobody is asking for.
Commits with messages like refactor auth (final, really), cleanup before v2, switching state mgmt are the tell. You're not shipping. You're sculpting marble that nobody will ever stand under.
If your most-starred repo has fewer stars than your number of refactor commits, you're in the trap. Ship a broken thing this week. Watch what happens.
Almost every profile had at least one pinned repo that was abandoned 18+ months ago. The AI called these out by name every single time: "This is your top-pinned project. Last commit: Q3 2024. It's giving 'I peaked in college'."
Pinned repos are the most public part of your GitHub. They're your highlight reel. And most of us are highlighting projects we wouldn't actually defend in a code review.
If you can't open the repo today and ship a meaningful commit in 20 minutes, it doesn't belong pinned. Replace it with something newer, smaller, and more honest.
The roast is funny because it's specific. The AI isn't generating random insults — it's reading the artifacts of a thousand small decisions and pointing out the story those decisions tell. That's also what every senior dev does when they open your profile to evaluate you. The AI just says the quiet part out loud.
None of this means your GitHub is "bad". Mine got roasted hard. The point isn't to clean it up to look impressive — it's that your profile is already telling a story about you, and you might as well know what that story is before someone else reads it.
It takes 12 seconds. The AI reads your actual repos, languages, commits, and stars — no signup, no email, no "premium tier". You get a shareable card you can post or send to the friend who keeps subtweeting your code.
Roast My GitHub →I pulled 50 GitHub usernames from a mix of sources: 20 from the GitHub trending page, 10 from "Show HN" threads, 10 from the byline of recent dev.to posts, and 10 from a random sample of awesome-list contributors. I fed each one through github-roast, captured the output, and tagged the recurring themes. The patterns above were each flagged in at least 8 of the 50 profiles. The full breakdown lives in a follow-up post.
If you want to make this less brutal and more useful, pair the roast with two related TinyTools: the Resume Tailorer to make sure your written story matches what your GitHub says, and the Cover Letter Generator for when you want to explain (away) the Friday Night Hero pattern.