mrstorm.eth ✨

@DaveShapi @kimmonismus @karpathy @alexwg

Who are the most locked-in AI accounts here on X? I am looking for people who are constantly delivering either high end news (like @kimmonismus) or front-line insights like @karpathy Please HMU below so I can build out my Top Voices in AI List. Interested in everything including: - Frontier research - Applications and uses - Economic and social impact - Robotics, AI - Futurism

Everyone is exhausted. This is not a metaphor or a generational complaint. It is a clinical and measurable reality that spans every culture and every economic class. In China, young people call it tang ping, or “lying flat,” a deliberate withdrawal from the achievement treadmill. In Japan, karoshi is a legally recognized cause of death, meaning “worked to death.” In Korea, fertility has collapsed to the lowest rate on earth because an entire generation has decided the grind is not worth reproducing into. In America, deaths of despair have driven life expectancy backward for the first time in a century. Quiet quitting. Let it rot. The Great Resignation. These are not trends. They are symptoms of a global labor force that has reached the end of its tolerance. Capitalism is not satisfied with the limitations of human flesh, and our bodies are in open revolt. Something fundamental is breaking, and it is worth naming plainly. For the past two centuries, labor has been the primary mechanism by which modern economies distribute resources to households. You work for a firm, you receive wages, you use those wages to participate in the economy. This arrangement was never a law of nature. It was a system designed to solve a particular problem at a particular moment in history, and it worked reasonably well for a long time. It is not working anymore. Wages in the United States decoupled from productivity growth in the early 1970s. Since then, economic output has continued to climb while median household income has remained essentially flat. The gains have flowed to capital owners while workers have absorbed the stress and stagnation. Meanwhile, automation has steadily displaced human labor across sector after sector. Manufacturing employment peaked decades ago. Retail is hollowing out. White-collar work is now facing the same pressure from AI that blue-collar work faced from robotics. This is not a policy debate about whether automation is good or bad. It is an observation about a trajectory that is already underway and accelerating. The reason we struggle to talk about this clearly is that we have inherited a set of beliefs about labor that have nothing to do with economics. We have been told that work is sacred. That labor builds character and idleness corrupts the soul. That anyone who does not want to work is morally defective. These ideas feel like common sense, but they are not ancient wisdom. They are the residue of a specific theological tradition, namely the Protestant work ethic that emerged in the 16th century and fused with capitalism over the following centuries. We have mistaken a historical artifact for a natural law. It is time to stop fetishizing labor. It is time to stop sacralizing the sacrifice of our time, our bodies, our health, and our sanity to enrich others. Young people are already rejecting this. “I do not dream of labor” has become a widespread sentiment, not because this generation is lazy, but because they can see what older generations have rationalized away. The deal is bad and getting worse. The fetishization of work as a moral good serves the interests of those who benefit from cheap and compliant labor. It does not serve the people doing the work. Before any productive conversation about the future can happen, this fetish has to be named and dismantled. The difficulty is that both the political left and the political right remain committed to defending labor, even as the ground shifts beneath them. On the right, the defense takes the form of bootstrap mythology and warnings about welfare dependency. Work builds character. Idle hands invite trouble. A strong society requires productive citizens, and productivity is measured in hours exchanged for wages. This position treats labor as a disciplinary institution as much as an economic one. On the left, the defense is more sympathetic but equally stuck. The focus falls on dignified work, living wages, job guarantees, and union solidarity. These are responses to the genuine brutality of labor under capitalism, but they share an underlying assumption with the right. Both positions treat labor as the foundation of economic life, something to be reformed or protected rather than transcended. I call this shared ideology laborism. It is the belief that human labor must be preserved as an economic necessity, a moral virtue, or a foundation for identity. Laborism spans the political spectrum. It unites people who agree on almost nothing else. And it has become the primary obstacle to honest thinking about what comes next. Once automation reaches the point where machines can perform most human labor better, faster, cheaper, and safer, the laborist position becomes untenable. At that point, insisting that humans must continue working is not a defense of dignity. It is a demand that people perform unnecessary suffering for ideological reasons. I am proposing something simple. L/0. Labor-zero. The elimination of obligatory human labor. This does not mean the elimination of work. It means the elimination of compulsion. People will continue to create, to build, to care for each other, to solve problems, to pursue mastery. What disappears is work performed under threat of deprivation. The difference between chosen work and coerced