Hacker News: jjaskulowski

New comment by jjaskulowski in "Mercury 3: Diffusion at scale"

jjaskulowski — Mon, 19 Oct 2026 18:36:22 +0000

If that's true, that's the real answer, or more like anty-answer because you're saying basically: hey, nothing should stop them.

New comment by jjaskulowski in "Mercury 3: Diffusion at scale"

jjaskulowski — Mon, 19 Oct 2026 18:34:51 +0000

Any comparision would make sense only if you made them - Mercury and Luna - the same training data and then the same post process, alignment. It's nice it's fast. I support it. Just saying I don't know how to judge that.

New comment by jjaskulowski in "[dead]"

jjaskulowski — Sun, 11 Oct 2026 20:33:01 +0000

I'm done with documentation! I'm done with recording demos! I decided that my app will be self documenting just like my code.

If I can ask agent to create an e2e test set, I can ask it to record hundreds of demo recordings and embed them everywhere in the app, wherever they're needed.

Look:

https://pub-1ba258680d824481855fc86b1ecbafe4.r2.dev/demos/ho...

https://pub-1ba258680d824481855fc86b1ecbafe4.r2.dev/demos/bu...

https://pub-1ba258680d824481855fc86b1ecbafe4.r2.dev/demos/se...

They have music, they have slides, live demos, narration. They are faster to produce than any studio can do. They will be everywhere in the app.

I was thinking to make this thing a product of itself but then again, programmers don't like to pay and it was so easy to do with AI.

The best part. The voice is poor but the moment I find a better one I will just rerender. The moment my app changes, I will just rerender. The momement something stops working the same way in the app it will just not rerender and I will now.

Try it! 10 / 10 worth it.