Offloop claims a SOTA benchmark result on GDPval (which models $2.4 trillion/year of US paid work) with a 4-person team

A four-person startup called Offloop just posted what they say is a state-of-the-art result on GDPval, a benchmark that tries to measure how well AI handles tasks representing $2.4 trillion worth of paid work in the US economy each year. That’s a big claim from a tiny team, and the benchmark itself is new enough that we should treat the number with caution. But the positioning is what matters here.

Offloop isn’t selling another chatbot or another coding assistant. They’re pitching themselves as the system that “makes teams whole” by running the recurring operational work nobody owns: follow-ups, feedback loops, customer updates, the kind of stuff that falls through the cracks when everyone’s busy. Their post frames the problem as work that’s essential but orphaned, and their solution as an agent that just handles it.

This is almost word for word the thesis behind what we’ve been building into the Omni Command Centre: autonomous systems that take ownership of the repetitive, cross-functional work that bogs down small teams. The difference between a tool and a command centre is whether it waits for you to remember to use it or whether it sees the work and does it.

Whether Offloop’s benchmark score holds up under scrutiny or not, their framing is sharp. They’re not talking about productivity gains or efficiency. They’re talking about covering the work that doesn’t get done because no one has time or it’s not clearly anyone’s job. That’s the real bottleneck in most businesses under 50 people. If a four-person team can build something that credibly attacks that problem, the question isn’t whether the technology works. It’s how fast you can get it running in your business before your competitors do.