it’s a paradox, but ads subtract.

“To predict the consequences of an invasion and seizure of Greenland is unthinkable because it’s like asking ‘what would happen if gravity smelled like Roquefort?’ And yet!” @profmusgrave musgrave.substack.com/p/the-gr

@landley do small services have to have anything to do with cloudflare?

in reply to @landley

How expensive would it be to run a full-size DeepSeek-R1 on ones own hardware or rented server?

Instead of paying giant service providers, could small communities of people affordably run their own, for privacy and ideally customization?

you know you are strong when everybody hates you.

You'd think before sending repatriation flights, someone would make a phone call? It wouldn't be, like, "Hi! We're here! Can we come in?"

@Phil @realcaseyrollins you guys are all so nice.

in reply to @Phil

abundance is awesome and i support it, but presuming abundance will overwhelm distributional concerns is a bet we’ve already taken that works out poorly, both from an abundance and distributional perspective.

i am radically for the equitable abundance agenda!

@realcaseyrollins the humans are mostly very nice people. but each humans is a potentially dangerous criminal alien with respect to nearly all communities.

in reply to @realcaseyrollins

The American consumer looks forward to valuable innovations.

potentially dangerous criminal aliens are mostly very nice people.

the US military is an example of an effective, high quality labor pool reinforced by a well organized benefits state. sets a dangerous example.

does deep seek low training cost presage the possibility of models built and trained by nonprofits with broad public support, with no plutocrat or aspiring plutocrat anywhere in the pipeline?

@ZaneSelvans it’s insane.

in reply to @ZaneSelvans

in prospect revenge is more appealing than remedy. but it’s like a bag of cheetos — you’ll find you wish you’d gone for the salad after you’ve binged.

@admitsWrongIfProven AI models that are less costly in compute and bandwidth to train, in compute and memory to run, that can be run locally on consumer hardware rather than via API in the cloud. New Chinese models like DeepSeek and Bytedance’s are examples. But other, prior models are runnable on consumer hardware too. If API pricing reflects cost, though, the new Chinese models are lighter. I’d like to see a global competitive race to get lighter!

in reply to @admitsWrongIfProven

“I’m not a doctor, but I play one on TV. Today I will be performing your heart surgery.”

sorry, i love light AI. suck the value out of all the plutocrat’s capex. may all their billions turn to dust while we run whatever fancy autocomplete we want on our laptops.

just encountered a Rolls Royce SUV that looks like basically every other oversize SUV, and i think to myself wtf are you guys doing to your brand?

looking forward to the Lamborghini minivan.

@sqrtminusone that’s the problem. incompetence doesn’t prevent them starting wars. just because they don’t win ‘em doesn’t save anyone from the pain.

in reply to @sqrtminusone