When AI Starts Building AI
Self improving AI is no longer a research curiosity. Kunal Bhatia, CEO and co-founder of Hexo Labs, joins Alan Shimel on why enterprises must own their own intelligent stack. Furthermore, Kunal explains how SIA, the first system of its kind, trains both the weights and the harness of another agent at once.
About Kunal Bhatia
Kunal is on his third AI startup. Consequently, he brings a decade of AI perspective, from an early Alexa style device in 2014 to a general purpose learning app built on GPT-2. Furthermore, that pattern of shipping early and iterating hard now shapes the Hexo Labs roadmap.
How self improving AI actually works
Hexo Labs started as a consulting business with a strong contract research pipeline. As a result, Kunal and his co-founder productized the work into SIA, the first self improving AI system. Meanwhile, they shut the profitable services business down to focus on the platform.
Alan digs in. In addition, the meta agent trains both the harness and the weights of the target agent. Therefore, if the meta agent points at itself, the loop starts to recursively self improve and behaves like a full time researcher who never sleeps.
Why enterprises need self improving AI
Alan and Kunal look at the strategic stakes. Meanwhile, Satya Nadella, Alex Karp and a wave of open weight releases from China all point the same way. Consequently, enterprises must own their intelligent stack rather than rent frontier intelligence from a handful of vendors.
Explore more AI coverage and the latest Techstrong TV interviews. Kunal also unpacks the Shopify for frontier AI vision, in which platforms let merchants own their storefront and compound their advantage. Meanwhile, teams that own their weights, data and production context build a durable moat as the model layer commoditizes. Learn more at hexolabs.com.
