caught covid, had a week to kill, and was annoyed at both nft hype and social media companies. built a decentralized social platform where users can mint posts as nfts. gun js for the database, polygon for the chain, react for the frontend.
the week-long constraint was the point — force decisions fast, ship something real, don't overthink it. most projects die in the planning phase where you spend three weeks deciding between database options and never write a line of code. this one died in the "nobody actually wants to mint their tweets as nfts" phase. but the build was fun and the tech stack held up.
gun js was the right choice for decentralized data — it's a graph database that syncs across peers without a central server. posts live on the user's device and propagate to whoever's subscribed. no single point of failure, no company that can shut you down. the tradeoff is that data consistency is eventual — two users might see different versions of a post for a few seconds before sync catches up. for a social platform that's acceptable. for a banking app it's not.
the polygon integration was the weak link. minting an nft costs gas (even on polygon it's fractions of a cent, still feels wrong for a social post). the user experience of "sign this transaction to publish a thought" killed the spontaneity. social media works because posting is frictionless. adding a wallet signature to every post is like requiring a notary for every text message. technically interesting, practically dead on arrival.
sometimes the best projects are the ones you know won't matter. the tech worked, the concept was sound, the timing was just wrong. nft social platforms still don't work in 2026. maybe they will in 2030. or maybe the idea was always better as a thought experiment than a product. either way, i learned more about decentralized systems in that week than any course could teach.