Research
Working papers on how a person and an AI build a company together. We test our own methods with data, and publish the results whether or not they flatter us.
Working papers
-
The Day One Great Decision Loop
Does a five-step loop of research, simulation, independent review, an intent check and a report improve the decisions an AI co-founder makes? In our model it cut decision loss by 23.9% at base settings and helped in 90% of 10,000 tested settings. Independent review is the fragile step: it stops helping once a false alarm changes more than about 16% of reviewed game-changing decisions.
Reviewed by independent AI agents in three rounds; not externally peer reviewed. Code and results: research/code/wp1.
What every paper here includes
- The method, with the math written out, and every assumption labeled by the strength of its evidence.
- The code and stored results, so anyone can reproduce every number.
- The limits and any conflict of interest, stated plainly.
- The review it received and how we responded, including what changed.
- A clear statement of whether it has been externally peer reviewed.