Mastodawn

@adrianhon I recall some story where Pokemon Go's thing-to-be-photoed was placed near/on objects since agency wanted to see. So even legit stuff can be misused.

Robin Palotai 5d ago

Lobsters 6d ago

Stealing from Biologists to Compile Haskell Faster

https://iankduncan.com/engineering/2026-05-30-stealing-from-biologists-to-compile-haskell-faster/

#Haskell #Programming #OpenSource

Stealing from Biologists to Compile Haskell Faster - Ian Duncan

The trick that makes GHC’s ApplicativeDo optimization faster is structurally identical to how biologists predict RNA folding.

Ian Duncan

Robin Palotai May 30

muesli May 30

The difference between science fiction and reality is that science fiction has to make sense.

Show thread

Robin Palotai May 30

@knowprose @Yuras I know this was not the question, but a problem with AI for producing output (here, code) is that it doesn't have a vision.

Now when I set out to a task, I have a vague idea in my head. It often turns out to be wrong, but my instincts and vision along the way help me correct it.

If I were to hand off the task to the AI, it would perform it as-is (with ~20% of my instructions botched).

Now, if I could formalize that vision well enough for the AI.. a different question.

Show thread

Robin Palotai May 30

@barubary @benjaminedwardwebb aside, I got there by looking at the foldl package, whose Fold class has a Cosieve instance... clicikng Cosieve says "A Profunctor p is a Cosieve on f if it is a subprofunctor of Costar f." which is something to wrap my head around at a later time...

Show thread

Robin Palotai May 30

@barubary @benjaminedwardwebb

... now the question is, does a Star which has 'Applicative f' as a constraint instead of just 'Functor f' exists and named? Then it could have a class method 'pureStar' defined as 'Star . rmap pure', to achieve the original goal.

Unless #Haskell people know a name for that class, we might call it AppStar or PureStar :)

Show thread

Robin Palotai May 30

@barubary @benjaminedwardwebb
guys, you won't believe this, but I accidentally got nearer to solve this.

I believe things of shape 'a -> f b' are more generally called a Star (if we treat a->b as a Profunctor, and f is a Functor). See https://hackage.haskell.org/package/profunctors-5.6.2/docs/Data-Profunctor-Types.html#t:Star

It literally says "lifting a Functor into a Profunctor". Now, obviously since f only has Functor constraint in Star, we don't have pure (so we need to construct by 'Star . rmap pure . f').

(TBC)

Data.Profunctor.Types

Robin Palotai May 28

If you want performance, or the problem space is sufficiently large/open, you need tradeoffs.

When you have tradeoffs, it bleeds into your interface. Interface and implementation can't go irrespective.

When implementing, you'll keep thinking, will users be able to achieve what they want, and how they need to call this?

When writing interface, think how will the implementation be able to satisfy this?

That's why things are complicated, for me at least, in #programming.

Show thread

Robin Palotai May 27

@johnzajac reality check from Hungary, busy indeed, but will it work and to what extent, a different question. Some of those things are really entrenched. This is only the planning phase, with high hopes. Let's see. (And, well, not literally everything. There is some intersection)

Robin Palotai May 27