Laredo wants to use gen AI to automate dev work

Developers are embracing AI — some surveys suggests, at least. The Q&A site Stack Overflow polled devs on their attitudes toward code-generating AI tools and found that the vast majority were quite positive. Seventy-seven percent of developers feel favorably about using AI in their workflows, per the survey — citing benefits like increased productivity and faster learning.

GitHub Copilot appears to be gaining traction where it concerns AI-powered tools, and to a lesser extent Amazon CodeWhisperer. But in spite of tech giants’ efforts to corner the nascent space, entrepreneurs are launching their own takes.

Take for example Laredo Labs, a startup developing an AI-driven platform for code generation. Leveraging an AI model trained on data from around a hundred million software projects, Laredo writes code based on high-level natural language commands — writing, editing and deleting code to accomplish dev tasks while documenting progress.

Laredo was co-founded in 2022 by Mark Gabel and Daniel Lord. Gabel was previously the chief scientist at Viv Labs, which Samsung acquired in 2016 to beef up its Bixby voice assistant. Lord, meanwhile, was a platform engineer at Siri prior to Apple’s purchase of the startup.

“We’ve always cared a great deal about our craft and have always strived to make better software, faster,” Gabel told TechCrunch in an email interview. “My background in AI-driven software engineering — and the sudden increase in AI scale — created a unique opportunity to make a massive leap in software development tooling.”

And that’s how Laredo came about. Gabel and Lord built their own models, user experience and — to train those models — what they claim is one of the most comprehensive software engineering data sets in existence. Currently available in private preview, Laredo’s platform can complete “repository-level” tasks from instructions or even text taken verbatim from an issue tracker, Gabel claims.

“Laredo is a ‘full stack’ machine learning company,” Gabel said. “We’re introducing an ambitious new AI-driven developer experience.”

Now, generative coding tools — like all tools underpinned by generative AI — can be legally risky.

Microsoft, GitHub and OpenAI are currently being sued in a class action lawsuit that accuses them of violating IP law by letting Copilot, which was trained on billions of examples of public code from the web, some under a restrictive license, regurgitate sections of copyrighted code without providing credit. Liability aside, some legal experts have suggested that AI like Copilot could put companies at risk if they were to unwittingly incorporate copyrighted suggestions from the tool into their production software.

It’s not clear if Laredo’s models were trained on copyrighted code or what the startup’s indemnification policy might be in the event a customer’s sued over IP regurgitated by Laredo’s models. I’ve asked Gabel for clarification and will update this piece once I hear back.

Legal hurdles aside, Laredo is entering an increasingly competitive field, as alluded to earlier. One recent new entrant, Sweep, aims to automate basic dev tasks very similar to the way Laredo appears to be doing it — using generative models trained on large data sets of coding projects.

But Gabel thinks that Laredo has a fighting chance.

“The software engineering space is an enormous market, with room for many participants, and Laredo … currently occupies its own niche,” he said.

Laredo is pre-revenue. But it’s raised $8.5 million in a seed round co-led by Radical Ventures and Horizons Ventures. A portion of the new capital will go toward hiring — expanding Laredo’s team from eight people to ten by the end of the year.

source

Leave a Reply

Your email address will not be published. Required fields are marked *