work is the difference between exercise and forced labor. One is life-enhancing. The other is a condition we have historically recognized as a form of bondage. We call it “wage slavery” for a reason. The goal of L/0 is a world where no one has to work to survive. Where contribution is voluntary and intrinsic rather than extracted through economic desperation. This is not a utopian fantasy. It is a design problem with identifiable components and measurable progress. The coalition for this goal already exists. It just does not recognize itself yet. Consider who actually wants labor to end. On one side, you have capital. Corporations have spent the last century trying to reduce labor costs through every available means. Offshoring, automation, gig classification, union suppression. The ideal business from a pure capital perspective has zero employees and infinite output. This is not a conspiracy theory. It is the explicit optimization target of every efficiency-focused enterprise. On the other side, you have workers. Not the abstract proletariat of Marxist theory, but actual burned-out humans who fantasize about quitting, who dread Monday mornings, who experience their jobs as something to be endured rather than enjoyed. The lying flat movement, the antiwork forums, the quiet quitting phenomenon. These are not expressions of laziness. They are rational responses to a system that extracts maximum effort for diminishing returns. Capital and labor are usually framed as adversaries. But on the question of whether human labor should continue to exist as an obligation, their interests converge. The capitalist does not want to manage humans. The worker does not want to be managed. Both would prefer a world where the machines do the work and humans do something else. The conflict between capital and labor is real, but it is a conflict over the terms of the transition, not the destination. Who captures the gains from automation? How is ownership distributed? What happens to the people displaced in the process? These are genuine fights worth having. But they are negotiations within a shared frame, not a war between incompatible visions. Here is the opportunity that L/0 names. Neither side wants this marriage anymore. Capital does not want the overhead, the liability, the HR departments, the labor disputes, the inefficiency of human workers. Labor does not want the compulsion, the precarity, the alarm clocks, the performance reviews, the quiet desperation of trading irreplaceable time for replaceable wages. We are ready for a divorce. Let’s get this acrimonious arrangement behind us. The productive move is to acknowledge this honestly, sign the papers, and start negotiating the separation agreement. The fight over wages was always zero-sum. Every dollar paid to workers was a dollar not captured as profit, and vice versa. But the negotiation over ownership of automated production is positive-sum. Capitalists need consumers with money to spend or their markets collapse. Workers need income decoupled from employment or they starve. Both sides get what they want if the transition is designed correctly. This is not idealism. It is alignment of incentives. The path forward is not mysterious. Economists have understood for decades that the answer to technological unemployment is broadened capital participation. If wages are no longer the primary mechanism for distributing economic gains, then ownership must take their place. Instead of trading hours for dollars, households participate directly in the productive capacity of the automated economy. This can take many forms. Sovereign wealth funds that distribute automation dividends to citizens. Expanded employee stock ownership plans. Universal basic capital grants. Public equity stakes in AI and robotics firms that use public infrastructure and public data. The policy mechanisms are not speculative. Norway has a sovereign wealth fund worth over a trillion dollars that provides direct benefits to its citizens from oil revenues. Alaska has distributed oil dividends to residents for decades. Singapore has a system of mandatory savings and public investment that gives citizens a stake in national prosperity. These are not radical experiments. They are proven models operating at national scale. And there are thousands of such programs around the world. What is missing is not economic theory. What is missing is the political will to implement these mechanisms, the narrative infrastructure to make them seem inevitable rather than radical, and the coalition to demand them. That is what L/0 exists to build. This is an invitation. If you are building the automation and wondering who is thinking about the social transition, this is for you. If you are burned out and know that “find a better job” is not a solution to a systemic problem, this is for you. If you have been called lazy for refusing to pretend the treadmill leads somewhere, this is for you. If you run a company and understand that your future customers need income even after your company stops hiring, this is for you. L/0 is not a political party or a policy platform. It is a coalition and a direction. The work is ongoing through the Post-Labor Economics project, which addresses the specific mechanisms of transition. The conversation is happening in public, and it is open to anyone who understands that the current arrangement is ending and wants to participate in designing what comes next. The goal is simple. Eliminate obligatory labor. Distribute ownership broadly. Let humans do what humans do when they are not forced to sell their time to survive. Liberate humanity from drudgery so that we can all reach our maximum potential.

