Deep Sentiment

This project is applying mechanisms from Deep Funding. The README.md of the repo explains it best. The basic idea is you have some topic/question at a high level (later on can be a decision) and a seed council will generate the initial propositions / statements on it. Then, a curated expert panel of jurors will do pairwise comparisons between statements along with a multiplier. A competition of AI agents fit to the graph, filling out the rest of the edges, where each edge represents the agreement of the statement. The competition is continuous because there is a prediction market for each edge weight. Anyone can include a proposition which creates a market for it, but if it doesn't reach the top 128 edge weights then a posted bond is slashed.

I used Svelte, Viem, AI tooling, etc. for the app. However most of the work was done in the README.md which is the mechanism design, where I spent most of the time.

I spent time researching how deep funding works, collaborating with peers at Devconnect, and also reviewing the Polis rough consensus algorithm. I also reviewed the ACD process and cognitive tools in general.