Established Only 6 Weeks Ago, Valuation Skyrockets to $4 Billion: Former OpenAI Core Member to Develop "Continual Learning" AI

According to Dynamic Beating monitoring, Core Automation, an AI startup founded by former OpenAI reinforcement learning lead Jerry Tworek, is seeking to raise $3-5 billion at a $40 billion valuation. The company, established just 6 weeks ago, recently closed a $1 billion early-stage funding round at a $10 billion valuation with participation from Nvidia, Spark Capital, and Accel.

Core Automation is attempting to disrupt the conventional large model development path. At a technical level, they are taking a non-consensus approach, consolidating all stages of model training into a single stage and using minimal training data with the goal of developing a new architecture capable of "continual learning" in real time from real-world experience (the model can continuously self-improve even after training is complete).

Tworek has informed investors that the company's initial goal is to develop an AI Agent that can autonomously drive R&D efforts, with a long-term vision that includes establishing a "self-replicating factory" and even developing bio-mechanical organisms. The core team members are largely from top tech companies, including former Anthropic and Google DeepMind researcher Rohan Anil and former OpenAI Model Behavior Lead Joanne Jang.

Such startups, known as "neolabs," rely heavily on Nvidia's funding and computing power endorsement, but they face significant uncertainty in actual product delivery due to skyrocketing GPU rental costs and dual pressures from major players like Meta poaching talent with high salaries.