In a year filled with constant announcements regarding developments in enterprise software and generative AI, there exists an often-overlooked aspect of software development that rarely makes headlines: maintenance.
The reality is that virtually every newly created and released software requires ongoing vigilance and manual updates to ensure it remains secure, functional, and efficient, while also avoiding the accumulation of “technical debt” over time.
Traditionally, these tasks have fallen to developers within software organizations. However, a fresh startup called Grit believes it has found a superior solution: a generative AI-powered developer assistant. Today, the company is unveiling its open beta, accompanied by a $7 million funding round led by Peter Thiel’s Founders Fund and Abstract Ventures, with backing from Quiet Capital, 8VC, A* Capital, AME Cloud Ventures, SV Angel, Operator Partners, CoFound Partners, and Uncorrelated Ventures.
Grit’s eponymous AI tool automatically scrutinizes a program’s codebase, monitors it over time, and suggests updates and enhancements, almost as if it were a member of the development team.
Morgante Pell, CEO, and co-founder of Grit, explained the innovation: “Traditionally, software engineering was very artisanal. You have experts who go in with scalpels and modify code line-by-line. But what we’re seeing with generative AI is it’s much easier to generate new code. So we just need new tools that are able to move code at scale, like bulldozers.”
So, how does this technology function?
Morgante Pell is intimately familiar with the challenges of maintaining software, having previously worked on infrastructure for Google Cloud.
He was struck by the realization that many of their customers, not being tech companies themselves, still relied heavily on software crucial to their businesses, necessitating constant maintenance. This sparked the idea that if maintenance could be automated, it could tackle the unglamorous work engineers are often burdened with, allowing software to run smoothly.
“Right now, engineers are interrupting their more interesting work to do maintenance,” Pell stated. “It’s the work nobody wants to be doing in the first place.”
With the permission of a CTO or authorized developer, Grit can be installed as a GitHub app or linked to GitLab, scanning a company’s code repository and building an index.
Pell clarified that the index is “stored ephemerally,” and a highly optimized search tool is used to understand where changes are needed in the application codebase, based on predefined objectives.
Grit’s signature app utilizes a natural language query interface, allowing developers to express their high-level goals while Grit handles the nitty-gritty implementation details.
Grit not only automates changes but also presents proposed changes to developers or developer teams, seeking approval. If developers wish to modify the proposed changes, they can simply communicate with Grit in natural language, treating it as another member of the team.
Pell explained, “Instead of an engineer having to go in and proactively make a change, Grit can just look and say, ‘OK, you’re out of date on this version, and we’re going to suggest the upgrade and we’ve already generated the change to do the upgrade.’ So the engineer, all they have to do is click one button and say ‘approved.’ They don’t even have to open their editor to do their change.”
Early results suggest Grit’s promise. Although only a year old, Grit has already saved considerable time for customers like Faire and PromptLayer. Projects that were estimated to take six months of engineering effort were completed in just a week with Grit’s assistance.
Grit’s primary customer base currently comprises later-stage technology companies and some fintech firms. The tool is particularly beneficial for modernizing old codebases and is ideal for teams of hundreds of engineers looking to offload maintenance tasks efficiently.
Notably, Grit’s angel investors include Guillermo Rauch from Vercel, Scott Belsky from Adobe, and entrepreneur Sahil Bloom.