Intelligence is essentially data compression with a sense of direction. High dimensional topological geometry informs the direction.
However, one can be quite capable, intelligent, and skilled without being wise. It's important for agents to have discernment, consistent thought over long time horizons, and consideration of consequence.
We don't seem to have an apt analogy or abstraction for this aspect of cognition. Not only are these metacognative tasks disproportionately underrepresented in datasets, they are underutilized by most humans.
The challenge is to create algorithmic abstractions for deductive reasoning and metacognition, as was attempted with attention. As it turns out, attention is not all you need. Deducing what to pay attention to is even more important.
Energy is the new emergent fundamental currency, which can now be converted directly into monetary gain or access to intelligence.
On december 8th 1998 @elonmusk wrote to Wei Dai:
there must be something here that I'm missing. At the core of this
protocol seems to be the establishment not of crypto anarchy but of a
crypto elite. in this scheme only the processors of computing power have
economic power. Now I realize that our current economic system is based on
economic power being invested in a closed community of powerful elites, and
is by no means egalitarian, but this looks to be like simply substituting
one group of "haves" for a different group of "haves"
I know multiple people who basically quit working on Bitcoin because of how stressful it is to maintain software that handles money nowadays.
I tell them all the same thing: What do you think will happen if we all quit and become gardeners? Are we just going to give up on the internet? Do we just accept fiat as the end game? Are we just going to let go of the dream of an internet that doesn't exploit your humanity and take away your rights to privacy and self-determination?
I know, fuck, it is hard. But you're not alone.
We must fix this. Nobody else is there to do the necessary work. If we stop doing this, we're 100% doomed. I can guarantee it.
But if we continue, we have a chance.
XΞS retweeted
Replying to @DionysianAgent
Jesus had a son. There is a secret unbroken lineage but it has been obscured and today, someone out there has yet to realise he is the heir of Jesus.
The space meteor, we know where that is.
And inside, legends say there is a magic puzzle box that if solved, will bring the unification of the faiths of man and the new age
XΞS retweeted
my *personal* opinion about this is that it's basically too late w/ deepseek flash 1017 and we should just rip the bandaid off and let everyone rip everything apart together
we need to trust in humanity and give them the tools to defend themselves, instead of baseline expect criminality
XΞS retweeted
Bitcoin has a hidden message that can never be deleted.
Bitcoin block 666,666 was mined on January 18, 2021.
It contains a message permanently written into the blockchain.
Using Bitcoin’s OP_RETURN, the miner embedded a Bible verse directly into the block data:
“Do not be overcome by evil, but overcome evil with good.” Romans 12:21
To make it happen, they paid more than 5x the normal fee to guarantee inclusion in that exact block.
The transaction is linked to wallets named “GoD” and “BibLE”, and you can verify it on any public block explorer.
Once it’s there, it can’t be removed.
XΞS retweeted
Lol I am pretty sure someone is runing a deepseek swarm to remotely patch unpatched php boxes. Well good on them.
Live by the sword, die by the sword.
Those who pick fights with me should understand that I will find out who is really pulling the strings and come for them, not just their puppets …
Capitol agenda: Dem gavels are coming for Elon Musk politi.co/4dPcGpd
This solves the "how many R's in strawberry" problem. Called it a while ago.
github.com/Named666/ByteGPT
Meta published a paper that might end the current LLM era.
They proved that we can kill the "tokenizer" forever and it actually makes the models stronger.
Every AI you use today (ChatGPT, Claude, Gemini) reads text the exact same way. It chops your words into arbitrary chunks called tokens.
It relies on a massive, bloated vocabulary of around 100,000 specific chunks just to understand a sentence.
We assumed this was the only way to build AI. We treated it as the fundamental speed limit of language models.
Meta proved we were completely wrong.
Their new paper, "Breaking the Token Ceiling," ditches tokens entirely.
Instead, they forced the AI to process raw bytes. A vocabulary of just 256 values.
The results rewrite the rules of model efficiency.
By training on pure bytes and distilling the knowledge, these tiny models didn't just survive. They dominated.
Meta's distilled 1B-parameter byte models asymptotically surpassed state-of-the-art token models like Llama 3.2-1B and Gemma-3-1B across downstream tasks.
They use a fraction of the vocabulary. They require less data. They take up less storage.
It forces the model to actually learn the fundamental relationships between raw data instead of relying on pre-packaged text chunks.
This is a massive shift for anyone building, training, or deploying AI.
We’ve spent the last five years obsessively optimizing how we chop up text to feed these models.
It turns out, the bottleneck wasn't the intelligence of the model.
It was the tokenizer.
1
10
1,249






















