Whoops, the first example should be `f{ x: 0, y: 1 }` but you get the idea.

… the following are equivalent:

f{ x = 0, y: 1 }
f{ x: null, y: 1 }
f{ y: 1 }

The evaluator remembers that the field argument `x` has a default value, so it takes care of defaulting `x` if `x` is absent.

However, records or record types in isolation do not support defaults; just record arguments.

Grace supports named arguments for function definitions, but it's still treated as a function of a record.

For example, if you define:

let f{ x = 0, y } = x + y

… then f is inferred to have type:

f : { x: Optional Natural, y: Natural } -> Natural

… and …

Grace solves these issues, although it's not clear if you're talking specifically about Haskell or not

ember is the dostoevsky of lewding with models

my partner @ember.pet

bullying ember with my super long wifi password that she has to enter with a remote

fish want me

my tits

girlfriend snapped a nice picture of my tits

I am here to say that coding agents are not, in fact, good at generating boilerplate

men experiencing midlife crisis will get a nicer car but not nicer tits smh

it's alright!

you're the best choto

ember recommended gemini to me because she said the reasoning was better but honestly i'm just sold on the fact that it doesn't break/hang all the fucking time like chatgpt does; that's good enough for me to switch

ember being egregiously hot on the couch

give me attention

thank you @ars-synthetica.bsky.social

love this

mistress chief

Screenshot review feedback from Bluesky thread (@david-cornelson.bsky.social) Screenshot review feedback from Bluesky thread (@david-cornelson.bsky.social) - review.md

Your comment is noted and responded to (by Claude): gist.github.com/Gabriella439...

this post has a lot of my thoughts on the subject: haskellforall.com/2026/02/beyo...

… but the short version is that I believe that interactions with code can and should be AI-assisted, but not necessarily *AI-mediated* (and the post explains the distinction between the two)

clearly you also firmly believe reading is unnecessary right now

"im not owned! im not owned!!", you continue to insist as your face is shoved down into the mattress and your hands are pinned behind your back

(like your code)

well if claude said it then it must be true

LLMs have definitely never said plausible-sounding things that are wrong and fall apart under scrutiny

I think you misunderstand where I’m coming from. I’m not anti-AI. I literally built a domain-specific programming language for prompt engineering (github.com/Gabriella439...) and I was the lead for prompt engineering R&D at my last job.

I’m specifically down on *agentic coding*, not AI as a whole

I read it. The task they measure in the study is creating ads. You can read the Abstract or Methods sections if you don’t believe me

ember: i would like a little earthquake, as a treat

This study is not about agentic coding, though

> And some of those studies, by MIT, show productivity boosts.

Link?