twalla

I have a 5090, once you start splitting weights/cache amongst devices or start having to rely on unified memory you quickly enter the land of diminishing returns and tradeoffs around cost, power consumption and speed. I'd just go for whatever the cheapest single device that will fit your chosen model weights + context at the speed and price you are comfortable with, which for most folks will probably be a single GPU and a model in the 30b neighborhood (qwen/glimmer)

there's eventually a point in organization size where it makes sense to have some kind of centralization to make it easier for teams to actually be on call for the systems they build without grinding them to dust.

You joke, but they already have birds https://www.dexerto.com/counter-strike-2/cs2-adds-pet-chicke...

The default spacing is too narrow for my liking but what's really cool is in the amount of customization that's possible, including lookalikes of some popular paid programming fonts (I'm partial to Iosvmata).

https://github.com/ahatem/IoskeleyMono

https://github.com/shytikov/pragmasevka

https://github.com/N-R-K/Iosvmata

Retaliation for Ukrainian attacks on Yandex datacenters most likely - it communicates to EU states that Russia has capabilities to physically impact their infrastructure via cyberattack, which should act as a deterrent for sending aid or admitting Ukraine into the EU.

The real price of anything you see on 1stdibs is about 40-50 percent lower than what you see. Their fees are egregious and if a piece is unique or the seller name has enough info it's almost always very easy to find the seller and contact them directly via instagram. Also, if you want more market price information go look at sold prices on liveauctioneers (just keep in mind buyers premium is about 25-30 percent)

People are bombarded with crappy genAI graphics and even if they can't articulate what about it they dislike aesthetically, I believe most are able to draw the short line from the bad, piss-filter, fake-ghibli slop they see to the social/environmental ills most associate with AI. I have mixed thoughts on shame as a tool for social cohesion but in my mind it's a positive that use of genAI for artistic purposes seems like it's nearly universally panned and people trying to pass off the output as their own are (rightfully) clowned.

The text exchange in the article with the thinly veiled threat is exactly how you'd expect someone disempowered (and it's not like working artists pre AI times had a ton of capital to begin with) to act when they are given some power. As we continue displacing, or attempting to displace, individuals up the value chain, I think we'll continue to see echoes of this kind of behavior. When the luddites can't break into the factory to smash the machines, smashing the outputs and users of the machine seems like the obvious choice.

180km is pretty bad but you also trained in a race shoe that deliberately makes material tradeoffs to increase running efficiency.

You may like this: https://www.washingtonpost.com/video-games/2022/10/13/nba-sp...

Women get an absolute deluge of interest - when I was dating it wasn't uncommon for me to see women with inboxes with 10k plus messages/likes. There's definitely an aspect of "just pick the hottest 2 or 3 people you're shown and close the app" because you're overwhelmed and over half of the profiles are poorly filled out/sending low effort or vulgar messages. I think most men do poorly because they don't understand the female gaze and they don't take the time to create a profile with half decent photos that communicates who would want to date them and why.

And nobody with actual taste is going to take any of those recs seriously in the first place.

Looks like you're right - I've only been on the buy side of this equation. The company has regular tender offers but that's significantly less liquid than my earlier comment implied.

There's a fairly liquid secondary market for OpenAI stock - see: Forge, Hiive, EquityZen

That’s odd because spend on all of those things per capita is consistently lower with better outcomes where they’re socialized.

You’d be surprised. I got an absolute boat anchor of a 2019 MacBook when I started a job in early 2022. I requested a new one in late 2023 because company policy is to refresh laptops every 4 years - whoever was on the other end of my support ticket interpreted the policy as four years from issuance and denied me.

depends on how you define equally, but probably between 100 and 200 miles https://runrepeat.com/state-of-ultra-running

whatever you say bro - https://www.thelancet.com/article/S0140-6736(25)01186-9/full...

Can you go into more detail on this? I’ve been putting together the data pipeline for an ASR fine tune but your approach sounds more interesting.

that spectrum is really more of a horseshoe than most would think

There's a graphic illustrating the typographic homogenization of a bunch of fashion houses in the late 2010s early 20s like you described in the article below. It's come full circle and there's been a return to maximalism and signaling heritage/legacy among some brands. Largely, I think, as a reaction to ever-accelerating trend cycles, planned obsolescence, the cursed nature of millenial targeted d2c brands and especially the looming specter of big tech - all of which have some sort of association with the clean, cool/pastel, sans-serif, minimalist kind of vibe.

https://www.designandpaper.com/they-are-calling-it-a-sans-se...

The whole cycle of aesthetics is about signaling and countersignaling. This whole "look" is also about what it _isn't_ (bright, sterile, sans-serifs, weird corporate memphis purple squiggly people). It's part of the reason all the tech bros won't shut up about "taste" now and Palantir has a storefront with a chore coat drop (someone thought they cooked lol). I don't know how you'd even begin to turn the understanding around contextualizing why something that used to be "in" is "out" or why things like tension/juxtaposition/irony/nostalgia work into something an LLM can grasp. The irony here is that by pushing LLMs in this direction they're just more rigidily codifying the whole aesthetic and giving it a stronger, rancid AI stink which will only bring about the new thing sooner once people see this whole look as the visual equivalent of the em-dash.

It seems like you’re dramatically overcomplicating what could be a 2 bash scripts and a single LLM call. Natural language and LLMs are great for searching the problem space to find a solution, once you find the solution, shrink the stochastic parts (the MD saying “check GitHub” or whatever) and grow the deterministic parts (a bash script) as much as possible.

"some routine tasks like coding and debugging will fall back to Opus 4.8."

disappointed but not surprised

What are the odds these nonprofits were chosen due to proximity to communities with contentious datacenter buildouts?

these are lovely my guy

I'd love to own SpaceX - what I don't want to own is all the unprofitable, toxic dogshit its ketamine-addled CEO folded into it that has nothing to do with putting stuff into orbit or selling Starlink.

James P. Carse - https://www.goodreads.com/book/show/189989.Finite_and_Infini...

There is indirect exposure to Anthropic/OpenAI through all the incestuous partnerships between hyperscalers and NVDA

Hilarious how this guy treats the “confession” as some sort of smoking gun rather than the exact same stochastic slot machine that enabled him to score an own-goal on his prod database.