Filligree

I’m already edging in the article’s direction on my desktop. I recognised a lot of this, and it makes a great deal of sense.

The only question is if it’ll still make sense in six months, or if an even more unreasonable approach will be the correct one. Feels like a singularity.

The programmer adjusted the way it works to make it slightly less OP. Just so we could all be sure the bug was now a feature.

I wish I could be sure this was a joke.

The purpose of a prompt like that is to have it rephrase it in its own words, so you can spot any misunderstandings and correct them.

If you're of the belief that this is a contradiction in terms, then what can I say? It works.

But that wouldn't have put him in prison for life.

Usually it means the opposite. Books that are old and valuable tend to be out of copyright, so they do see new printing runs.

They are. They can’t release it; that would be copyright infringement. But they’re absolutely planning to make further use of the book later.

Obviously. Copyright infringement is settled law.

B is illegal, and Anthropic ate a billion dollar fine for trying it, so you can’t even claim they don’t want to.

Because that’s illegal.

I use urllib.request. It works fine, it’s always available, and I don’t need to push it through an audit.

They definitely are successors to some of those. Two Point Hospital for one.

It never quite recaptured the charm of the original, but that might just be because I’m older.

I am not, but would be interested to hear. Any articles?

The authors of the alternative books, which you might have bought instead if your weren’t reading old books like a high seas pirate.

That keeping the physical copy intact would mean creating a copy, and they don’t have the copy right.

That is an element, but given the PR cost I’m sure they would use non-destructive scanning if they could.

No, the core problem is that keeping a scan is legal if and only if you destroy the physical copy.

For watermarked long-form text, it is actually possible. Makes the watermark more fragile, but the math is considerably more forgiving than usual.

The word means what it does. I’m not changing my use of language just because some people misunderstand the dictionary definition.

Isn’t it MIT-licensed?

That’s not theft. The author explicitly told people he wants them to do this; the company is just taking him up on the offer.

> The strange part is that nearly all the decently large publishers seem to be completely unable to predict when a game will be successful, because they seem unable to determine when a game is actually fun to play. So they will dump hundreds of millions into a game that nobody wants to play, get close to zero sales, and end up shutting down the game, laying off the studio, and having to find another way to make back the $100m+ they just burned.

I know a guy who works with one of those publishers.

According to him, almost nobody at the publisher actually games. The ones who do, limit themselves to gacha games or Bejeweled-tier games. Gaming, at least at that one, is considered childish.

Haven’t ever seen a warning like that in KDE.

I want to learn if it would one day be possible to engineer something like that. Unfortunately I think that even the theory is post-singularity territory, so... maybe in the 2030s?

Their stock price won’t matter once OpenAI launches the paperclippers.

Not everybody uses squash and merge. Regular merge, I imagine, would work fine.

Read the tests.

Doing a pass where you just ask the AI to sanity-check the existing tests (against rules like “test against the spec, not the implementation” can also help.

Will, it depends on the AI. Anthropic used to have a lecture-length system prompt for their models to explain this stuff—part of the secret sauce for Claude Code—and famously found that the 5 series models no longer need it.

As usual, if you use anything but the best model available I’m going to state that the better ones do better. If you do use the best model available, then I’ll just mention that Fable still has limits and still needs some guidance.

One thing it does not do is deliberately build tests which test nothing at all, or which restate the code under test. I mention this because certain other models absolutely would.

Using a third-party launcher (e.g. Prism) mostly fixes that nonsense, at least.

Less efficient but faster, which is a fun combination. At least they seem to have improved FPS.

It’s not about the LLM being modern or not. 3.5 Flash is fairly new, but it’s also a flash model. It’s not designed to be knowledgeable.

People keep doing this. Pointing at the known limitations of cheap/fast LLMs and pretending they’re universal is not, in fact, valid reasoning.

Yeah, we’ve seen that as well.