Average Message

People can be louder about what they think through the proof of humanity they have beneath their messages. At times, people in certain regions shout certain common messages, and/or have common grounds in their voices that might not be clear in first glance. Our solution makes is possible for user to see what individuals or groups of individuals that live in certain regions think. This would be especially meaningful during times of hardship, political discourse, or war. Where there is misinformation, we could strengthen democratic opinions by giving real human (non-bot) voices a new, simple platform.

On the "blockscout" branch, we also have blockscout integration which handles the case where messages would be handled on-chain and explored through blockscout (rather than using a conventional postgres database). Its experimental, but we did our best and believe it qualifies for the pool

We create embeddings of messages sent by people (in certain regions), and then calculate the average embedding. Then, we LLM-generate a certain amount of sample 'average candidates' , and then accept the cosine-closest thereof as the `` decoding` of aforementioned average.

We make use of worldcoin as a means of proof-of-humanity, and also for pseudonymization which enables constraining each user to e.g. 3 messages per day..

We have a flask backend hosted through hetzner (server in EU, safer), using IAC. We also have NextJS frontend hosted through github pages.

All work was done within the hackathon