Working on co-author, an AI tool to help creating content (stream 387)
Summary
In this stream, I experimented live with AI-assisted “vibe coding,” trying to one-shot a .NET Aspire app that generates blog posts from videos and extracts keyframes automatically. I compared tools like OpenCode and Copilot, talked through architecture choices (containers, FFmpeg, UI), and debugged issues in real time with the chat. In the end, I realized a fully one-shot approach wasn’t ideal for a project this complex and decided to pivot back to a more iterative, pair-programming style when using smaller (less performant models).
📺 - Twitch archive - stream no. 387
Replay
Project
All the code for this project is available on GitHub: co-author - https://github.com/FBoucher/co-author
Game Results
- @marcipopsis: 0
- @tbdgamer: 0
- @theunoriginaljerk: 0
- @procrastoholik: 0
- @surlydev: 0
- @stoney_eagle: 0
- @undefined_process: 9.04
- @cmgbeanie: 48.83
- @dj_surly: 54.59
- @phrakberg: 63.20
- @my_friend_asks: 73.16
- @fboucheros: 75.48
- @lurkydev: 84.17
- @thecliptographer: 86.29
- @jtsom: 89.03
- @groversaurus: 89.69
- @codebymistakes: 90.65
- @coppersbeard: 90.89
- @fredda_the_cat: 92.99
- @mcnets: 93.13
- @jeffs_hat_stand: 94.03
- @a_friend_asks: 95.59
- @therealsurlybot: 99.42
Statistics
- 🏆Best score: @therealsurlybot with 99.42
- 😭Biggest loser: @surlydev with 17 drops and no high score
- 🍀Luckiest: @therealsurlybot with best score 99.42 and only 16 drops
- 🎖️Super participant: @surlydev with 17 drops