I published my first engineering case study today. It breaks down how I built the AI clipping system behind ByteCap’s Magic Clips. What looks like “upload a long video and get short clips” ended up being a pretty interesting engineering problem. I wrote about how I approached: • finding and ranking moments across long videos • combining transcripts, audio, scene changes and visual signals • face and person tracking for automatic 9:16 reframing • moving computer vision workloads from Google Video Intelligence to self hosted infrastructure as costs grew • handling FFmpeg jobs, concurrency, cancellation and failures in production • turning the final result into an editable video project I’ve spent years building software, but I haven’t done a great job of writing about the engineering behind it. I’m going to change that and publish more of these. First one is here alimansour.com/engineering/a…

New senior dev joined the Bytecap team. Already took my chair.

Nothing kills trust faster than hiding the cancel button behind support emails or retention flows. At Bytecap, it’s one click. If you want to leave, you should be able to leave.

You can now import videos directly from TikTok, Instagram, Facebook, X (Twitter), and Twitch. Just paste a video link to automatically generate AI clips, captions, and ready to post short form content in minutes. bytecap.io

Don’t let one platform decide whether your content is good. Exact same videos. Facebook Reels: 200k+ views TikTok: under 100 Post everywhere before judging your content.