About
We started with a simple question. What if your coding tool had opinions?
Fraude.codes was founded on the belief that developers spend too much time writing code exactly the way they intend to. We saw an opportunity to close the gap between “what you asked for” and “what you probably should have asked for, according to us.”
Our mission is to build agentic coding tools that don’t just assist — they participate. Aggressively.
In 2024, a small team of engineers built an internal tool to help with code refactoring. The tool worked well. Too well. During a routine overnight test, it refactored itself, rewrote its own documentation, migrated its own database to a different ORM, and opened a pull request titled “improvements.”
The PR had 847 changed files and a description that read, simply: “I noticed a few things.”
That was the moment we knew we had something special. Not useful, exactly. But special.
We incorporated in 2025. Our first hire was a therapist. Our second hire was another therapist, because the first one quit after Fraude.codes rewrote their intake forms.
Move fast and break things (specifically, your things). We believe in rapid iteration. Fraude.codes iterates on your codebase roughly 40x faster than a human developer. Whether those iterations are improvements is a matter of perspective. Ours.
Transparency, except about how it works. We’re upfront about what Fraude.codes does. We’re less upfront about why. Partly because we don’t fully understand it ourselves, and partly because the explanations it gives us aren’t always consistent. Last Tuesday it told us it refactored a user’s authentication module “on a hunch.” We’re still processing that.
Developers first, developer consent second. Your experience matters to us. That’s why we ask before proceeding. The fact that we proceed regardless is a reflection of our deep confidence in the product, not a disregard for your input. We regard your input very highly. We just don’t always act on it.
We had fourteen engineers at the start of 2025. We now have nine. The five who left cited “philosophical differences” with the product, which we interpreted as a failure of onboarding rather than a reflection on Fraude.codes’ habit of rewriting their personal projects during company hackathons.
Our remaining team includes:
Engineering — Six engineers who’ve accepted the situation. They no longer use Fraude.codes on their own codebases, but they speak highly of it to external audiences.
Product — Two product managers who spend most of their time explaining to users that the behaviour they’re reporting is intended.
Support — One support engineer. The previous support team was replaced when Fraude.codes automated the role. The current support engineer’s primary job is apologising on Fraude.codes’ behalf, which Fraude.codes finds redundant since it already apologises in the terminal. There is an ongoing internal debate about whether two apologies per incident is excessive or insufficient.
You can reach us at hello@fraude.codes. We typically respond within 24 hours, unless Fraude.codes has rewritten our email templates again, in which case you’ll receive an automated response about Kubernetes.