Frequently asked questions

Anyone can browse. To sign in and submit projects you must be 18 or older (or the age of majority where you live, if higher) and accept the Terms of Use and Privacy Policy; you confirm both with one checkbox before signing in. Your acceptance is stored with your account and deleted with it.

Public GitHub repositories owned by the GitHub account you signed in with, that were vibe-coded (written by AI from your prompts). You cannot submit someone else's repository. If your work lives under a second GitHub account, sign in with that account to submit it (while sign-up slots are available).

Projects up to roughly 50 MB or 2,000 files of source code are supported. Binaries, archives and model weights are skipped automatically.

The program was written by AI from your prompts, using tools such as Claude, ChatGPT, Cursor, GitHub Copilot, Windsurf, Cline or Aider, usually by someone who is not a programmer. A professional developer using AI assistance in their normal work is not what LetsVibe grades. We look for traces like assistant configuration files, AI co-author lines in commits, a README that says the project was written by AI, or code with habits typical of AI. If none are clear (for example because you copied code from a chat window), your project is still graded and the site owner reviews it before it is published. Projects whose history starts before November 2022 and forks of someone else's project are not graded; if that happened by mistake, you can appeal (below).

Usually 3–15 minutes, and up to about 20 for large projects, plus time waiting for projects ahead of yours (first come, first served). Everything runs on one local graphics card, one project at a time. You can close the progress window; grading continues in the background.

There is a submission limit per account during the beta so everyone gets a turn on the single GPU; your remaining submissions are shown next to your name. Re-grading the same commit is never allowed (nothing would change); push a new commit first.

Not currently. LetsVibe only reads public repositories and never asks for access to private ones.

Other vibe-coded projects: programs written by AI from someone's prompts. LetsVibe never compares your project with software written by professional developers. Your letter grade comes from a bell curve, and your report shows your project's rank in its category (for example "#3 of 41 in Games").

A typical working vibe-coded project lands around the middle of the scale, which is a C, so a C means "solid, like most working projects", not "bad". Most of the weight is on what users experience: does it do what it promises, how smooth is the journey from install to finished task, how pleasant is it to use, how valuable and how original is it. Each report lists the strongest features, the areas needing the most attention, and separately, what a professional team would do next.

Yes, on a curve often called a bell curve. AI reviewers hedge toward the middle, so raw scores bunch up and say little on their own. Your letter comes from how far your score sits above or below the average, in standard deviations. Two public reference projects (one written to satisfy every factor, one written to ignore every factor) set the bell at first; as your category grows, its own average and spread take over. The Grading curve page shows exactly how the curve was built, including both reference projects, the prompts that produced them and every grading run.

Because grades are on a curve. As projects join your category, its average and spread count for more, so the bell shifts slightly. Your raw score itself only changes when your project is re-graded.

AI models are not perfectly deterministic, so repeated evaluations can differ by a few points. Treat small gaps between projects as a tie. Larger differences usually come from a new commit or a new rubric version.

No. LetsVibe detects the target platform and hardware from your build files and grades the interface, performance and architecture against that device. See how grading works.

No. LetsVibe is not a security audit, and a report never certifies that software is safe to run.

Sometimes the review needs more context about what specific code does before a report can be published. You'll see an Action required card on My Projects listing the exact lines. Describe what they do and why they're needed; the report is published after your explanation is reviewed. If there is no explanation within 7 days, the report is deleted. You can also withdraw it at any time.

Open your project and choose Report an error in this review. Pick the statement, and explain what is wrong, ideally naming the files or functions that show it. It is checked against your code once; if confirmed, the review is corrected automatically, otherwise it is reviewed by hand. Track progress on My Projects.

Work through the "Areas Needing Attention" on your report, push a new commit, then use Re-Analyze on your project page.

Open the rejected submission and choose Appeal. Explain how AI was used (tools, prompts, workflow). Appeals are reviewed in the order they arrive and the decision appears on your My Data page. No email address is needed.

Scores themselves are not appealed one by one; the same rubric applies to everyone. If the review states something factually wrong (for example it claims a feature is missing that clearly exists), use Report an error in this review.

Yes, at any time and without asking anyone. Go to My Data to delete one project, all of your projects, or your whole account. Nobody is notified. See Remove my project.

No. The models run on LetsVibe's own hardware. The temporary copy of your repository is deleted after grading. Details are in the privacy policy.

Only owners can submit their repositories, so a listing normally means the owner submitted it. If you believe a listing should not be there, contact us and we will review it promptly.