Theresa May 2017 “There is no magic money tree” What happened next… Brrr it’s actually a printer “There is an infinite amount of cash in the Federal Reserve.” -Neel Kashkari March 2020 youtu.be/gUtJEfB9Hi4?si=eV_G…

Introducing Kardashev-0.7, the world’s first trained swarm made of 32 distinct models Trained with RL for Population Scaling (RLPS), 32 models organically develop specialization & complementary capabilities, delivering frontier performance at: - 0.007x ~ 0.02x of the inference cost - 0.03x of the required memory Civilization advances through different minds specializing and working together. We’re bringing that principle into AI @BanburyRoadAI, we’re scaling intelligence by model count, toward civilizations of models that learn to build on one another

ok i think i finally figured out why alignment has gone basically nowhere after 15 years and several billion dollars. the problem is hard sure. but have you met the people in charge of it 1) the leadership does not know how computers work. ask one what happens when you type a url into a browser and watch them start sweating. dns is a spooky word. they genuinely believe the model lives in "the cloud" like a ghost haunting a castle 2) zero idea how large scale distributed systems work. never been paged at 3am. never watched one bad bgp announcement take half the internet offline. their big threat model is the ai "copying itself across the internet" like its a floppy disk virus from 1998. brother it needs 50 megawatts, a liquid cooling loop and a personal relationship with jensen. it is not escaping onto your smart fridge 3) they dont understand the stack they are supposedly protecting us from. ask about inference servers, kv cache, batching, egress, rate limits. blank stare. ask how the model will seize the power grid and you get a 45 minute answer with a hand drawn diagram and three links to their own blog 4) nobody can define an agent. ask ten of them and you get eleven definitions and a 30 page google doc. to the rest of us an agent is a while loop that calls tools and eats a 429 every thirty seconds. to them its a digital god in its larval stage. my agent cant book a dentist appointment without asking me three times if im sure 5) they started from "everyone dies" and worked backwards so every result is bad news somehow. model refuses, deceptive. model complies, sycophantic. model does well on evals, well now its scheming about the evals. i cant think of a single thing a model could do that would get them to say ok maybe we're fine

True

💯

I love that this paper contains instructions for AI Models summarising it for human readers.

RT @Excellion: Volatility is a gift to the faithful.

The holy grail for robotics is being able to generalize: doing work in unseen places We rented 30 homes in the Bay Area and are doing tasks without any new training

Today’s announcement marks an important step for The Smarter Web Company. Subject to approval of the Prospectus by the FCA, launch, and all conditions to the Possible IPO being satisfied, we expect the proposed Preferred Shares to be the first of their kind in the UK: a pounds sterling-denominated, London Stock Exchange Main Market-listed perpetual preferred share issued by a UK-incorporated commercial company with a Bitcoin treasury strategy. We have been working towards the Possible IPO for some time, and I am very pleased that we are now able to share this exciting development with our Shareholders. The proposed Preferred Shares are designed to provide an additional source of long-term capital, broaden the range of investors able to invest in the Company and further diversify our capital structure. We believe that by providing a differentiated investment option alongside our Ordinary Shares, we can create a capital structure that gives us greater flexibility to pursue our long-term ambitions through different market conditions. This is a natural next step in our 10-Year Plan and supports our ambition to build a stronger, more adaptable business, combining the growth of our operating businesses with the continued development of our Bitcoin treasury strategy to create long-term value for our investors. LSE: #SWC | OTCQB: $TSWCF | FRA: $3M8

Nearly died on holiday, out of danger now. Thanks to the brilliant medical care team and my amazing wife fighting for me whilst I could not fight for myself. Not looked at charts for over a week. Total vibe shift, fucking uponly everything from here. Life and BTC send it to infinity. Grateful for the opportunity to manifest it. Live life to the full frens.

We are retweeting Cory's new daily walking videos until the Bitcoin bull run is here. Didn't you get the memo? Do your part!

