Interesting, but of no use as I can't replace its OS.

And liquid ass.

Doesn't CC BY require to say what you changed?

If a person would ignore one part of a license, they may also ignore other parts.

... Then why would they care that it's ND?

Not really. You're supposed to attribute the original creator. And if I did make major changes/additions I'd find some way to flag that. But you don't have a Github-style change log.

Not really, but you can't just distribute a changed version as if it were original. That's just as illegal as if it were ND.

It also cannot do tasks it wasn't trained for. It can extend texts, read images and click on a desktop, but only because it's made for that.

I don't think that's strictly true, as I can give it a new gui or tui program it wasn't trained on and it will learn it. Unless you're talking about general abilities like sight, but the same is somewhat true of humans.

If you consider the data on which an LLM was trained on to be points on a very highly multidimensional object, the claim is that the LLM can interpolate a convex hull spanned by those points, therefore recovering a subset of consequences attainable from those points. Obviously this hull includes completely novel points that were not present in the initial data set, so the output of the LLM goes beyond its initial training. And yet, there are clearly points outside a convex hull spanned by any finite number of points, such that we can imagine not all possible outputs are attainable using this method.

The claim is furthermore that truly original thinking, the infamous leaps in understanding and creativity, happen by attaining points outside such a convex hull.

It's hard to rigorously verify or disprove this claim. Hopefully this helps build an intuition of why the claim is not as shallow and obviously wrong as it may seem initially.

I'd call it interpolation on a high dimensional manifold. Convex hull is too simple a shape.

But yes, metaphorically I think that's right.

This makes me realize there is a higher bar we need to achieve with AI still. The ability for the model to evolve through interactions more on a hourly or daily basis. The models are accelerating but inference doesn’t modify the model.

Humans cannot do tasks they are not "trained" for.

speak for yourself

This is false, unless you think the same for humans

Why is there this trent of vibe coded TUIs?

Thought it would be fun to make one.

Some people like to show off.

Why not just use system-ui? What's with the irrational Inter love? Is it because it looks like the Macintosh font?

Maybe it is not listed by the OSI, but it is 'open source' in practice, if the source is also in the public domain. The FSF's definition of free software makes more sense not only ideologically, but also because it does not try to be exhaustive.

One problem is that a large part of Europe doesn't understand the concept of public domain. I'm not legally allowed to not own my creation, except by transferring it to someone else. That's why licences like CC0 exist.

But why exactly are AI agents using the terminal? Can't they use accessibility tools to read GUIs?

Sometimes. Sometimes they can read screeshots, too.

But why waste tokens when bash (etc.) exists and many of the command line toools are configurable to what they emit when used?

I said TUI in the title, not CLI.

And

> and the tooling to build good TUIs finally got good at the same time

Was in the thing you replied to to ask

> But why exactly are AI agents using the terminal? Can't they use accessibility tools to read GUIs?

Which is what I replied to.

Often there is the Inter font. Corners are usually quite rounded, there are lots of gradients. Also, I sometimes see weird descriptions of what a particular UI element does, even though it's obvious. It's possible that the LLM is putting them as a comment to the operator.