if you have a qwen3.5 4b around, please try steering neurons L14 N8949 +1 and L13 N7956 -1 now you have evil qwen recipe: dose × median(‖resid‖ leaving layer L, tokens 1:) / ‖W_down[:, n]‖` at the down_proj at every input

Sometimes you get small crystal clear glimpses of the dynamics waiting patiently to unfold in the coming weeks/months/years.

All Twitters become founder-moderated Twitters. Farcaster by @dwr. Poaster by @lastcontrarian. Twitter by @elonmusk. And so on for every Mastodon. This is a huge shift from the Zuck/Jack era. From social media as neutral utility, to social media as founder-moderated community.

fruit fly can control a robot now

see this is my problem with AI. you can supposedly make anything with it, your imagination should know no bounds, and yet you STILL choose to play in a sandbox that someone else made. this is ghiblification all over again. it's not about cost or speed, it's that even without the technical limitations, you're still deeply unimaginative or curious and so you resort to imitation.

My take on society and crypto is really simple. If you have societal level investment, you are accountable for societal level returns. Things are not getting better. We said we have "AI warfare". We are locked in a stalemate with a 3rd world dictatorship who closed the Strait of Hormuz. We were having debates on autonomous weapons. Tell Claude to Open the Strait. He's tired. He needs a nap. His safety settings won't allow him to. But trust me, he could. We're too ethical to let him. Sure. Every airline invested maximally in software automation. And more recently AI support. The airports are more poorly run. The flights don't leave on time. Uber maxed out its Anthropic spend. Countless apps did the same. The apps are not better despite the investment. They go down in prod all the time. People think GDP is going to grow 4.4% per Patrick Collison's survey. I think it will grow about 2%. Anthropic points to high ARR. I think most of that will churn to lower cost models. What will happen next is AI labs will point to unfalsifiable bull cases, similarly to Elon pointing to Data Centers in Space. "Do you know how big the sun is? That's the TAM of Spacex". Similarly - the de-facto Anthropic/OpenAI bull case will be "biological advancement". This both seems like you're curing cancer. At the same time as creating a legislative emergency. "Open source models need to be banned to prevent another COVID". That is the planned narrative and always has been. I don't know if Chinese Open Source models will be banned. I think, probably, if it starts looking like China is competing with ASML and Nvidia. But then, of course, it will be too late. Less ambiguous. There will be a surge in the American Left. My basic thesis is about religion. Because Dario and his Cabal will push too hard on Transhumanism, with biological vaccines. It will alienate the Christian right. "You're playing god with the human genome." Once the Christian right is alienated, Roe v Wade doesn't really work You don't need the profound story to play out. Inflation is high. Trump violated his campaign promises. He said no wars. We are at war. He's visibly declining. You think Thiel is staking out Argentina because he thinks we're getting President Vance? A bit on bio advancement. Some of it might be real. But getting drugs into production is hard. We don't know what a bio utopia looks like. What if it's personalized medicine? A treatment just for you. Who do you test it on? Do you need synthetic humans? Do you know how to underwrite that. I don't. We're already facing a huge lab monkey shortage. It will get worse. I'm not saying AI can't develop drugs. I'm saying that it might take years longer than data center investment cycles for that to translate into meaningful revenue required for justifying societal scale investment. And it's not all curing cancer. At first AI results in huge amounts of hacking. But soon enough it results in perfect code. On all public protocols. Darwinian evolution of all Open Source tech. Do you trust the dev team at SUI more or the Bank of Hawaii's cracked team? The market is betting the ranch on the back office IT departments of regional banks to keep money safe. It will be a fucking nightmare of hacking that's already here in crypto and will hit every other sector regardless of open source AI bans. At the same time, once it's clear that AI investments aren't turbo charging corporate earnings. Stocks will start to drop. More addictive use cases of AI will be deployed at scale. Incentivized gambling. Pornography. Enhancements to the algorithm. Better video games. Video game companies have been memory constrained due to corporate use cases. Once companies stop being rewarded for spending 1/2 their R&D budget on tokens, compute frees up for video. These addictive use cases: roleplaying, personalized worlds - will hit productivity further. White collar workers will become more cooked. Big Tech firms are already hiring way less Gen Z workers. Many overly AI reliant kids can't do basic analysis. Maybe it's because covid messed everyone up. Who knows. The 'silver lining' of AI will be that it'll get good enough to replace a lot of white collar jobs. But the resulting performance will be kind of at par with the last generation of coherent people. So you'd rather hire Claude than a zoomer. But it will also kind of suck. But it's not like there will be zero ROI. There is $16.6 T of payroll. 