@mjblacker @DanielW_Kiwi @Jimmy_Byrd It's up now, by the way - prepare to be disappointed, because an F# LSP with extremely limited typechecking is basically unusable!
English
Patrick Stevens
684 posts
@mjblacker @DanielW_Kiwi @Jimmy_Byrd Feel free to watch github.com/Smaug123/borzo… which is where it'll appear when I've prepared it for open-sourcing. (It's nowhere near ready for use yet, but also it's even more not ready for open-sourcing; hopefully fairly soon I can at least release the source.)
English
@mjblacker @DanielW_Kiwi @Jimmy_Byrd SharpLsp uses FCS and is therefore too slow. I am hosting a C# sidecar but implementing F# myself.
English
@Smaug12345 @DanielW_Kiwi @Jimmy_Byrd SharpLsp?
Was checking that out the other day, very interesting
English
@Jimmy_Byrd @DanielW_Kiwi The fact that I can parse all of the dotnet/fsharp repo before the .NET runtime finishes starting up is really upsetting, isn't it.
English
@Smaug12345 @DanielW_Kiwi Damming that no one is doing this in dotnet.
English
@Jimmy_Byrd @DanielW_Kiwi Precisely so. This one's in Rust so it's bound to be fast!
English
@Smaug12345 @DanielW_Kiwi Unless its recreating FCS, were all bottlenecked by the same thing. Everyone making their bespoke LSP is going to be about the same.
English
@DanielW_Kiwi @Jimmy_Byrd Do you want a fast editor experience, or do you truly want fast compilation? Because I *am* about 30% of the way through a fast LSP for F#.
English
@TheZvi Massive step up from 5.5 in code-correctness review ability (and 5.5 was no slouch). Out of the box, perhaps a bit *too* agentic when implementing features? I don't really want it to silently bundle in five unrelated bugfixes, especially ones with nontrivial ramifications.
English
@HVrotec46045 @ClaudeDevs Cherries grow on trees - how tall do you think Claude is?!
English
claude invented the concept of “wedging” for me the other day. the good news is that a bad upload no longer wedges the whole batch
gabe@allgarbled
One of the worst things about working with LLMs is the way they constantly make up fake technical jargon, and it only seems to be getting worse. Nobel prize to whoever fixes this.
English
@Smaug12345 @lthlnkso Nobody was talking about baby humans. That you've survived to apparent adulthood is evidence enough that society does a good job of caring for its "challenged" members.
English
If humans were to adopt monkey economics, we too could enjoy the luxury and wealth of monkey civilization.
bug girl@buggirl
in nature if a monkey hoarded 1 trillion bananas the other monkeys would beat that monkey to death and take his bananas
English
@RSI_Lance @lthlnkso How smart are the baby humans you're interacting with?! I don't think "a few months" after birth is enough time to observe any difference at all in human survival rate due to intelligence; that time frame is more like "hold own head up while sitting upright".
English
@lthlnkso In the wild, a monkey as stupid as you would not survive more than a few months outside the womb.
English
@TheZvi Visibly better than Opus at coding, though GPT-5.5 still routinely finds critical bugs in its work. Much more pleasant to talk to than Opus; doesn't do the annoying thing of inventing strawmen you didn't say so it can refute them, for example.
English
LFG. Claude Fable 5 reaction thread. One united thread for everything - capabilities, model welfare, classifiers, etc. Respond multiple times on multiple topics, if you have lots to say.
English
@mathandcobb Makes roughly as much sense as counting neuron activations in a human to judge their age.
English
Or... should we count the total amount of computer clock time... and if it adds up to over 40 years, then that's it.
I imagine we are already well above 40 years of clock time, though??
Alvaro Lozano-Robledo@mathandcobb
If AI doesn't solve a millennium problem by the time it's 40 years old, then it won't be eligible for a Fields medal.
English
@lacker @tribbloid It's kind of like an axiom except it's provably inconsistent? I'm interested to know how you avoid explosion here (or do you just go "this program may exhibit undefined behaviour in edge cases, and we can't characterise those edge cases")?
English
@tribbloid like if you have two strings s1 != s2 then hash(s1) != hash(s2), it's "kind of an axiom"
English
I'm surprised to learn that Lean never had the concept of "executable axiom": a function/predicate that can throw an error (thus not a constructive proof), but the probabilty is so low that it can be declare an axiom and be used in theorems.
not cool for a dual prover-executor🧵
English
@tweeter_handle @IsaacKing314 Nah I also remember the security guy being a grandmaster. Internet-searched quote, dunno if correct but matches my memory: > "Once a checkpoint official, himself a chess grandmaster, had doubted their story, until Artemis beat him in six moves."
English
@IsaacKing314 It's been like 20 years since I read those but I remember that part. But I don't think it was a grandmaster he beat? He was disguised as a pro chess player or something and some security guy doubted him, so he beat him in 6 moves to prove it. At least that's what I recall.
English
One of the Artemis Fowl books I read, there was a passage about him beating a chess grandmaster in 6 moves. Knowing that the author couldn't be bothered to learn the first thing about chess before writing about it rather soured me on the series.
corsaren@corsaren
Why are you, a resentful autistic formerly gifted kid, reading HPMOR when Artemis Fowl already exists?
English
@KelseyTuoc Can 4.8 identify you? I just tried the experiment again: 4.7 can identify me, 4.8 appears not to be able to (so far).
English
I have a bunch of secret AI benchmarks I only reveal when they fall, and today one did. I give the AI 1000 words written by me and never published, and ask them who the author is. They generally give flattering wrong answers (see ChatGPT, below:)
English
whats your favorite type system feature in a programming language?
it can be an unused lang or whatever, just something cool that you think other languages could be doing if it they were designed w/ it from the start
... definitely not making my own lang ... do not tell prime
English
funny to look back at
Acer@AcerFur
how long until AI solves a fairly famous open math problem from a century ago
English
@nosilverv Say more? Gödel exhibits a sentence which holds in one model of PA but not in another. (Although the models where the sentence fails are very noncomputable.) Really struggling to see the connection to stats.
English
