utxo the webmaster ๐Ÿง‘โ€๐Ÿ’ป's avatar
utxo the webmaster ๐Ÿง‘โ€๐Ÿ’ป
_@utxo.one
npub1utx0...50e8
Wisp: https://wisp.mobile Nostr Archives: https://NostrArchives.com Haven: https://github.com/barrydeen/haven Rektbot: npub1r3kty2vkh247jgdu63wgkcsnktdtp9hc3e962eudg0getgvxs4gsz4uytc All Projects: https://github.com/barrydeen
There's an AI conference happening near me, so I check it out Speakers are purely like mid level managers at banks and telecom companies I couldn't imagine a worse line up of speakers, other than the lineups at bitcoin conferences
Dgx sparks are not slow, that is fake news Especially when you factor in prefill, which is half the speed in big context New software optimizations drop almost every week that push the speeds up, little things like math kernels etc Currently getting single stream 60 Tok/sec decode and 11,000 Tok/sec pp running deepseek on two sparks People just looking at memory bandwidth and drawing conclusions don't know what they're talking about
Suppose fiat currencies did collapse and everyone moved to bitcoin Do you really think the powers that be, be it armies or the mob, will just allow us to be the new rulers/bankers of the world because we were early? No, we'll probably get guillotined
I wish I was smart enough to understand what will happen with all the bond stuff Japanese yen bailout to prevent them from dumping treasuries, then some scammy "debt buy back" (with what money? More debt?) while crossing $40 trillion Either way feels sketchy af, not sure "just buy bitcoin" is the answer My best guess is that interest rates are about to go up a lot, which is bad for everything including bitcoin But I'm retarded who knows gonna go mow my lawn now
Next version of DeepSeek flash will have vision and made small improvements across the board on other benchmarks, perhaps out in next two weeks GLM5.3 weights come out in two weeks, performance exceeds opus5 on agentic tasks Qwen3.8 coming out with a midsize model, probably 120b-a10b Gonna be an epic
โ†‘