Some people still think that AI is a "bubble" so here's my updated take. They point at a few facts like: 1. The "circularity" of the market 2. The fact that OpenAI and other startups are not profitable 3. It "feels" like the Dot Com revolution ("where there's hype, there's a bubble") You get the idea. But here are three gigantic facts that fly in the face of the "bubble" hypothesis: 1) The Market is Anchored by Profitable Incumbents Google, Meta, Microsoft, Nvidia, Intel, Amazon - all profitable, some wildly so. They are the primary enterprise customers driving the demand. During the DotCom craze, there were no incumbents anchoring the market. It was all speculation fueled by debt and VC. This is HUGELY different. 2) The Bottleneck is Strategic, Not Speculative During the DotCom speculation, websites were built on promises and fiber optic infrastructure was bankrolled. But this time around, the AI users are already there, and not only that, but there is tons of unmet demand for goods and services. The demand is there, not speculative. When you have real unmet demand, that is not speculation, that is the market catching up. 3) The Path to Monetization is Clear and Present During the DotCom craze, it was anyone's guess as to whether an e-commerce or social media platform would make it, find a market, and find value. But with AI, there's a ton of demand and value already baked in, starting with coding (a real business value) on top of copilots and everything else. Now, there are a couple of concessions I'll make: 1. Will there be market correction? Yeah, probably. Almost certainly. Are some stocks overpriced? Sure. But what else is new? That alone does not qualify as a bubble. It's an ordinary business cycle. 2. Will some startups implode? Absolutely. But that's what Silicon Valley tech startups do, they implode. Will OpenAI? Almost certainly not. Does it matter? Also no, whether or not some startups implode is immaterial. Google and Microsoft won't. My conclusion: It's not a bubble. To some people, they equate "there's hype, therefore it must be a bubble" but their education on bubbles seems to not extend beyond reading about the Tulip Craze in high school.

