A new Finnish startup called GitHits has secured €1.5 million in pre-seed funding to tackle one of the most frustrating problems plaguing AI coding agents: hallucinations that produce code looking correct but failing in practice. The company, spun out of Softlandia Venture Studio in late 2025, is launching a beta version of its CLI (command-line interface) product today, with a commercial release planned later this year.

Why Do AI Coding Agents Hallucinate?

Modern software development extends far beyond a single codebase. Developers rely heavily on frameworks, libraries, software development kits (SDKs), and open-source dependencies to build applications. The problem is that today's AI coding agents like Claude Code struggle to inspect these external components effectively. When agents can't find real examples of how to use a dependency, they make educated guesses based on patterns in their training data, producing code that looks plausible but doesn't actually work.

"Coding agents are great at navigating your local codebase. The problem is that modern software doesn't stop at the repository boundary. A large part of the system lives in frameworks, libraries, SDKs, and other open-source dependencies. Agents can't inspect those nearly as well, so AI has to guess, and it produces code that looks correct but doesn't work in practice," explained Olli-Pekka Heinisuo, CTO at GitHits.

Olli-Pekka Heinisuo, CTO at GitHits

Heinisuo brings serious credibility to the problem. He previously developed opencv-python, a software package downloaded over 100 million times that was used by NASA in its Ingenuity helicopter during its Mars missions.

How Does GitHits Solve the Hallucination Problem?

GitHits is building what it calls an "AI-native, version-aware index of all public open-source code." Rather than competing directly with tools like Claude Code, Cursor, or Codex, GitHits complements them by providing a specialized search layer that gives AI agents access to working implementations and dependency-level context from real open-source projects.

The platform offers AI coding agents a toolkit for finding working examples of open-source implementations and inspecting software components, including their dependencies and known vulnerabilities. By grounding AI outputs in actual code rather than model-generated guesses, GitHits helps agents reduce hallucinations, end retry loops, and consume fewer processing tokens.

Steps to Understand GitHits' Competitive Position

• Market Gap: OpenAI, Anthropic, and Google have left a gap in the market by not providing specialized open-source code search for AI agents, creating an opportunity for a focused solution.
• Differentiation Strategy: GitHits focuses exclusively on code search, whereas competitors like Exa (which raised $250 million at a $2.2 billion valuation in May) build general-purpose search for AI agents.
• Team Pedigree: The founding team includes serial entrepreneurs and open-source veterans with deep experience navigating the exact problem they're solving.

The funding round was led by Vendep Capital, with participation from Trind VC and angel investors including Peter Sarlin, Zach Shelby, and Jerry Liu.

"We'd been watching GitHits since it was just an idea, and what convinced us was the team that formed around it. Olli-Pekka is a quiet legend in open source and has lived inside this problem for years. At this stage you invest in people, and this was an easy call," stated Timo Felin, Partner at Vendep Capital.

Timo Felin, Partner at Vendep Capital

What's Next for AI Coding Infrastructure?

GitHits' launch reflects a broader shift in how the AI industry is approaching coding assistance. Rather than relying solely on large language models to generate code from scratch, the next generation of tools is adding specialized layers that ground AI outputs in real-world implementations. This approach reduces the risk of deploying code that looks correct but fails in production, a critical concern for enterprise teams adopting AI coding tools.

The company plans to use its new funding to expand its version-aware index of public code and prepare for its commercial release. As AI coding agents become more prevalent in enterprise development workflows, tools that help them access reliable, real-world examples of working code may become as essential as the models themselves.