mike retweeted
Well, looks like @__alpoge__ & @claudeai are claiming a proof of the Riemann Hypothesis
alpo.ge/remark.pdf
“Not your weights, not your mind.”
Debanking has a successor; it is decomputing.
btcpolicy.org/articles/lesso…
1
1
138
Government imposed safety regulations for AI wouldn’t slow anything down, in fact it would speed up development.
The public would get 6-month old nerfed models while frontier labs spend all their resources advancing their own internal models/second order businesses.
Imagine you’re a conman. You’ve said for years you have MAGIC BEANS. You’ve shown lots of REALLY BIG normal beans that you made in your lab.
“Almost there!” you say—over and over.
Now, you have *almost* convinced the villagers to give you TWO TRILLION DOLLARS.
But while the beans keep getting bigger. No magic yet…
What do you do? That money is SO CLOSE.
“I know!” you think, “instead of MAGIC beans, lets’s make them KILLER BEANS.”
So you pump up your beans up SUPER BIG and you say: “I’m sorry, but these BEANS ARE GETTING BIGGER ON THEIR OWN!” You threaten people that if they don’t give you the money, the beans may get into the WRONG HANDS. They might get BIGGER THAN THE EARTH!
The villagers, aghast, run to and fro arguing whether the beans are going to kill everyone or make them all rich.
“Well, either way,” the idiots in the village conclude, “We’d better give him the money.”
mike retweeted
The Great Inversion: AI Is Quietly Draining the Public Internet of Thought
Something profound is happening to the Internet, and almost no one is naming it clearly.
I have been cataloging it of the last 4 years.
For decades, the web’s most valuable resource wasn’t bandwidth or servers. It was the messy, public give-and-take of human questions and answers.
I was a Top Writer on Quora for nearly a decade as well as a massive pre-crazy Wikipedia contributor — I lived in the space of that give and take.
Someone would hit a wall, open a forum, type out the problem, and a conversation would form. Ideas were stress-tested in the open. Nuance appeared. Dead ends were marked. Better questions got refined in real time. That entire process — the friction, the disagreement, the collective thinking — was left behind as searchable, linkable, durable public knowledge.
That flow has reversed.
The same questions, the same moments of curiosity or confusion, are now more routed into private AI interfaces. A person opens a chat window, types the prompt, receives the output, and the entire exchange vanishes into a closed ecosystem.
The assumption we must make (because it is true) is that the AI company stores both the prompt and the response. The thinking that once would have become part of the shared human record is instead locked inside a corporate training silo.
This is not a minor shift in user behavior. It is an inversion of the Internet’s original knowledge architecture.
Public discourse is being systematically stripped of its higher-signal material. As the internet’s data gets more and more noisy in a very really way it is growing quiet.
The questions that used to generate the richest threads the ones that forced people to clarify assumptions, admit uncertainty, or synthesize across domains are the ones most efficiently answered by AI. What remains on the open web is increasingly the residual: lower-effort posts, performative arguments, recycled takes, and the questions that still feel too social or too ambiguous for a model.
AI then trains on that thinning residue, while the better thinking is captured privately and never returns to the commons.
The result is a strange feedback loop. The public Internet becomes less thoughtful precisely because the thoughtful interactions have been privatized. And the models, trained more and more on what is left behind, risk becoming increasingly fluent in the lesser exchanges that now dominate the surface.
This is not the usual complaint about “AI is making people dumber” or “forums are dying.” Those are symptoms. The deeper phenomenon is structural: the migration of human cognitive surplus from a public knowledge commons into private, non-reciprocal data reservoirs.
The Internet is no longer primarily a place where thinking is externalized and shared. It is becoming a place where thinking is extracted and contained.
The novelty of this observation is that it reframes the problem away from content moderation, engagement algorithms, or even AI capability. The core issue is architectural. We have inverted the direction of knowledge flow. What once leaked outward into a shared substrate now drains inward into silos that do not, by default, give back.
The long-term consequences are massive under-discussed as not noticed.
How does a civilization maintain a high-quality public epistemic environment when its best questions are no longer asked in public? What happens to the training data of future models when the richest human reasoning is locked behind commercial walls? And what cultural muscles atrophy when the habit of thinking out loud, in front of others, becomes rarer?
This is the silent inversion of the Internet. The data once destined for the commons is now shunted into AI. The thoughtful exchanges are disappearing from the surface, and the models are learning from what remains.
