ricericerice

That doesn't seem to match the arena.ai rankings, who I'm assuming has a far larger dataset of rankings than you currently hold.

https://arena.ai/leaderboard

It's not like Claude's subscription is super clear, either. All of these pricing models are a bit too opaque for my taste.

Indeed, while struggling with VCS is a right of passage for any tech company, I'm not sure why what they described couldn't have been built as a diff/patch API in front of git.

just tell them to point their agents at it /s

you have no idea how bitwarden works, do you...

by that logic, every time you send a password over a TLS connection, you're publishing it outright too

you can easily finetune your own*

*if you have a sufficiently sized and quality dataset for the specific classifications you're targeting

i have used racknerd for years - no fancy features like a real cloud provider, but you get a stable VM for very competitive prices, and they're not super overprovisioned like some cheap providers

https://www.racknerd.com/specials/

I found the egregious LLM generated text a bit of a turn off, too.

If you don't care about your website, why should I?

I think the concern is less over this specific feature and more about what this means for the project roadmap, now that they've chosen the route of built-in rather than first-party extensions.

Add in the context that they're also building their own inference provider[1], and there's certainly a non-zero risk that the project trends away from the barebones platform to build on that people signed up for.

[1] https://radius.earendil.com/

got a link to a laptop that can run K3?

i think you underestimate how much you "just getting it" stems from being terminally online - this is a HN model release thread after all

my feeling is you're most likely wasting money using Opus for implementation. The plan and task breakdown should be specific enough that Sonnet can implement without you noticing a difference.

How do you verify in practice then? Wouldn't you need the original prompt so you can reobtain the likely token distribution to validate again the random key(s)?

After some research, I think I understand what you're getting at here - BERT being a model for encoding text but not architecturally feasible to generate text with it, which "LLMs" (the lack of definition here is resulting in you two talking past eachother), maybe more accurately referred to as GPTs, can do.

Also the irony of your comment when it in itself was confidently stated yet void of any content was not missed either - consider dropping the superiority complex next time.

i believe `jj fix` is your -x equivalent, though im not familiar enough with it to comment on how similar the semantics actually are

jjui is definitely the best contender ive seen for a jujutsu ui. although, with the ergonomics of jj, and a decent merge editor setup like meld, i haven't really needed a ui at all

there's 1,000,000 validators (defined as a public key), but you can run multiple validators per machine. Most estimates that crawl the p2p network to index nodes comes out at like ~20,000 machines

doesn't invalidate your point but it at least shaves off a few orders of magnitude

and a single PG node is not a fair comparison, we're talking 100% uptime global networks. Visa does about 70,000 transactions per second - how many servers do you think they run across their infra?

choose-your-own-adventure style chatbots

you can set up the watchman fs monitor to snapshot whenever a file changes on the filesystem, its saved my ass plenty of times

I mean, ChatGPT could easily be profitable today if they wanted to, but they're prioritizing growth

why even use vim at this point? the LLM ecosystem there is decent, but definitely less polished than using a modern IDE