"Appreciate it" / "cool carry on"

Is exactly the sort of confusing non choice options cookie banners give. Along the lines of

"Accept all" / "more information"

Then I'll retract my claim but it is bad UX to make a popup on the landing page that looks like an evil cookie banner.

> XXXX is completely buggy and unusable

Which is amusing because experience shows that XXXX is not buggy and very much usable.

https://code.visualstudio.com/docs/agents/agents-window

Something went wrong with music and culture in recent times. Participation became consumption. Everybody got their own headphones, channels, and separate cultural bubbles. Concerts became about filming a DJ twiddling a USB controller.

By the lake we tried to get people up and dancing, and one of the girls led a reggaeton/zumba/salsa session. I had one woman come up and ask for advice on where to go to get dance lessons. But most people sat there watching, clearly wanting to take part but scared. People have learned that creativity and participation are not welcome.

The most amazing thing was a little 10-year-old girl who just sat herself down in our group of adults. She was so happy to see people singing and dancing. We chatted to her for a while, and then it turned out she could play guitar, so we gave her one and she jammed along. Her mother was observing from a distance and was happy to see her daughter connecting and participating with strangers.

I don't think the issue is between AI and authentic music. This argument about authenticity in music is ages old. It's more about the imbalance in participation between producer and consumer. If AI music allows someone with less formal musical skills to feel like they are joining in and making something, then maybe it has its value.

Still, I'll always be more impressed watching someone play their trained fingers over a piano or guitar. There is more magic in that than prompting an AI. But if the music is just a backing track to some other participatory activity like dancing, then the equation is different again. I honestly couldn't tell — or maybe care — if many of the Bachata songs played at parties are fully or partially AI-generated. I suspect a lot are. But most of the reason I'm there is not to fetishize the authenticity of music, but to hang out with friends and dance and have a good time.

The human / llm needs to have some form of error correction signal. Either you have a corpus of tests or proof system that prevent regressions.

If you have a working system with no tests or validation and let a human loose on it then it will break. How is this different?

Will these new systems remove the middle man skimming that MasterCard/Visa has been doing to small businesses?

> computers handle math just fine with a finite allowance of digits.

Go try and write yourself a robust algorithm to do booleans on polygons or calculate a voronoi diagram. The finite nature of floating point is the mother of all leaky abstractions and bites you in the arse any time you think you are smart enough to roll your own algorithms.

For *most* users, fully distributed and disconnected is a bug not a feature.

1. Training data is the source. 2. Training is compilation/compression. 3. Weights are the compiled source akin to optimized assembly.

However it's an imperfect analogy on so many levels. Nitpick away.