i wanted to challenge myself on this one because i was just let go last week from eliza labs/ai16z. i had been working there for almost a year and had bottled up ideas I wanted to unleash. this idea seemed like a really good challenge because getting the live calls is really difficult to do, with interruption, realistic voice, and all, and there is a lot of potential to make great content from recorded prank calls. so i had to think thru alot of user problems as well as who would want to consume what type of content and worked back from there.

technically, once i had the concept sketched out, i ran into tons of problems. i decided to use tanstack as ive heard alot about it and hadnt tried it. it turned out to be a good play. not that next js would have been much worse but i got to learn a new framework and new concepts.

i had an old way of connecting twilio<>openai that worked but had really garbage connection quality so i scrapped the whole thing out for a new method and it was really exciting when it worked.

figuring out what models to use was a key component and how to process the audio into a video. i went thru like 3-4 iterations of approaches to processing the audio into video, weighing pros cons of cost, time, code complexity. i like where i landed.

and getting all the payment options to work seamlessly across all chains and doing the fhenix encryption smart contract was really fun and rewarded. all and all a great hackathon.