Generative AI Startup Helps Devs with AI Agents, on OpenAI's Radar

Generative AI has already changed how businesses work and disrupted well established workflows. Expanding the gamut of innovation is a new Israeli Generative AI startup using advanced Large Language Models (LLMs) to transform the software development life cycle. By facilitating the creation of AI agents, Fine allows developers to create their own squad of virtual teammates, and 10x their development speed.

Today’s AI coding assistants landscape is diverse and maturing rapidly, with preferred choices like GitHub’s Copilot, and more recently released OverflowAI. While these assistants boost productivity by 50% or more, they only focus on the next few lines of code. “Fine, on the other hand, provides you with agents to which you can delegate complete software tasks,“ explained Haggai Hofland, Fine’s CEO and Co-Founder.

Fine was founded in 2023 by serial entrepreneur Hofland (previously co-founded Raycatch, acquired), Jonathan Harel (VP R&D), and Dan Leshem (CTO and previously co-founded Plantt, an LLM startup “before it was cool”). Their platform helps software developers build software with AI Agents, pioneering the move to Software 3.0. By offering a robust and secure way to utilize Generative AI and LLMs, they significantly increase the velocity of software engineering.

“Each agent is an expert in a specific software task. You can add your codebase as context and enable each agent to run with extreme accuracy. This improves productivity by 500%,” added Hofland. “Imagine you're considering a framework transition, or pondering the addition of a new feature or API. These tasks, that usually take a few days of work, are done easily with Fine. Simply build an agent and have it execute these tasks. All you need to do is lay out the technical specifications and let the agents take it from there.”

With hundreds of users already leveraging Fine and contributing to an open collection of agents, the emphasis now is on customization and collaboration. New users can fine-tune existing agents to mesh seamlessly with their unique workflows. According to Hofland, Fine is built for developers but comes with enterprise features out of the box. “We are fully committed to privacy - your code is not stored on the platform, so no one has access to sensitive information. Not even us.” said Hofland.

Hofland, Harel and Leshem are also the creators of Dark{mode}, a web series produced in partnership with Wix, Appsflyer, Taboola, AI21 Labs and other leading tech companies. It covered the topic of developer experience and served as a source of inspiration for their foray into Generative AI and software development.

The startup is currently bootstrapped with a team of six developers. The product is offered for free to individual developers, and through a seat-based pricing model for enterprises.

Addressing the skepticism that many developers hold about the accuracy and quality of AI agents today, Hofland argues that the technology is advancing at a record clip. “We now need to build the application layer that will be able to realize its full potential,” he noted. “We’re focused on implementing development tasks, knowing that the models are particularly strong in this field. We’ve built a new workflow that’s adapted to the problem and the technology, and we’re now working on a solution that provides an optimal user experience,” he added.

Hofland is bullish on the future ubiquity of AI agents across various sectors. “It’s clear the world is moving in this direction,” he said. Substantiating his vision, OpenAI recently recognized Fine in a list of promising Generative AI startups for software development and coding.