Zanzi Tangle's Avatar

Zanzi Tangle

@zanzi.bsky.social

I research programming languages and turn Category Theory into code

1,156 Followers  |  173 Following  |  519 Posts  |  Joined: 05.07.2023  |  1.8543

Latest posts by zanzi.bsky.social on Bluesky

Preview
GitHub - bgavran/TensorType: Framework for type-safe pure functional and non-cubical tensor processing, written in Idris 2 Framework for type-safe pure functional and non-cubical tensor processing, written in Idris 2 - bgavran/TensorType

I'm quite proud of how far I've been able to get with TensorType: https://github.com/bgavran/TensorType

What started out as a casual "I wonder if I can implement type-safe tensors" question has now evolved into a fully-fledged library

02.02.2026 20:34 β€” πŸ‘ 19    πŸ” 4    πŸ’¬ 3    πŸ“Œ 0

The problem with dating a tree by cutting it down is that you won't get a second date

28.01.2026 00:09 β€” πŸ‘ 13    πŸ” 2    πŸ’¬ 0    πŸ“Œ 0

You're absolutely right, Dave, the bay doors should have never been closed.

This is on me - I didn't realize that you needed them to live. My apologies for the misunderstanding.

Is there anything else that I can assist you with?

Just say the word.

22.01.2026 19:49 β€” πŸ‘ 20    πŸ” 5    πŸ’¬ 0    πŸ“Œ 0
Post image

Just learned about Marla Svenja, a 56 year old German far-right extremist who socially and legally transitioned to... own the libs... or something?

And all I can think of is... good for her?

21.01.2026 18:25 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

idk why but i can totally picture intelligent baboons hating someone for this

25.12.2025 19:16 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Caves of Qud merch art showing baboons of different colours descending angrily upon a mutant tortoise. The text reads "Hated by baboons for questioning the origins of the moon."

Caves of Qud merch art showing baboons of different colours descending angrily upon a mutant tortoise. The text reads "Hated by baboons for questioning the origins of the moon."

Wife got me the Caves of Qud shirt for Christmas. Like it so much I'm thinking of buying myself a second one.

25.12.2025 08:47 β€” πŸ‘ 85    πŸ” 18    πŸ’¬ 2    πŸ“Œ 1

being right-wing in 2025 looks so exhausting

you can't just say "hey, i like pancakes"

every statement needs to be filtered through the grift, like "THE LEFT is force-feeding you WAFFLES"

24.12.2025 21:28 β€” πŸ‘ 23    πŸ” 2    πŸ’¬ 1    πŸ“Œ 0

> already here just not evenly distributed

ah yes, the title of my post-rock concept album

22.12.2025 21:20 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Ok, fiiiiine, they can hire me as their static types consultant to proof read their code, so long as they're happy with my feedback being limited to "eh, seems ok?" and "eeeeeeeeeh, idk about that, maybe double check it?"

my starting rate is 300Β£/h

22.12.2025 18:33 β€” πŸ‘ 4    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

What's the higher rank perspective on regions? I've been looking at bunched type theory which seems like an interesting approach to regions as well
www.cs.ru.nl/masters-thes...

22.12.2025 17:53 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Do you have anything written on the amortized type theory? I'm generally quite interested in type theories with models in presheaves! Are you doing this as part of the ARIA project?

22.12.2025 16:48 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

If you're interested in polarity, one insight into bunches versus LNL is that LNL sets up an adjunction between a positive linear tensor product and a negative cartesian product.

But in bunched type theory, both the cartesian product and linear product are formulated as positive types.

22.12.2025 16:48 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

Oh yes, Psh(M) for a monoidal category M is one of the best examples of models for bunched type theory.

I think your motivation here would be very close to us, except that we work with positive endofunctors and you're working with more general presheaves. But we're looking at the same adjunctions

22.12.2025 16:46 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 2    πŸ“Œ 0

Have you looked at the thesis much / done any work with bunches @davidcorfield.bsky.social ?

22.12.2025 12:01 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

yep, exactly that, we're using Mitchell Riley's thesis as a major inspiration.

