Cc @cunkonlife.bsky.social

But my personal challenge here was:
→ Run everything locally
→ On a first-gen Apple M1
→ Using only open-source stuff (Whisper, NLLB…)

No APIs. No cloud. Just a CLI and a mic.
Not because it’s easier.
That’s the only way to meet the constraints and learn something cool on the way.

I would love to!

It is exactly the kind of question I ask. I already have gSuite, and I wonder how protected my data are. (Note the same applies for Microsoft).
I think we both agree that contract don’t always protect us… I wonder about the thickness of the line.

BTW: I think I am “protected” by the contract I have with Google (beside patriot act, which is something, but I am already working with gSuite…).
Are you aware of studies or deep analysis that exposes problems with those contacts ?

This is a very good point. My laptop cannot run a good enough LLM to do this (the transcript works because it is using another model, but the generative part don’t). I rely on my contract with Google and I do not send confidential information.

I made a similar tool recently that is using google’s API for live transcript, and then Gemini to act on the transcription flow.
I also vibe coded it and I fully agree with you. I allows new ideas to pop with a very short feedback loop.

github.com/owulveryck/l...

I genuinely hope you have something more to sayyyyy 😉

Thanks

Is there a recording of this talk?