Apptruth
Developer ToolsDiscover your app's sneaky behaviour before your users do
What the community said1
Martin SailliezWhat works
The demo result is concrete and technical, not vague: "Free-plan limit can be bypassed, 96% confidence, API accepts project #4, fix in projects/create.ts" reads like a real finding, not a mockup. Read-only GitHub access plus a fix-ready prompt (not just a warning) is the right shape for a vibe coder who wants a next step, not a report to interpret.
The biggest problem
The confidence score (96% in the example, "9.6/10 issue confidence") is presented with zero explanation of what it's measuring or how it's computed. For a tool whose whole pitch is "catch behavior you didn't expect," an unexplained accuracy number is exactly the kind of unverified claim the product is supposed to protect against.
Top suggestion
Add one sentence next to the confidence score explaining what it's based on (static analysis, pattern match, LLM review of the diff). Founders will trust a number they understand more than a bigger one they don't.
Is it useful for vibe coders?
Yes, more than for experienced devs. Someone who vibe-coded a Stripe/auth flow with an agent and never manually reviewed the logic is exactly who benefits from a tool that catches "the free plan limit can be bypassed" before a user finds it.
How accurate is it? (i noticed 98% accuracy)
I didn't see a 98% figure on the page itself, only a 96% confidence score on the one example finding shown ("Free-plan limit can be bypassed"). If 98% is quoted elsewhere, it needs the same treatment: say what it's measured against (false positive rate on a known test set, real bugs caught, etc.), otherwise it reads as an unverifiable number like the ones I've flagged on other roasts this week.
Would you pay for it?
At $29/mo, maybe, for one high-stakes repo before a launch. I'd want to see the confidence methodology first, paying to trust a black-box score is the same leap of faith the tool is meant to remove.
I guess you didn't try out the real flow then ;) Sign up (free account) and run a test on your github repo, it will find you exactly the what, why, where of every finding (with a deeper explanation and exact file location). its very detailed. lmk once you do, maybe you're assessment will change :)
Sign in to join the discussion.
Want to roast back, or ship your own?
Reading is open to everyone. Posting a roast or a project takes a free account.
Sign in, it's free