Ving

@BasedBeff you cannot outaccelerate me

rainy night in taipei

European dying because of no AC *meanwhile in china

it is so over for american AI. at this rate the us is gonna lose the short term AI race by jan 2027

@zanoga can you test glm 5.2? i think it would run really well on that system at 4 bit quantized. maybe with a bit of post training we would have some 4.8 level shit run locally

Finally finished building my AI datacenter! 🚀 32x3090s across 4 servers (8 GPUs each), all connected over InfiniBand. The whole setup is solar-powered with a massive battery bank and generator backup. More technical details and benchmarks coming soon.

every time a space image was used for a model it was a disappointment

@markksantos @threejs can you try codex harness, i found that opencode harness tends to over trying to save tokens

SAKANA FUGU ULTRA vs. CLAUDE OPUS 4.8 RESULTS Prompt: "build a really high quality single html file crossy road game with three.js" Sakana Fugu Ultra: - Tokens Used: ~89k ($7.32) - Time Elapsed: 22 minutes - Issues: inverted directional turn, wonky camera, no sfx, not identical to Crossy Road game Claude Opus 4.8 Ultracode: - Tokens Used: ~940k (~$37.85) - Time elapsed: 79 minutes - Issues: got stuck twice in a retry loop (had to prompt to self-correct), wrong character position after restart, difficult from the start (whereas Fugu's version got significantly more difficult as you progress, which is the correct behaviour) I think in terms of application functionality, quality, and design, Opus won. In terms of model speed and performance, Fugu on Opencode won. What are your thoughts? Comment down below who you think won! ⬇️

i didnt know you can play angrybirds on the handheld console on eva air flights

@elonmusk @teortaxesTex won’t take that long

Against all odds. Congrats @elonmusk and @SpaceX.

when Le Chaton Fat max effort fast mode ?

@claudeai @Padierfind lets go

Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use. Its capabilities exceed those of any model we’ve ever made generally available.

life in a third world country

@TxoriAGI @jasperdevs they use gemini 1.5

@jasperdevs Hmm weird, it depends which models they are using I do think is probably a cheap one, but how cheap

you could probably use poke 24 hours a day for 500 years straight and it would cost them $5 in API costs

@jussy_world @Arcium ico?

Soon we'll see the biggest TGE in crypto since the start of the year $ARX from one of the leading teams in confidential AI While most projects focus on AI applications, @Arcium is building the infrastructure layer If you're bullish on: • NEAR: $3.5B • VVV: $900M • RAIL: $130M You should be 10x more bullish on $ARX

@nicosolgetta thats like 50 dollars

lel niggas saying trenches ded bet im up 400 sel

@cxnversion only gay ass men paying 30k luke belmar

best 30k ive ever spent

@heyibinance @hongshen6666btc 怎么样？

@hongshen6666btc 有

问个问题：在币安买美股有分红吗