Lwerewolf

Flash is runnable in various gpu + cpu offload configurations. Not saying that 27b is bad (far from it, especially if they get n-grams in it with... hopefully today's release), saying that flash is runnable at decent speeds on way less than you'd imagine.

As for opus... I'm the weirdo that refuses to get an ant sub.

Mac user from May this year (maxed out mbp16, you can guess the use case). The dock sometimes doesn't want to unhide, the window manager sometimes thinks it's still there when trying to maximize something (not "fullscreen"), the taskbar (or whatever the thing at the top is called) sometimes blacks out completely when switching workspaces (clicking on it kinda fixes it - not sure if clicking _anywhere_ on it fixes it), still no shortcuts for stage manager, alt(cmd)-tab is still practically useless for me, workspace animations are still always there, etc, etc, etc.

Basically it's anything but "just works". It's still an extremely nice general purpose machine.

Oh, forgot. Spotlight takes over a second to autocomplete sometimes. Everything but apps is disabled from the settings menu for it, and I exclusively use it to launch (or bring back in focus, cause... "alt-tab") applications. Super annoying.

I mean, you do own your work in a way... it's not so different.

>Instead, think like an owner.

Exactly. LLMs make it, I'd say, _possible_, to own way more parts of the code than before, and understand even more.

This one is 2x m5 max, so ~1.2TB/sec.

There are resident systems, and they're of exactly the type listed in the article (i.e. water heaters). Residential... area heating pumps, so to say, are moving towards propane (afaik). I've been looking into getting a propane monoblock air-to-water pump, with an antifreeze loop towards an indoors heat exchanger / water tank.

I came to this relatively late (2013-ish, windows kernel development, scouring OSDev, etc) so I thought that was always the right name for them. Prior experience was mostly... higher-level langs.

As others have mentioned, nothing's stopping any other major provider from offering it. Given its popularity, you can guess how that'll develop. So, overall, irrelevant.

A regular yearly replay is in order for me, too. Already done with SC2... just SC1/WC3/DoW(2)/RA2(mental/flippedmissions/etc)/RA3(uprising)/HOMM3/etc/etc/etc. Ah well.

The model weights are MIT licensed.

Welcome to the club. If you're _really_ competitive in cs2, I'd swap out to a 9800x3d setup, but it's still a maybe. Very little reason to upgrade right now other than to run LLMs.

You need latency for token parallelism, not bandwidth. Hence actual RDMA that bypasses the software TCP stack (ROCe or whatever).

You just need social stigma on bad tires. It's the single most important thing on your vehicle, for literally all the reasons out there.

I literally had Fable tell me that I've checked out a nonexistent upstream commit of LMDB - two days ago. Sol called its BS, of course.

I mean, a big RAG setup with a smart LLM ought to do it, right? Or even just locally provided knowledge database - given the size of LLMs, what's a clone of wikipedia and whatever else you'd need?