Author here. I've been encoding video for about 25 years, and FFmpeg has been part of most of that. Started at M6 Web in France, 4 servers in a basement producing HLS before hardware encoders had it.
It's partly about the game, partly an homage to people like Fabio Sonnati (who documented FFmpeg for 15 years on a WordPress blog), Jan Ozer (whose book is literally called "Eliminate the Guesswork from your Streaming Video", and I built a game that turns encoding back into guesswork), and Fabrice Bellard who started it all.
I emailed FFmpeg to ask permission. No response yet. Which tracks, they barely have time to patch bugs found by Google's AI.
I've been writing a weekly streaming industry newsletter for 68 editions. Most of my time wasn't writing — it was information logistics: tracking sources, searching my own archive, remembering which expert I'd cited and when.
Bookmarking tools keep dying (RIP Pocket). If you write a newsletter, your sources are your most valuable asset and they live in tools you don't control. I built an MCP server to own that layer — sources, notes, briefs, article tracking. SQLite-backed, your data stays yours.
The design principle: the MCP handles everything around the writing (collecting, organizing, retrieving). It never generates content.
Caveat: I daily-drive a more advanced version (Supabase, custom schema, 50 tools) to build two paid industry reports (20K+ datapoints, 172 companies, 34 countries). inkwell-mcp is the extracted portable core. It type-checks, but I haven't battle-tested it with anyone else's workflow yet. Issues welcome.
25 years in streaming, built M6 Replay (50M+ users), several CTO roles.
Full write-up: Full write-up: https://streamingradar.substack.com/p/my-newsletter-has-50-a...
The newsletter: https://www.streaming-radar.com
The full MCP with 20K+ datapoints (bidirectional, 4 slots): https://lens.streaming-radar.com/mcp
Happy to answer questions about the architecture or MCP design decisions.
We knew better the sub-saharan countries for instance, so our network and focus is to go there first. We have a lot of work to have the best catalog from there first, and then we could extend.
I wrote the full story behind this: https://www.streaming-radar.com/p/i-built-a-wordle-for-ffmpe...
It's partly about the game, partly an homage to people like Fabio Sonnati (who documented FFmpeg for 15 years on a WordPress blog), Jan Ozer (whose book is literally called "Eliminate the Guesswork from your Streaming Video", and I built a game that turns encoding back into guesswork), and Fabrice Bellard who started it all.
I emailed FFmpeg to ask permission. No response yet. Which tracks, they barely have time to patch bugs found by Google's AI.
CLI in Rust: https://github.com/ludobos/ffmpuzzle-cli
I also wrote about the invisible French engineers behind streaming (Bellard, VLC, GPAC): https://www.streaming-radar.com/p/les-streamers-de-lombre-30...
I also wrote a piece on the invisible French engineers behind streaming (Bellard, VLC, GPAC): https://www.streaming-radar.com/p/les-streamers-de-lombre-30...