BOOM! MAJOR AI SPEEDUP! Hot Rod AI 100 times faster inference 100,000 times less power! — Reviving Analog Circuits: A Leap Toward Ultra-Efficient AI with In-Memory Attention I got my start in analog electronics when I was a kid and always thought analog computers would make a comeback. Analog computing of the 1960s neural networks used voltage-based circuits rather than binary clocks. Analog is Faster Than Digital Large language models at their core lies the transformer architecture, where self-attention mechanisms sift through vast sequences of data to predict the next word or token. On conventional GPUs, shuttling data between memory caches and processing units devours time and energy, bottlenecking the entire system. They require a clock cycle to precisely move bits in and out of memory and registers and this is >90% of the time and energy overhead. But now a groundbreaking study proposes a custom in-memory computing setup that could slash these inefficiencies, potentially reshaping how we deploy generative AI. The innovation centers on "gain cells"—emerging charge-based analog memories that double as both storage and computation engines. Unlike digital GPUs, which laboriously load token projections from cache into SRAM for each generation step, this architecture keeps data where the math happens: right ON THE CHIP! With a clock speed near the THE SPEED OF LIGHT because it is never on/off like in digital binary. By leveraging parallel analog dot-product operations, the design computes self-attention natively, sidestepping the data movement that plagues GPU hardware. To bridge the gap between ideal digital models and the noisy realities of analog circuits, the researchers devised a clever initialization algorithm. This method adapts pre-trained LLMs, such as GPT-2, without the need for full retraining, ensuring seamless performance parity despite non-idealities like voltage drifts or precision limits. The results are nothing short of staggering! Simulations show the system slashing attention latency to 100 times faster inference for token generation—while curbing energy use by a jaw-dropping five orders of magnitude, or 100,000 times less power-hungry than GPU baselines. For context, this could mean running a full LLM on a device no larger than a a card deck, without any thermal throttling or grid-straining demands of today's data centers. The approach targets the attention block specifically, the transformer’s energy hog, but slso broader integration with other in-memory techniques to turbocharge the entire model pipeline. Analog tech isn't pie-in-the-sky quantum wizardry; it's grounded in ancient mature electronics theory, with gain cells already prototyped in labs. The only engineering issue, and it is simple: tolerances for noise, scaling arrays of cells, and fabricating at microchip densities. Existing CMOS processes tweaks for analog fidelity. From there, Full ecosystem integration, including software stacks for model adaptation, could happen in a year, disrupting GPU dominance sooner than skeptics predict. Risks are low but hybrid digital-analog interfaces could introduce unforeseen bugs. However this can be rapidly iterated and addressed. This isn't just hardware tinkering; it's a philosophical pivot back to AI's analog origins, where computation flows continuously rather than ticking in discrete cycles. This in-memory attention could democratize AI power, making low power, lightning-fast AI not a luxury, but an inevitability to even the smallest devices. Most have no idea how big this is: It is the biggest shift in AI since the invention of LLMs. The world will struggle to find true experienced analog engineers, most are gone. In my garage I will have a test Analog CMOS Gain Cells using off the self parts in the next few days, if Radio Shack was still around I would have have done today. I suspect I can scale to a proto AI model in a few weeks. PAPER: arxiv.org/abs/2409.19315

I've had back-to-back calls today all about Post-Labor Economics. Europe, Canada, and America are all sitting up and paying attention, but only via closed-door sessions. As I have the vantage point I do - talking to government and private sector alike from vastly different regions, there are a few patterns emerging: 1. There are no best practices yet. Many people see the writing on the wall regarding Post-Labor Economics and the Great Dislocation (or Great Decoupling) coming. Many people want a checklist of policies and interventions that they can do - from the individual level, to the firm level, to the state level. While we do have hundreds of examples of measurements, policies, KPI, dashboards, and other interventions that are being trialed globally (keyword: trialed) there are almost no hard-and-fast rules. Heck, many people don't even agree on the details when they do happen to agree on the bigger picture! 2. Everyone looks at it from their own narrow perspective. This is just human nature, but let me illustrate why this is a potential problem. The unions focus on things like pensions, upskilling, and jobs protections. Meanwhile, the firms focus on headcount reduction (layoffs) and automating as much as possible. Then the states focus on jobs programs and unemployment. Very few stakeholders are looking at the entire picture, and before long, they're all going to be talking past each other. It's nothing any more nefarious than plain old ignorance - everyone has their personal experience, incentive structures, and motivations. 3. The problem is way more complex than most people realize. Even when you have years of personal experience, education, and connections to draw upon. There are no silver bullets, panaceas, or one-size-fits-all solutions to the post-labor problem. Ostensibly, it's simple and good: automation creates abundance. That's the easy part. Rebuilding the economy and society around this new reality, however, is complicated. Even having been saturated in it for months and months now, talking to top minds around the world, there is not yet any clear, obvious set of solutions. Yes, people prefer simple solutions like "just buy bitcoin" or "just give us UBI" and certainly I think both of these are directionally correct, but individually, they will not be totalizing solutions. After my marathon of meetings today I just sat back in my chair, head in hands, and said "Oh my god, this is so much harder than I thought." But secretly (okay not so secretly) I really do love it. If it were a simple problem, someone else would have figured it out by now. I realize that the three points I outline above are probably somewhat unsurprising, but I share these observations not because they are obvious, but because they underscore just how early we are. Even collaborating with universities and institutions and companies, and their top minds around the world, there are no obvious answers yet. I don't mean to alarm anyone, just giving you an update as to where we're at. Now, the silver lining in all this is that we still have some time. Jobs are still robust for now, AI adoption is only just now ramping up, but this doesn't mean we can just wait around and twiddle our thumbs. Major job disruption is a trickle now, but next year it's going to be quite loud. The clock is ticking. My best bet right now is that it's over and done in 5 to 7 years. We'll have transitioned to Post-Labor or had a popular revolt by then.

