Swiss startup LogicStar has raised $3 million in pre-seed funding to bring its innovative AI-powered software maintenance tools to the developer market. The startup, founded in summer 2024, is focused on creating autonomous maintenance capabilities for software applications, diverging from the typical AI agent use case of code co-development.

According to LogicStar CEO and co-founder Boris Paskalev, the startup's AI agents could potentially partner with code development agents, such as Cognition Labs' Devin, in a mutually beneficial business arrangement. Paskalev highlights the issue of code fidelity, which affects both human developers and AI agents, and sees an opportunity for LogicStar to improve the development process by automatically detecting and fixing bugs in deployed code.

The startup is building its platform atop large language models (LLMs), taking a model-agnostic approach to maximize its AI agents' utility. This allows LogicStar to leverage different LLMs, such as OpenAI's GPT or China's DeepSeek, to resolve specific code issues. Paskalev believes that the founding team's technical and domain-specific knowledge will enable them to build a platform that can resolve programming problems that challenge or outfox LLMs working alone.

In practice, LogicStar's technology performs an analysis of each application it is deployed on, using classical computer science methods to build a comprehensive "knowledge base" of the software's inputs and outputs, variables, and dependencies. This enables the AI agent to determine which parts of the application are impacted by a bug, allowing it to narrow down the functions needing to be simulated to test potential fixes.

Through its "test-driven development" approach, LogicStar's AI agent can run thousands of tests to identify a "failing test" and ultimately land on a fix that works. While the actual bug fixes are sourced from LLMs, LogicStar's platform enables a "very fast executive environment" that allows its AI agents to work at scale, separating the most effective solutions from the rest.

LogicStar's initial target market will be enterprises, where its "silicon agents" will work alongside corporate dev teams to handle app maintenance tasks, freeing up engineering talent for more creative and challenging work. The startup's pitch touts a "fully autonomous" app maintenance capability, but Paskalev confirms that human developers will still be able to review and oversee the fixes suggested by the AI agents.

The startup is currently testing its alpha version with a number of undisclosed companies, with plans to expand support to Typescript, Javascript, and Java in the near future. The pre-seed funding, led by European VC firm Northzone, will be used to demonstrate the technology's effectiveness with its design partners, focusing on Python.

Michiel Kotting, partner at Northzone, expressed his confidence in LogicStar's potential, stating that "AI-driven code generation is still in its early stages, but the productivity gains we're already seeing are revolutionary. The potential for this technology to streamline development processes, reduce costs, and accelerate innovation is immense."

LogicStar is operating a waiting list for potential customers interested in getting early access, with a beta release planned for later this year. As the startup continues to develop its innovative AI-powered software maintenance tools, it is poised to make a significant impact on the industry.