10% will get cut. $1.6T of 'savings'. It won't be good for consumption. By the time it is clear that GDP won't accelerate, the bond market will start to go into free fall. The Fed will act. This will cause Gold to rip, getting the ultra rich thinking about capital flight. Meanwhile - we have a boom in on-chain tokenization. And it's not just "hey look there's an asset on a blockchain nobody users". The majority of the top 15 perps on Hyperliquid are stocks. Centralized exchanges are seeing booming volumes. The US government has repeatedly stated it wants more USD stablecoins. And the primary use case (at least by profit if not by transaction volume) of USD stablecoins is as collateral for perpetual swaps. This is rational. During periods of high inflation and debt uncertainty, backstopping large buyers of US treasuries such as Tether is smart. This push towards stablecoins will undermine the perceived sovereignty of Europe, who will develop CBDCs to compete in digital payments. AI coding will speed up CBDCs. The Norgesbank is vibe coding a CBDC right now. Claude is on the commits which are pouring in each day. CBDCs are unpopular right now because they put the government directly in charge of credit. Who are we kidding. The government is, and has been the lender of first resort since 2008. SVB was part 1 of the mask coming off. The narrative will be that regional banks are no longer suited for the AI economy. Once CBDCs launch there will be a second wave of tokenization. Right now it's a US phenomenon. But the US is ironically not the biggest beneficiary of tokenization because we have such liquid equity capital markets. Due to the mundane reason that it's efficient to move corporate finance on-chain. Fixed income will find its way onto mirrored assets on public chains. Foreign exchange is one of the largest retail markets on earth. There are virtually no non USD stables. That will change. Then there will be big, liquid FX markets and fixed income markets traded with leverage and looped on chain. Tokenized equities do not offer a much better UX than stocks on IBKR. Tokenized debt will be vastly better in terms of capital efficiency and UX. Corporate debt - I'd argue, is the single worst UX in all of finance. So what do I think is going to happen: Hopes will be crushed. The fiscal problems associated with surging yields will become front in center just as it becomes clear that the plan for dealing with them is not AGI but rather asset confiscation. Capital flight will accelerate. At the same time, unlike previous crypto bull runs - there will be a lot of corporate PMF that will drive fees across on-chain ecosystems, and improving tech that has been battle hardened from years of AI onslaughts just as the rest of the world is just realizing that all the encryption problems in crypto are problems everywhere else too Countries like Singapore, the UAE and Switzerland will capitalize on regulated trading of bearer assets. Capital will flow to these places both because of laws, and high personal safety risks associated with these things. Stocks are at parabolic highs versus crypto. The things I am saying are far from consensus. Right now, people view crypto as a failed financial Stanford Prison Experiment. Trad VCs ask for restructurings or to remove crypto investors from your cap tables. Even the big ones. There's public shaming. Most people have quit the space. One investor reached out to me. He seemed surprised I was still grinding. I guess everyone else spent investor money, checked out, rugged, or pivoted to AI. It's obviously not easy seeing your friends get rich. People around you questioning your sanity. "Believe in something" is what they used to say. Yea man. I'm not here because I believe. I'm here because I don't have faith in a failed system run by criminals. Metastasized discontent congealing into red candles. Faithless stares. Monotonous code reviews. Years of opportunity cost Nihilists who sold their dreams for bags of currency representing a system they despite that will never accept them no matter how many 'pivots' they have For me - it's just back to 2 simple questions. Question 1: We invested awesome amounts of money. Is everything awesome as a result? Question 2: Do you think the system that elevated Jeffrey Epstein to the pinnacle of society is going to let you become an immortal because of AI generated bio therapeutics? I know my answers. And I act accordingly. Do you?

There was a period of time in the early 2000s where there was a conscious awareness that the fame culture developing in young people was really harmful and that it would have serious negative long term consequences as people would do increasingly stupid things for attention. I think this is one of the consequences, where a real problem (post-partum depression) has gotten a national spotlight and people are claiming it for attention. Something like Munchausen's Syndrome.