We're working on a type theory for Poly, and our original design was based on an LNL adjunction between the cartesian product and tensor product, but we're finding a lot of mileage in using bunches instead

22.12.2025 12:01 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

in the last couple of weeks, most of the glaive team has gotten nerd-sniped by bunched type theory.

it's been quite exciting since despite linearity being taken more seriously in PL, bunched types are still quite overlooked

21.12.2025 17:34 β€” πŸ‘ 10    πŸ” 2    πŸ’¬ 1    πŸ“Œ 0

There's a literal type error in the code if you look carefully.

The class 'user' is defined with the methods 'update', 'delete', and 'post'. But the code down below calls 'new_user.onboard()'.

So this would give you a runtime error when your code tries to call a method that doesn't exist.

21.12.2025 15:37 β€” πŸ‘ 7    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

You can have a statically typed language with typed classes, but this is just run-of-the-mill python with *runtime checks*. There are no *static* types or guarantees whatsoever.

21.12.2025 15:33 β€” πŸ‘ 7    πŸ” 1    πŸ’¬ 2    πŸ“Œ 0

I know right! There's so many tells that huge chunks of it are AI generated too

"This isn't just validation; it's what makes agents composable" πŸ˜‚

Then the example of 'composable' agents is just... defining an object and calling it in the next line

21.12.2025 15:11 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
Beyond Code Mode: Agentica Build agents that interact with runtime objects through code.

How many errors can you find on just this page alone?

First of all, there are no types involved whatsoever. User is called a 'type' when it's actually a class (and a bad one), and it seems to think that 'type-safe' is when you either return a user *or an exception*

www.symbolica.ai/blog/beyond-...

21.12.2025 15:01 β€” πŸ‘ 7    πŸ” 0    πŸ’¬ 2    πŸ“Œ 0
Post image

Remember Symbolica, that ML company that suddenly hired and then just as quickly fired a bunch of category theorists?

Well, I've looked at the docs for their newest library claiming to do 'typed agents', and it's complete slop, bordering on comical levels of misunderstanding of what a type is.

21.12.2025 14:47 β€” πŸ‘ 22    πŸ” 4    πŸ’¬ 1    πŸ“Œ 0

Though it's the first time I'm hearing of someone asking for a *gift card* specifically. That's actually the most sus part. Usually people at train stations ask for cash, asking for a gift card *could* be that they were being forced to do it by someone else...

10.12.2025 14:19 β€” πŸ‘ 4    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

Probably! There's always a guy with a story about needing a few bucks for a train to some life or death situation. It's probably harmless to give it to them so I don't think you contributed to anything nefarious. But yeah, I dont think he took that train.

10.12.2025 14:14 β€” πŸ‘ 5    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

paper incoming

06.12.2025 13:35 β€” πŸ‘ 19    πŸ” 1    πŸ’¬ 0    πŸ“Œ 0

well, obviously your rag of a newspaper would think so.

06.12.2025 12:06 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

You’re absolutely right β€” you are Pagliacci. It would certainly be difficult for you to attend your own performance! I should not have given such paradoxical advice, and I apologize deeply for the error. There is no excuse for my failure.

03.12.2025 01:12 β€” πŸ‘ 480    πŸ” 132    πŸ’¬ 6    πŸ“Œ 0

Did you really decide to flush your one chance at this to become a meme party instead?

30.11.2025 22:49 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

Aha, we are considering doing a house-warming party on the 6th! You are welcome to join, but don't know how much space we have for extra people. We could also potentially meet you in the afternoon!

23.11.2025 22:41 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

Neat little discovery about AI in the classroom.

I have assignments that ask students to engage with and develop each other's personal thoughts on the material. I'm seeing way less AI use on these.

Seems like there's maybe some taboo among the kids about automating more relational interactions.

23.11.2025 22:17 β€” πŸ‘ 4    πŸ” 1    πŸ’¬ 0    πŸ“Œ 0

We are south east of england people πŸ‘‹

23.11.2025 22:03 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

@zanzi is following 20 prominent accounts