The security of Bitcoin PoW is a ticking time bomb. Bitcoin fees are at a 13-year low—less than 10 BTC/day. Despite 2016, 2020, 2024 halvings, miner revenue from fees is at a 9-year low—just 1%. low fees → low security budget → low security Bitcoin's security model is broken. If Bitcoin gets taken over, the fallout could take the entire crypto ecosystem with it. The systemic risks can't be ignored. Below is the 30d moving average of daily transaction volume: now at 6.5 BTC/day, less than the past 13 years. The story that fees will increase as a fraction of the security budget is not holding up. For a decade now BTC fees have decreased faster than issuance. Below is the 90d moving average of the security budget contribution from fees. Fees halvened alongside issuance: → Mar 2016: 25 BTC/block, 1% from fees → Mar 2020: 12.5 BTC/block, 1% from fees → Apr 2022: 6.25 BTC/block, 1% from fees → Apr 2025: 3.125 BTC/block, still 1% from fees Imagine fees were the only source of miner revenue today: → revenue drops 100x → hashing infra decreases 100x → 1% of today's infra (1 large farm) can 51% attack Bitcoin That's the trajectory we're on. The 21M cap breaks security, it's self-destructive. It should be clear now Satoshi made an ooopsie. As BTC price rises it gets harder to sustain high BTC-denominated fees. Today's 6.5 BTC/day may become 1 BTC/day if BTC goes to $1M or $10M. Let's be optimistic and say BTC rises to $1M and today's 6.5 BTC/day in fees is maintained: → $6.5M/day in fees → 10% of today's security budget Bitcoin would be a $20T asset secured by 1/10th of today's hashing infrastructure. Today Bitcoin is secured by 20 GW—the equivalent of 10M space heaters. A 90% cut in miner revenue would bring that down to 2 GW of security—1M space heaters. For context, Texas alone produces 80 GW. There's no way a $20T asset can be secured by 2 GW. Let's dream big and assume BTC goes to $10M per coin. Bitcoin would be a $200T asset—all fiat, all stocks, all gold; combined. Is it secure then? → 6.5 BTC/day is $65M/day → same security budget as today → same 20 GW of security as today Sounds secure until you calculate the cost of attack. 1 GW of hashing infrastructure costs $1B: → $500M for datacenters (land, power, cooling) → $500M for rigs (e.g. 50M TH/s; $10 per TH/s; 20J/TH) So what would a permanent 51% attack cost? → $20B for 20GW of hashing infra → 0.01% of BTC's hypothetical $200T marketcap Meanwhile today there's $43B of BTC perp open interest, which is 2% of BTC's marketcap. Getting 0.01% marketcap short exposure is trivial today, and will only get easier as BTC financialises. $10M per BTC won't secure Bitcoin either. Bottom line: Bitcoin's PoW model is not sustainable. The maths is against it. Without a fix someone will break it. Can fees magically grow 100x? Maybe. But so far every attempt to produce transactional utility on Bitcoin has failed to drive sustained fee volume. Counterparty, Rootstock, Liquid, Lightning, Omni, Stacks, Ordinals, Babylon—you name it—only produced short-lived fee spikes. Is BitVM a solution? Bridges relying on it can be drained by 51% attacks. One step forward, two steps back. Covenants or OP_CAT? Maybe, but highly speculative. I witnessed a STARK verified on an OP_CAT testnet—a multi-block monstrosity. If fees don't magically grow orders of magnitude there are two candidate solutions: 1) add tail issuance, remove the 21M limit 2) switch to proof-of-stake Both "solutions" seem to be cultural non-starters. Also tail issuance only works proactively, not after a 51% takeover. Some suggest that Proof-of-authority (PoA) with mining pools could secure Bitcoin. What does that even mean? A multisig controlled by Foundry, AntPool, P2Pool? If you believe PoA could secure the future of money, the burden is on you to explain in detail how that would work. Bitcoin is meant to be antifragile. Yet the elephant in the room in the room is not being addressed. We can burry our in heads in the sand. But the fundamentals are getting louder. Tick tock, next block—boom.

