Lightrun, a Tel Aviv-based startup that helps developers debug their production code from within their IDE, on Wednesday announced the launch of its first AI-based tool: the Runtime Autonomous AI Debugger. The new tool, which is currently in private beta, aims to help developers fix issues with their production code within minutes instead of hours.
In addition, Lightrun also on Wednesday disclosed an $18 million SAFE round it raised last year from GTM Capital, with existing investors Insight Partners and Glilot Capital also participating. This brings Lightrun’s total funding to date to $45 million. It’s our understanding that the company plans to raise a Series B round next year.
“Until now, we slashed [Meant Time to Recovery] to say 30 minutes, maybe 45 minutes on average, based on how we measure ourselves and customer feedback,” Lightrun CEO and co-founder Ilan Peleg told me. “Now, we’re going to automate everything from the moment you have a ticket that was raised, up until finding the root cause down to a single level of granularity, like which of your single lines of code is responsible for this very specific root cause.”
Over time, Peleg said, Lightrun would like to extend this to also using generative AI to fix bugs automatically. For now, though, that’s not yet an option, but given how quickly the technology has advanced, it’s probably just a matter of time.
To do this, Lightrun is fine-tuning existing models to focus on debugging, something the company can do in part because it gets insights not only from the code itself but also the entire monitoring and observability stack. Looking ahead, the company also plans to connect this system to other enterprise inputs like ticketing systems. “There’s so much data in the enterprise landscape that is somehow related to troubleshooting or debugging — and that’s missing from the Copilot-like solutions,” Peleg said. Most of the Copilot-like chat interfaces, he argued, only look at the code but don’t have enough insights into the context to present the best solutions.
As Peleg noted, the team went through quite a few iterations before it felt like its system was ready for day-to-day use. About half a year ago, Lightrun started experimenting with existing models to see where generative AI could help its users. But at the time, the solution was far too expensive to offer as a product. “Now we have tuned our system […] so that it’s not going to add significant cost to us for the solution, which is why we’re talking now. In the past, I didn’t feel comfortable to announce something that was not yet there.”
At least for now, these generative AI features will simply be part of the existing Lightrun solution for the users in the private beta. Peleg stressed that the company wants to prove that the system does indeed bring value to users and isn’t trying to optimize for monetization in the short term.