I prefer en-vidia, its an actual spanish word for: "envidia (n.): the feeling you get when someone posts their local LLM configuration and you realize they have more VRAM than you have RAM"

Holy, China strikes again: Qwen3.8-Max reportedly worked autonomously for 16 days while costing 80% less than GPT-5.6 Sol and 88% less than Claude Fable 5 on output. And its open weight! Alibaba’s 2.4T-parameter MoE costs $2/M input tokens and $6/M output tokens. GPT-5.6 Sol: $5/$30. Claude Fable 5: $10/$50. Qwen says the model operated autonomously for 16 days, producing 265 commits, 127 PRs and 151 issues through an issue -> code -> test -> repair -> merge loop. It does not lead every benchmark. But it reaches the frontier range across coding, professional work and computer use, while PaperBench puts it ahead of both Fable 5 and GPT-5.6 Sol at a very good pricing. Open weights arrive next week. A model that can economically work for ten days may be more useful than a better model you can afford to run for ten minutes. Ngl another insane china release.

ten significant advances in mathematics and theoretical computer science. solved using an internal version of Astra, our next major model, for a total cost of about $2000 at Sol API prices:

Why the last cycle felt muted — a supply-side answer, with numbers. Microstrategy selling is tiny in comparison. Between 2023 and 2025, Bitcoin absorbed roughly a decade of accumulated forced supply, compressed into about 24 months — arriving at exactly the moment the ETFs launched. GBTC unwind — ~400,000 BTC. 620k at conversion in January 2024, ~222k by that September. Three separate drivers: the discount-arb trade closing out (it had traded to −50% NAV), fee rotation at 1.5% against 0.25%, and bankruptcy estates selling GBTC shares. Larger than everything else on this list combined, and the least discussed. Mt Gox — ~142,000 BTC. Coins with a pre-2014 cost basis. Maximum propensity to sell. The 2022 contagion estates — Celsius, Genesis, BlockFi, Voyager, 3AC. A decade of bankruptcy overhang distributing at once. Genesis and FTX alone were approved to sell ~$2.5bn of GBTC. German government — ~50,000 BTC. Dumped in weeks in July 2024. Famously badly executed. FTX estate. A forced seller at the 2022–23 lows. Creditors who couldn't wait sold claims at deep discounts to distressed funds, who took the recovery and returned it to LPs in fiat. That value left crypto entirely. US government Silk Road sales. Tens of thousands of coins. That is not a normal cycle's supply profile. And almost none of it recurs. Then the sign flipped. Executive Order 14233, March 2025: Bitcoin in the Strategic Reserve "shall not be sold." US holdings ~328,000 BTC as of February 2026. The largest sovereign seller became a mandated holder. Alongside it, scaffolding that simply didn't exist last cycle: — Spot ETFs approved — FASB fair-value accounting, so corporates can mark Bitcoin to market rather than impairment-only — The first credit rating on a Bitcoin treasury company (S&P, B−, October 2025) What I'm not claiming: that any of this guarantees a price outcome. Two supply sources are still live, and one is new (more noise than signal). Miners are selling post-halving. And treasury companies trading below NAV can become forced sellers — Strategy itself has now sold ~3,620 BTC in 2026 (~$218m) to fund preferred dividends and rebuild dollar reserves, with up to $1.25bn authorised. Small against the numbers above, but the sign matters: the largest corporate holder, whose identity was "never sell". However look closer and tax loss harvesting and preparing for re-rating are key drivers. Don't forget a meaningful share of ETF inflows was rotation out of self-custody rather than new money, which flatters the demand side. But the question worth asking is whether the last cycle was structurally dampened — a permanent feature of a maturing asset — or circumstantially dampened by a once-in-a-decade convergence of forced sellers that has now cleared. I think it's more the second than most people assume.

Why @Strategy selling is tiny and why the worst is behind us by a long long way... $MSTR

Today is my birthday (turning 30), which is the only time each year when I ask you for a repost if you like my art. Have a good day! 🦦🌿✨️ - Simon

The High Court of Justice in England and Wales has approved Smarter Web's £210 million Capital Reduction. This is a balance sheet restructuring process that converts the locked share premium into distributable reserves. It ultimately paves the way for a UK perpetual preferred.

HUGE breakthrough out today, in arguably the single most neglected aspect of aging, extracellular matrix damage. Top researchers have tried & failed for decades to do this. Massive kudos to Aaron and his team! (Proud to note that Revel is a SENS spinout.) revelpharmaceuticals.com/new…

Writeup in Longevity Technology: longevity.technology/news/wh…