Core Automation.

Core Automation is building an automated AI lab. The company's stated objective is systems that optimize and automate work, starting with research itself. The first customer is the lab. Automate the work of asking questions, running experiments, reading results, and deciding what to try next, then use that loop to build better systems.

That is a different shape from a normal AI company. Most teams start with a model or an application. Core is starting with the machinery of research: agents, evaluation loops, experiment infrastructure, and the operating model for a small team that wants to do work normally reserved for a much larger organization.

What they're building

The research agenda is aimed below the product surface. Core says it is pursuing new learning algorithms that supersede large-scale pretraining and reinforcement learning, and architectures that scale better than transformers. Those are not incremental claims. They are bets on the substrate of frontier AI: how models learn, how they are organized, and how much human labor it takes to move the frontier forward.

The method is recursive. Core starts by automating its own work. The more of the research loop the system can handle, the more ambitious the team can be. Each automation makes room for harder questions, and each hard question reveals the next thing to automate. If it works, the lab becomes both the product and the proof.

Why we backed the founders and team

We back founders who are willing to attack the bottleneck instead of the visible symptom. In AI, one bottleneck is research labor: the number of good ideas a small team can test, discard, combine, and push through to something real. Core's bet is that highly capable agents can change the throughput of research itself, not just the workflows around it.

Jerry, Rohan, Joanne, Anmol, and Julia have the right kind of ambition for that problem. The work will be quiet for a while. It is experiment harnesses, evaluations, new architectures, and the uncomfortable middle where most results are negative. But if small teams with automated research systems can take on work that once required entire organizations, the leverage is enormous. Core is building for that world.