Superintelligence research lab Hexo Labs today has launched SIA, an open-source agent framework that improves itself for task performance and autonomy. According to an OpenAI benchmark, SIA will accelerate the path to superintelligence by 350X.
SIA is designed to operate in continuous loops of learning and adaptation. According to the company announcement, SIA can create hypotheses, run experiments on those hypotheses, and evaluates the outcomes. It then updates its approach, and repeats the process.
“Today’s AI systems are powerful but share a fundamental limitation: every meaningful leap still depends on intervention of human experts to decide what to try next, interpret results, and refine direction,” said Kunal Bhatia, CEO and co-Founder of Hexo Labs. “But superintelligence will not emerge from static models. SIA learns from itself through execution and compounds its capability with every cycle.”
By releasing SIA as an open-source framework, the company is showing that “the infrastructure for self-improving agents should be transparent, aligned, and accessible to the broader research community.” SIA has already achieved top-tier results on MLE-Bench, a benchmark designed by OpenAI to evaluate the ability of an agent to train a machine learning AI model, Hexo Labs said..
To expand access and accelerate development, Hexo Labs has also introduced the Hexo Labs Grant Program. The program provides researchers with access to SIA, infrastructure credits for large-scale experimentation, and direct collaboration with the Hexo Labs team.
