bonsai

A premium grocery retailer wondering if it should enter a new market. A brand wondering if it should proceed with a bold advertisement – would be lauded for its creativity or lambasted for its insensitivities. An economist wondering under what conditions algorithmic collusion between LLMs might be possible. If your problem can be reframed as a turn-based game, Hypothesis is for you. Built as an alternative to traditional methods of gathering market feedback or conducting economic research (which are often time-consuming and costly), Hypothesis is fundamentally a custom simulator – each simulation is painstakingly curated by Muse Code and powered by Muse Spark 1.3 agents. To run an experiment, the user only needs to know how the experiment should be run – specifically, what does one turn/round look like? The LLM handles everything else, from clarifying the experiment, to building and executing it. Users can look forward to a custom dashboard containing experiment results, charts, and a downloadable Python Notebook should one want to replicate this experiment locally (to be run within the GitHub repo). Works across a variety of domains: Microeconomics, Advertising, Operations, and more.