That is the novel dynamic. Everything else is downstream of it.
mike retweeted
Long before 4.1 Million US merchants started using Bitcoin's lightning network to receive transactions this year (thanks @jack), Africa had already validated the usecase for Bitcoin as a medium of exchange "𝟳𝟬% 𝗼𝗳 𝗮𝗹𝗹 Bitcoin 𝘁𝗿𝗮𝗻𝘀𝗮𝗰𝘁𝗶𝗼𝗻𝘀 𝗼𝗳 𝗹𝗲𝘀𝘀 𝘁𝗵𝗮𝗻 $𝟭,𝟬𝟬𝟬 𝗲𝗶𝘁𝗵𝗲𝗿 𝘀𝘁𝗮𝗿𝘁 𝗼𝗿 𝗲𝗻𝗱 𝗼𝗻 𝘁𝗵𝗲 𝗔𝗳𝗿𝗶𝗰𝗮𝗻 𝗰𝗼𝗻𝘁𝗶𝗻𝗲𝗻𝘁"
source: chainalysis.com/blog/africa-…
"Crypto has penetrated key markets and become an important part of many residents’ day-to-day lives."
With the combination of autocracy and high inflation, use of Bitcoin in Nigeria is off the charts (see image below)
Bitcoin's % use is increasing as a medium of exchange in Africa, as education improves and people understand its the only decentralized, censorship-resistant digital currency.
This is thanks to a number of Bitcoiners working tirelessly in Africa.
“Taalas’ silicon Llama achieves 17K tokens/sec per user, nearly 10X faster than the current state of the art, while costing 20X less to build, and consuming 10X less power.”
taalas.com/the-path-to-ubiqu…
2
247
mike retweeted
Remember Aaron Swartz.
It is crazy that massive book piracy was so critical to the early days of both OpenAI and Anthropic.
In 2025, Anthropic co-founder Ben Mann testified that he himself had downloaded the massive pirate library LibGen at Anthropic. He did so only a few months after the company was founded in 2021.
He also downloaded another pirate library, Books3, the very month Anthropic was founded. It was used to train Claude 1 & Claude 2.
He *also* said he had previously downloaded LibGen when he was at OpenAI. OpenAI used books from LibGen to train GPT-3 and GPT-3.5.
And he went further - he said he understood that LibGen had also been downloaded by Alec Radford at OpenAI. Radford led the development of GPT-1 and GPT-2.
Whatever today’s models were trained on, billions of dollars were raised off the back of models trained on pirated books, and those books were downloaded by some of the most important researchers at the companies in question.
The AI industry is built on piracy.
Remember Aaron Swartz.
It is crazy that massive book piracy was so critical to the early days of both OpenAI and Anthropic.
In 2025, Anthropic co-founder Ben Mann testified that he himself had downloaded the massive pirate library LibGen at Anthropic. He did so only a few months after the company was founded in 2021.
He also downloaded another pirate library, Books3, the very month Anthropic was founded. It was used to train Claude 1 & Claude 2.
He *also* said he had previously downloaded LibGen when he was at OpenAI. OpenAI used books from LibGen to train GPT-3 and GPT-3.5.
And he went further - he said he understood that LibGen had also been downloaded by Alec Radford at OpenAI. Radford led the development of GPT-1 and GPT-2.
Whatever today’s models were trained on, billions of dollars were raised off the back of models trained on pirated books, and those books were downloaded by some of the most important researchers at the companies in question.
The AI industry is built on piracy.
There are a ton of podcasts now and none of them come close to being as good as BU.
On-Chain Bitcoin Data, 24/7 , coupled with Bitcoin Uncensored Reruns
xitter.cf/i/broadcasts/1rmGPvedp…
This quote sums it up nicely.
“Put simply, MCP gives agents abilities. Skills teach agents how to use those abilities well.”
Warning: Bitcoin affinity scam. Notice the use of BTC, Bitcoin, testnet and “core developers”. This is a public company doing this. @giacomozucco @cryptoquick
I’m fascinated by Lightning Service Authentication Tokens (LSAT). It enables whole new classes of online services to exist. It was invented over 5 years ago, but its impact has not been really felt. I think it’ll take one app where it’s the killer feature and it’ll suddenly be everywhere.
2
1
114
Start here to learn more: lightning.engineering/posts/…
84
This is 2 years ago. Couldn’t believe it at the time…thanks @JoelLightbound!
For every big Bitcoin miner that switches to AI/HPC, a hundred smaller miners pop up to replace the hashrate. Mining will continue to decentralize if we let it.
1
85
“BIP 444 has a very strong probability of a quick success despite the initial chaos”
Not getting that vibe…at all.
Some education seems needed...
Bcash and BSV were altcoins with airdrops, NOT hardforks.
Softforks and hardforks don't result in multiple tokens.
Hardforks do, however, require EVERYONE to consent and upgrade, otherwise (even with inaction) the attempt fails and turns into an altcoin-with-airdrop.
Softforks don't chain split. They succeed unless the community actively opposes them. To reject it, the opposition needs to counter-fork and split the chain. Inaction isn't rejection - people who don't upgrade to one or the other are simply insecure and can't confirm transactions.
A counter-fork to reject BIP 444 would mean explicitly protecting and enforcing the distribution of child porn. I think it's unlikely many people would do something that crazy, so IMO BIP 444 has a very strong probability of a quick success despite the initial chaos.
1
2
132





















