Factory.ai brings Autonomous Software Engineers to Market
the latest SWE-bench leader factory.ai brings up the topic again...
I will be covering the most important projects that hit the SWE-bench and the GAIA benchmark (and leaderboard).
A few months ago, Cognition AI, and the super-human championship coder, Scott Wu burst onto the software engineer automation AI hype scene with Devin.
The ‘AI is gonna take our jobs’ chants were strong that week as people watched the video, then open source tools like OpenDevin and Devika came out mimicking the UI where you could visually see the agent working in a terminal, a browser and an IDE in one window. It is not that these tools are new, it was the visual that made it more real.
Today, there’s a new offering in this market that isn’t getting as much of a splash called Factory. It seems like they have been in stealth mode until 6/18/2024 (2 days ago as of this writing) when they dropped Code Droid: A Technical Report. I have signed up for a demonstration, so I will be reporting back on my findings.
I have seen coverage of their stack through posts from one of the LangChain developer community leaders. And they have now brought Droids to market with SOC2 compliance. Here are the stats they boast on their site:
More general cognitive architectures, whereas right now the technology is not quite there but frameworks like this show promise when they can demonstrate new state of the art levels of quality and throughput of complex tasks.
More to come, I will be covering not just coding agents, but they are definitely one of the main ones that the larger companies are focusing on.