arXiv moderators: https://arxiv.org/moderators/

Editorial Advisory Council: https://info.arxiv.org/about/people/editorial_advisory_counc...

Moderation standards: https://info.arxiv.org/help/moderation/index.html

Thomas Dietterich's thread: https://x.com/tdietterich/status/2055000956144935055

the order here is backwards. publish the package first and let people apply without committing. right now GitLab gets the signal before employees even get the terms.

similar idea, but the failure mode is better. a branch mispredict burns cycles. a bad guess here usually just means no bonus tokens. https://arxiv.org/abs/2211.17192

As long as you're not bound on parallelism or bandwidth then it's "free", but if you're constrained on either resource then your lighter predictor model just needs to save you more cycles than it congests on average.

A bad guess still costs cycles, but the penalty is smaller compared to branch mispredict in the current state. But if we have some kind of pipelining, like if we have something that assumed the speculative decode is correct, then it'll be expensive again.

TUIs were supposed to be the simple option. now they're just web apps wearing a terminal costume

...but without the web's accessibility options, without good text editing, with very basic customisation options, requiring trusted compute instead of working in a sandbox...

They're a long way from web apps, far worse on most axes.

the wrapper is basically env var glue. You’re still betting the whole loop on Anthropic's closed client.

radiology already had its "AI beats doctors" moment. radiologists are still here. what changed first was the workflow, not the specialty. er is probably next.

I don't think radiology has had that moment at all. Computer programming is much closer, if not, at that moment right now.

no programming it's still just tool use for CRUD applications with react and tailwind

complex systems programming is just so unreliable and foolish to use LLMs to do anything important

companies adopting it for more safety critical systems are just already seeing the problems pile on and we're seeing news about it almost every day on Hacker News

If the tool can make something look smart but isn't necessarily correct, lazy employed humans will just defer to it, especially when their lazy greedy bosses tell them to, and everybody loses over time (except the stakeholders that just jump companies anyway after they made their money)

It's just sad to see these really unwise and inexperienced sentiments repeated ad nauseam

Disabling conflicting extensions, sure. Writing to global settings and leaving the mess behind after uninstall is not the same thing.

per-request was broken, yeah. but $10 of monthly credits is basically just a prepaid wallet with a reset timer.

heh the key move is the worm. you can't catch it by checking on a second box because there is no clean box.

cheaper than finding out their own 40B internal project has no users in two years.