csjh

Since static lifetimes last for the entirety of the program, it can sub in for any other lifetime (as it is guaranteed to exist until the end of the other lifetime)

hypixel is an impressive minecraft server, not mod

I disagree, IMO for any model that isn't open source model size is just an implementation detail. If someone released a 10 trillion parameter model that's better than GPT4 it isn't somehow inferior because it has more parameters.

Hard to imagine it was so major that it lead to him being fired while still being so quiet that it hasn't hit any news outlets

Never heard of `execution`, great tip. Is there some big caveat or just generally unknown?

If it couldn't take parameters what point would there be in it?

Only 3GB, interesting to see how small SOTA models in other domains are compared to LLMs like Falcon-180B.

I think it must be based on usage (some GPTs could be using more credits than others)

Claiming "many people" and proceeding to link to 4 posts that use Svelte as an example is a bit silly considering Rich Harris has said, numerous times, that the decision in Svelte's codebase should NOT be taken as advice for what to do in your codebase, because their decision only pertains to a very specific set of circumstances.

But it's an opt-in GC? It's not accidental by any means

That would be nice in a world where Mojo is more complete but it is nowhere near that level right now

Yeah, builtins seem like a bit of a specialized case. If I'm not mistaken, it doesn't even really give a Wasm-side method to create strings? Bit of a shame.

I think string builtins seem to be the direction the wasm committee is moving towards anyways? Unless I'm mistaken

Don't they need both physical access to device + fingerprints/face?

similar in ottawa

It "works", but there is a pretty ridiculous amount of latency.

I believe that's at least part of what the GC proposal is moving towards.

LLMs are definitely not AGI

Python pseudocode:

```py

output = prompt("My favourite animal is ") # returns a generator

start = time.time()

tokens = 0

for token in output:

    tokens += 1

    print(token)

print(f"Outputted {tokens / (time.time() - start)} tokens/second")

```

You measure how many tokens are output in a period of time and take the average?

What makes you think 3.5 is significantly smaller than 70B?

Is there a source on that? I've never seen anyone think it's below even 70B

ChatGPT wasn't surpassed by competitors in less than 6 months, what model would you say beat it that early?

Isn't that just how the view transitions API works?

Fortunately we’ll get revolutionary new products from it, such as calendar app (with AI integration), email app (with AI integration), resume builder app (with AI integration), and more!

I doubt the gains from React rewritten in Rust would be significant; the expensive code is the stuff you write in components

Also Typescript -> WASM exists in some form through assemblyscript

Would be cool for there to be a `llama2.f`, like https://github.com/karpathy/llama2.c, to demo its capabilities

WASM -> x64 as in a full WASM AOT compiler? Don't those already exist? What's the benefit to making one specifically for LFortran? Unless I'm misunderstanding

Super cool stuff though

No clue, maybe they performed too well :p

Based on the name, I’d imagine part of TinyGo’s raison d`être would be a smaller binary size. Plus, seems optimized for speed rather than speed-of-compiler