1 hour ago
1
A pair of researchers who spent years inside two of the world’s most powerful AI labs have stepped out to build something of their own. Andrew Dai, who spent 14 years at Google DeepMind, and Yinfei Yang, a former chief research scientist at Apple, have co-founded a new AI startup focused on visual reasoning and continuous learning, the kind of feedback-loop-driven development that could reshape how AI products evolve in real time.
The company is reportedly targeting around $50 million in seed funding, with that figure potentially climbing to $55 million as discussions progress. Striker Venture Partners is reported to be leading the round.
What they’re actually building
The core thesis here is deceptively simple: most AI models today are terrible at understanding the visual world. Dai has compared existing big-lab AI models to a 3-year-old child when it comes to processing visual prompts.
The startup’s approach leans heavily on rapid iteration cycles, borrowing from the vibe-coding methodology that has gained traction among developers building AI products at speed. Instead of training a massive model once and deploying it, the idea is to create systems that learn continuously from real-world feedback, getting smarter with every interaction rather than waiting for the next version update.
Seth Neel, a former Harvard academic, rounds out the founding team. The founders transitioned from their previous roles around late 2025 to early 2026, with initial funding discussions and public coverage occurring mainly between January and April 2026.
The company is based in Palo Alto. By focusing on visual reasoning and tangible environments, the founders are positioning themselves in the robotics and autonomous systems space.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) · 