We love laziness ! Thx ! haha

Good course ! But if you're still sitting through your courses at 1x speed, stop. Here's the method I use. I turn any training video into a clean text transcript, so I read a 45 min module in 5 minutes and keep it searchable forever. Three routes, cheapest first: 1. Grab the existing subtitles with yt-dlp. Takes 3 seconds per video, no download needed. One catch: use --sub-langs all, never fr. Loom tags the track as "en" even when the audio is French, and filtering on fr silently returns nothing. Subtitles give you the speech but never the screen, so this is perfect for talking-head videos and useless for screen demos. 2. NotebookLM for volume, free. One notebook per module, drop in your mp4s, pull the verbatim into a file. Solid up to roughly 6500 words per video. Two limits: no mm:ss timestamps, and silent failures (the source shows "ready" but the extraction comes back empty), so check every single output. Skill : https://github.com/teng-lin/notebooklm-py 3. Gemini API when you need precision. Real timestamps and an accurate readout of what's on screen. This is the one for tutorials and software demos. Free on the AI Studio tier, the only ceiling is the daily request limit. The setting that matters most on Gemini: media_resolution = LOW (cuts token usage by 3). Skip the second one and a 20 min video gets truncated with zero error message. The one rule I apply everywhere: always confirm the transcript actually reaches the end of the video. Truncation never announces itself.

@Daniel Agrici Gemini, on the free AI Studio tier. Specifically gemini-3.6-flash as the default, with gemini-3.1-flash-lite and gemini-2.5-flash as fallbacks when I hit a 503 or a daily cap. One thing worth knowing: rate limits are per model, not shared across the whole API. So when gemini-3.6-flash hits its daily request cap, switching to gemini-2.5-flash gives you a fresh quota and you keep working. That fallback chain is the reason I can transcribe a full module in one sitting on a free key. Two caveats though. Limits apply per project, not per API key, so making extra keys inside the same Google Cloud project buys you nothing. And Google's model catalog changes almost every month, so I list the available models before each batch instead of hardcoding a name in the script. Models I used a month ago have already been replaced by newer flash versions.

@Daniel Agrici I think it is helpful : A skill to stop your coding agent from burying the answer. ADHD-friendly output = https://github.com/ayghri/i-have-adhd

@Daniel Agrici Fun but helpful ! haha

A gauntlet-loop for plan ?

@Daniel Agrici Great for claude-seo, claude-blog and claude-ads !