Strange Lab

We get your company ready for Agents.

We turn the work record into a canonical event spine, find the workflows agents should learn, package the safe ones as governed skills, and check every rollout against what happens next.

The first output is an Agent Deployment Map: what agents can train on, what they can shadow, what they can run, and where a human gate stays in the loop. The world model sits underneath that map. Pick a point in the company's history, see only what was known by that day, test a different decision, and compare the branch with the later record.

You can try two archival replay environments now. One is for Enron, everyone's favourite accounting scandal. Using internal email, market, and news data, we can test real forks inside the company's history. We can look at the PG&E power deal and ask whether Enron should have held for a credit recheck or pushed it through. Or we can look at the California crisis, the day a preservation order lands on the trading desk, and test halting the strategy. Or ask what would have happened if Jeff Skilling had shown the books a year earlier.

We can also do this with a public-history world model built from current and analog U.S. macro records — inflation, jobs, rates, GDP, Treasury yields, and public releases. You can pick a cutoff, write a branch, and compare the LLM's evidence-only advice with the world model's consequence forecast.

Agents need more than context. They need company state, task boundaries, policy gates, and a score. That is the job of the work record.

Essays, Papers, and Code