@aramh @gabrielmfern Honestly I think perfection in a PL is fundementally impossible
English
Noam Y
11.6K posts
@VictorTaelin I think one the reason they do it is they want to lock clients in. They don't want ppl switching between models whenever another model is slightly better
English
Deleted again because misinformation 🥲
Gemini 3.5 Flash *is* available on the API. Yet, both the API and the CLI versions are 3x slower than on the IDE! See the video below.
→ Antigravity IDE: 4 seconds (smooth)
→ Antigravity CLI: 15 seconds (buggy)
So the point holds: they want you to use the visual IDE.
Problem is: it is 2026. NOBODY should be using IDEs anymore. Get over it. Let it GO. I’m certainly not launching a VSCode fork to use a model, no matter how great it is.
They invent a portal gun, only to lock it behind a taxi subscription, because they completely fail to realize their very product deprecates that other thing they think will make them money?
Cursor is a great example of a company that (sadly) is very likely fail because of that mindset. Composer is actually surprisingly good model. They should put all efforts in serving it. Yet, they keep locking it under an old school product that nobody wants to use.
And even these who DO use IDEs probably won’t necessarily pick YOUR IDE. And they shouldn’t. You do NOT need them to, to make money.
Your model is the product.
You keep chasing old business models.
Completely out of touch.
Meanwhile Anthropic is all charging at full speed to sooner or later surpass Google by just serving great models under an API
/ctrlv
Taelin@VictorTaelin
The new Gemini 3.5 Flash solved the HVM3's wnf bug in 1/3 attempts. This is my main test to take a model seriously. So far only the big models like GPT 5.5 solved it. And seems like it is 20x faster than Opus 4.6 ! Promising but Google will still find a way to fuck up
English
אני: היי שלום אני רוצה לעזוב את השירות שלכם, הוא יקר לי ולא מתאים לצרכים שלי.
חברה כלשהי, לא חשוב שמות: זה כי את לא יודעת מה טוב לך, ואני אסביר,,
עברית
Noam Y retweetledi
מרצים יקרים, הפסקה של רבע שעה היא לא פריווילגיה של סטודנטים עצלים שרוצים רק לנוח. היא לאיתים קרובות הזמן הנדרש להגיע מכיתה לכיתה, לא כולל עצירה בשירותים.
לכן לא, זה לא בסדר לגנוב חמש דקות מסוף השיעור כדי לסיים מהלך. זה גורם לנו לאחר להרצאה הבאה, ולהגיע לשם בלי נשימה.
עברית
@noam_yy It is very similar. It's basically a let that allows contraction.
English
In some sense `:` is the equivalent to λ, but it works backwards, because the scope is outside the expression body.
In λ-calculus the variable is usable inside the body: `λx x y` — you can use x inside the body, not outside.
But in CUE, the variable is usable outside the body: `x: {a: 1, b: 2}, y: x` — x appears outside the body of x (in the body of y) and it doesn't appear inside.
This is kinda like scopeless lambdas in HVM.
Aram Hăvărneanu@aramh
In λ-calculus you have to distinguish from abstraction and application. Traditionally abstraction uses a binder symbol, λ, and application uses juxtaposition, no symbol at all. CUE doesn't have anything resembling λ which might suggest it's only a first order language. But this is an illusion. In CUE, the λ (which is really more like a coλ) is implicit. It's everywhere you need it to be. Everything is generalized by default. But it is the equivalent of application which needs an explicit symbol, in CUE's case &, unification (or conjunction). & is not exactly like application and implicit generalization is not exactly like λ-abstraction, but it's close enough in what matters—you only need one symbol to distinguish between two things (and actually in CUE's case you can often dispense with & too, as there is an implicit & between all bindings). The order of a CUE program (in the sense of "how high order it is") is not an integer. Scopes don't nest into trees, they form a graph. The order is a multidimensional quantity. There is only a partial order between orders.
English
@VictorTaelin you should probably give it a different metric, like measure the AST size
English
@VictorTaelin It'd be hillarious if once you publish a harder benchmark , labs would just train on it. which may have been what you planned all along..
English
Introducing LamBench . . .
You asked me to make a benchmark, so I made it. It is a simple, old style Q&A consisting of 120 fresh λ-calculus programming questions. Some are easy, like "implement add for λ-encoded nats". Some are harder, like "derive a generic fold for arbitrary λ-encodings".
It measures:
- intelligence (% tasks completed)
- elegance (BLC-length of solutions)
- speed (completion time)
Basically what I care about, other than long context.
I made it today because I was excited about GPT 5.5.
It didn't do too well ):
(My first-day impression is that I can't tell the difference between GPT 5.5 and GPT 5.4. I would be lying if I said otherwise. I'd not be able to distinguish in a blind test. I need more time. It is much faster though.)
This is a new, simple bench, so expect be bugs.
Specially on OpenRouter models. I'll retest soon.
Also, it was born saturated. V2 will be harder...
↓ Link and more charts below ↓
English
@NitzhiaP אני חייבת לדעת אם זו בדיחה פנימית או מישהו שקוראים לו עזר וייצמן או הנשיא לשעבר
עברית
@DerJellyBelly איכ. סתם, פטריות זה אחלה בהרבה דברים, אבל חצילים לא נועדו לאכילה
עברית