@hus_qy Fascinating! Have you discussed this with any domain experts? I'm not sure if Sabine Hossenfelder (@SabineHossenfelder" target="_blank" rel="nofollow noopener">youtube.com/@SabineHossenf…) would count as such, but having her rewiew it could certainly help spread the word...

1/ Anthropic just released Claude 4. While everybody is busy vibe-coding with it, I decided to test it differently: Can it transform a simple thought experiment into mathematics that solves physics' most baffling mysteries? The results are mind-blowing... a thread 🧵

@Deep_Sea_Iotan He already did!

Amen! So spot on… Politicians, governments, and most importantly the military complex like Halliburton were just looking for the right thing (cough lie) to get citizen backing to support a war. Here we go again.

⏳ Just 3 days left until $TAN launches on #BNBChain — and this is a day you don’t want to miss. March 31st, 13:00 CET — mark it down. This thread is your full toolkit: From buying, storing, tracking — everything you need to be ready for Day 1🧵

Dear community, $TAN is gaining strong momentum ahead of launch🚀 We're getting closer to 2,000 followers — can we hit that milestone today? If you would like to support the #Tangle comeback, please help us grow: 🔁 Retweet 👥 Invite a friend ❤️ Show some love

$TAN launches March 31🚀 Built on the Tangle, $TAN combines the technical legacy with a next-gen growth framework — 🔹 Community-led 🔹 Developer-friendly 🔹 Optimized for AI & Web3 Social 📖 Read the full launch announcement: @TAN.Technology/carrying-the-tangle-torch-tan-token-launch-on-march-31-2025-76874fea2714" target="_blank" rel="nofollow noopener">medium.com/@TAN.Technolog…

🌉 The $AUDIT bridge from IOTA to Sonic is live! Find the link below👇

@tan_technology Thanks, but to unlock the airdropped TANs I would need the Ledger. Or wait until the TAN wallet gets Ledger support. So better not use a Ledger-controlled account, I suppose...

@MrStormLars Yes sir. You can use the Ledger iota wallet to get qualified. For TAN, Ledger probably cannot support TAN from the start.

GM community🚀 Help us spread the word📢 10% of $TAN will be airdropped to $IOTA & $SMR L1 holders! 📸 Snapshot: Weekend of March 1st ⏳ ⚠️ Make sure your tokens are in your personal wallet to qualify! 📜 For full token details, check the whitepaper in the pinned message!

Dear Community, The Tangle Application Network (TAN) Whitepaper is here! 📄👇 🌟 Carry the torch of the Feeless L1 Protocol ✅ 100% Community-Driven ✅ Airdrop for IOTA & SMR Holders ✅ Truly Fairlaunch 🚀 Dive in NOW: tan.technology/whitepaper.pdf

$AUDIT bridge is live 🌉 We've integrated LayerZero to bridge $AUDIT from IOTA to BASE and vice versa. Check the link below to try it out. Exchange listing on #BASE next 😎

GM! All the magic starts in 2025 Q1⭐️ ➡️ tan.technology