Cosine, a UK-based AI startup, has raised $2.5 million in seed funding and unveiled what they are declaring to be the "World's best AI Software Engineer". And they are justified in that claim! Their AI model, called Genie, has scored 30.08% on SWE-Bench, the industry standard for evaluating software engineering skills in AI models, setting a new state-of-the-art benchmark.
This score represents a significant leap forward, outperforming the previous best of 19.27% held by Factory Code Droid. Genie's achievement dwarfs other well-known AI models, including Devin scored which scored 13.8% and OpenAI's GPT-4, which scored 12.47% on the same benchmark.
Cosine's approach to developing Genie focuses on emulating human reasoning in software engineering. The company trained Genie on proprietary data that codifies human problem-solving processes, derived from real-world examples of software engineers at work.
Alistair Pullen, CEO of Cosine, explained: "Our breakthrough in codifying human reasoning allows us to train AI models to operate far beyond the narrow range of tasks currently available to software development teams."
Genie can solve bugs, build features, and refactor code either autonomously or collaboratively with human developers. The AI also integrates with GitHub, streamlining workflow by directly importing issues and creating detailed work specifications. This feature, combined with its precise file identification and in-place editing capabilities, makes Genie an invaluable asset for development teams looking to streamline their workflows and enhance productivity.
The $2.5 million seed round was led by SOMA and Uphonest Capital, with participation from Lakestar and Focal. Ellen Ma, Partner at Uphonest Capital, noted: "Cosine is fundamentally teaching AI to reason, providing companies with a true AI colleague."
Founded in 2022 and backed by Y Combinator, Cosine operates from San Francisco and London. The company plans to expand Genie's capabilities to cover more programming languages and frameworks, exploring both smaller models for simpler tasks and larger models for complex challenges.
As AI software engineering rapidly evolves, Cosine's Genie sets a new benchmark for the industry. However, with ongoing investments and developments in this field, further advancements and competition are likely to emerge.