@shaguncrypto $BTC is the exact opposite of a money laundering system. Off-ramping makes it practically impossible to avoid leaving a trace, pointing straight to the user or exchange complicity. Meanwhile, the banks and institutions calling $BTC a money laundering machine fit that exact label!
This entropy attack makes a very good case for covenants.
- Your Bitcoin sits inside a vault.
- Your normal key can initiate a withdrawal.
- The withdrawal is publicly visible but delayed (eg: 24-72 hours)
- During that delay, a separate recovery key can cancel the withdrawal and move everything to a secure recovery wallet.
- Only after the delay can the Bitcoin reach the requested destination.
This would NOT ONLY make individual HODLing entropy-error-proof, it would benefit Big Bitcoin.
If you have at least 100 $tao, you now can just chill and relax. Come back in 10 years, you’ll be a multimillionaire. No need to stress about the price now. Just worry about accumulating that many
bittensor:native is currently in the best state it has ever been and i hope it will continue to be here
-subnets cant exploit due to manual interference
-stale subnets don’t get emissions anymore
It became unattractive to extract, if we resume current state i expect exponential increase in quality
BLACKROCK CEO LARRY FINK HINTED AT SOMETHING CRAZY.
HE SAID SURGING DEMAND FOR COMPUTING POWER COULD SOON CREATE AN ENTIRELY NEW ASSET CLASS CENTERED AROUND BUYING AND SELLING COMPUTE FUTURES.
167k tokens per second training throughput on a single 8x 5090 box with parallax (on a toy 5b 1b active test moe)... That's quite high. Every GPU you add reduces compute for the others so that number goes up > linearly. Exciting times. And look at the beautiful loss curve, beating DDP baseline at matched steps (identical warmup/lr/dataset/etc.)
With 5.99 GFLOPS per token for this arch and BF16 denominator that's ~60% MFU equivalent.
3.068 nats so far at ~19b tokens (llama-3 tokenizer), beating the DDP baseline at matched token budget and getting there ~15% faster and 64% cheaper, and remember it's ternary routed experts...
Best guess as to nats gap is separate expert muon optimizers with the surrogate feedback+fold normalization etc.
Almost ready for the big runs...
I recommend that my friends buy one $Tao for two reasons: 1. the extremely tiny chance it goes on a Bitcoin run, and 2. more importantly, so you vote on distributed, uncensored, sovereign AI while “betting to learn”
Worst case, it's a $200 lottery ticket to what I believe will be one of the most chaotic and important technology shows of the next five years.
I'm long. Get educated. Have fun. Don't buy more than one. Don't blame me if you lose it all!
we got the full Kimi K3, 2.8T params, running on 80x RTX 5090s.
20 tok/s single stream, day one, untuned. Last week we took GLM-5.2 from 30 to 110 tok/s on this same fleet. This number will climb.
A first for open weights: frontier intelligence served with zero HBM, the scarcest silicon in AI. Just GDDR7 gaming cards, plain ethernet, and the official MXFP4 weights, nothing requantized.
The most powerful open model on Earth, on the most abundant GPUs on Earth. Any lab, startup, or university can now own it, probe it, fine-tune it, run agents on it.
@Kimi_Moonshot