Every company that runs software on the internet faces the same nightmare: something breaks in production, users can't access the service, and the engineering team scrambles to find out why. For a mid-size e-commerce company, every minute of downtime costs between $5,000 and $50,000 in lost revenue. For larger platforms, it's more. The clock is running, the CEO is asking questions, and the team is staring at thousands of lines of logs trying to find a needle in a haystack.

This is the problem we set out to solve when we founded Sonarly.

The hidden cost of production bugs

Software teams spend, on average, 30-40% of their engineering time not building new features, but investigating what broke and why. A senior engineer earning $200,000 a year spends roughly $60,000 worth of their time per year just debugging production issues. Multiply that across a team of ten, and you're looking at $600,000 annually in pure investigation time, before you even count the cost of the downtime itself.

The deeper problem is that finding the root cause of a bug (the actual line of code or system interaction that caused the failure) requires holding an enormous amount of context in your head simultaneously. You need to know what changed recently in the codebase, what the error logs say, what the database was doing at that exact moment, and how dozens of different parts of the system interact with each other. Even experienced engineers get it wrong. They fix the symptom, not the cause, and the bug comes back.

What Sonarly does

Sonarly is an AI system that investigates production incidents automatically. When something breaks, Sonarly reads the error, explores the codebase, analyzes the logs, and traces the problem back to its root cause. Then it proposes a targeted code fix for engineers to review.

What makes Sonarly different from simply asking an AI assistant like ChatGPT or Claude is the context it builds before reasoning. Sonarly doesn't just look at the error message. It maps the entire chain: which line of code triggered the error, who changed that line and when, what else changed at the same time, whether similar errors have occurred before, and what the system metrics looked like in the minutes before the failure. It connects dots that a human engineer would take hours to connect manually.

Sonarly also does something unusual: before proposing a fix, it actively tries to disprove its own conclusion. It asks itself, "is there another explanation for this?" and checks its reasoning for gaps. In nearly 1 in 5 cases, this step catches a wrong diagnosis before it becomes a wrong fix, saving engineering teams from implementing a change that would have made things worse.

What the results look like in practice

We tested Sonarly against the best AI coding tools available today on 200 real production bugs from 12 engineering teams. The results were clear: Sonarly correctly identified the root cause in 78% of cases, compared to roughly 53-56% for the best alternatives. On the hardest bugs, the ones involving multiple systems interacting in unexpected ways, the gap was even larger.

More importantly, engineers said they would merge Sonarly's proposed fix without modification in 51.5% of cases. For the alternatives, that number was around 22-24%. That means Sonarly doesn't just find the problem. It gives engineers a solution they can trust and ship.

For a team spending 35% of their time on debugging, this translates directly to more time building, faster incident resolution, and fewer 2am pages that ruin someone's weekend.

Who we are

Sonarly was founded by Alexandre Klobb and Dimitri Choudhury, engineers who spent years watching teams lose hours and companies lose revenue to bugs that should have been solvable in minutes. We are part of Y Combinator's Winter 2026 batch, working with engineering teams across Europe and the United States.

We built Sonarly because we believe that in five years, no engineer should have to spend their Saturday morning staring at log files. The AI should do that. The engineer should be the one who reviews the answer, approves the fix, and ships the improvement. Not the one doing the detective work from scratch.

Production incidents are inevitable. Spending hours investigating them doesn't have to be.

Sonarly is available for engineering teams of 3 or more. If your team deals with production incidents regularly, you can request early access at sonarly.com.

Alexandre Klobb & Dimitri Choudhury, Co-founders at Sonarly