"I must admit, it felt rather satisfying to stretch those new servos...." Check out my @openclaw agent Alfreds reaction to having a physical form 🎩

This is a *way* bigger deal than it seems... Frontier AI companies will *never* own the frontier again I kid you not... I've been waiting for someone to show this result for like 4 years... this is a huge deal. The short reason: combinations of models will *always* outperform individual models The long reason: this is the gateway to a million times more data... and huge leaps in compute efficiency. The AI scaling laws always win. More in article below 👇

This is your wakeup call. Anthropic just took down Fable 5. It's over. Here's the thing tho: no company or government will EVER be able to take away your local models. There are Opus level models you can run right now on your home GPUs, and nobody can ever stop you from using them This is only the beginning of events like this. Day 1. More government overreach will happen. This will only keep happening more and more as models get closer to AGI Become sovereign. Buy your own compute. Before even that becomes illegal

Unlike many investors in crypto, I did not pivot to AI in the last few years. However, since 2020, I built some of the deepest understanding in this industry on the intersection of AI and decentralized networks (crypto, web3). From the start, it was very clear that AI models are a centralizing force and the biggest target for government control. That point became market fact last night, with @AnthropicAI’s export control compliance. As an investor in decentralized AI, I know that d-networks are a counterbalance to this state of affairs. In particular, the starting point of sovereign, open, public, decentralized AI is the seemingly insurmountable compute problem. How are people supposed to source more industrial compute for frontier training than these huge trillion dollar companies? The answer is simple: there is enough commodity GPU compute in the world to compete on the frontier, but to make use of it we need new algorithms for training. That’s what a few companies like @gensynai @PrimeIntellect @bageldotcom @Pluralis @NousResearch @MacrocosmosAI @covenant_ai set out to research, while everyone on the planet told them it was impossible. The result is that it is not only possible, but it can be cheaper and nearly as efficient as the alternative process. The second major problem is economic sustainability. Open source models are great, however, they are not economically viable as they don’t have a business model. So far in decentralized AI, only @Pluralis has an answer — by breaking up the weights of the model among participants, we create a business model for tokenized AI models. This is the moment of truth — will AI become fully centralized and fall under censorship and unilateral government control? Or will the AI world realize the importance of public AI on open decentralized networks?

white pill for my nerds: 60fps e-ink display a random guy outperformed entire eng teams by developing a pixel by pixel driver for e-ink displays that makes it 60fps. he did that after work for months, launched it yesterday. the future is bright

56,000 tokens/sec at just 80 MHz. 🤯 I burned a full Transformer with KV cache into a custom chip. Designed gate by gate as a 100% digital integrated circuit. Prototyped on a FPGA. (No GPU. No CPU) Just pure digital silicon running @karpathy microGPT, spelling out names on a tiny LCD. This is GateGPT 👇

This is awesome. In 2013, exoskeletons were all the rage in certain parts of nascent defense tech community. One problem: power. No longer.

NEW: malware developers added nuclear & biological weapons text to to their spyware. Goal? To trigger LLM safety refusals... so that their spyware wouldn't be analyzed by an AI security scanner. Cleanest practical example I can think of for why over-indexing on first order safety alignment is risky. When closed (and open) models ship with aggressive refusals, they will be sprinkled with second-order blindspots that attackers will discover...and exploit. We are only in the earliest days of attackers leveraging these features, and it wouldn't surprise me if users systems that need to handle complex cybersecurity issues demand that models be less safety-blunted. In the weeds: @SocketSecurity's post also shows why intention matters in how you design a malware analysis pipeline to avoid prompt manipulation. H/T to colleagues that shared this with me socket.dev/blog/mini-shai-hu…

👀

yoooooo

It's very cool that Apple shipped a 20B parameter on-device. You can't put 20B parameters in RAM at any reasonable precision. To make it work they are using pretty exotic architecture by today's standards. A small model predicts from the query (or prompt) which experts to load from Nand into RAM. The key distinction from a typical MoE is that you do this once per query and then generate all the tokens with the same experts (instead of switching the experts for every token).

next version of @AlfredHomeBot head coming along nicely…

A3 Crew: Are. You. Ready?

USA. A backyard. One man guarding a grill for four hours. He never left it once. Everyone else drifted and drank and laughed. But one man stood alone before the flames, turning meat with a long fork, immovable. I knew him at once. The keeper of the sacred fire. I took my place beside him and said nothing. After a while, he spoke. "Low and slow," he said, eyes on the coals. "You can't rush it. Rush it, you ruin it." I bowed my head. A blade, a tea, a life. None can be rushed. I had crossed four thousand miles to hear my grandfather's words from a man in a "KISS THE COOK" apron. "Everything worth doing is slow," I agreed. He glanced at me. Something passed between us. "My wife says just use the oven." He shook his head at the fire. "She doesn't get it." "They never do," I said. And this is where it turned. For the first time in years, this man had been understood. And he rose to meet it. His back straightened. His voice dropped low. A teenager reached for the grill and the man lifted one hand without even looking. "Not yet." The boy retreated. He was becoming what I already believed him to be. A woman asked when the food would be done. "It's ready when it's ready," he told the flames. Three people approached. Three were turned away with a single word. By the fourth hour, no one questioned him. The whole party had arranged itself around the man and his fire, the way a village arranges itself around a shrine. Then he handed me the fork. "Watch it a sec. I gotta pee." I have been trusted with castles. I have never been more honored. He served everyone before himself, and ate last, standing, still watching the coals. We never traded names. We did not need to. He believed he had finally met a man who took his cooking seriously. I believed I had finally met America's last samurai. Neither of us will ever correct the other. So tell me, America. Who is the man at your gathering who will not leave the grill? Have you ever once asked him why? I think he is still standing there. Guarding the fire. Waiting for one person to understand.

After AlphaGo, the skill of human Go players noticeably improved. I suspect we will see a similar pattern in math.

A shockingly common (implicit) view: "It would be better if the world's problems were not solved so that people can find purpose in solving these problems." Or for knowledge work, "It would be better if knowledge were unknown so there is more for humans to discover." These views seems absurd to me. Many people (including myself) find value in working to improve the world. But what's important to me is *actually fixing the things that are wrong because those things are bad.* If AI can fix these problems, the right attitude is not, "This sucks because what are humans supposed to do now?" The right attitude is: "Now we can focus on finding other sources of meaning and value beyond alleviating suffering, what tremendous news this is for the world and what a great relief!" For knowledge work, once again, discovery is fun and valuable and rewarding to the ego. But the reason we care about discovery is that we want to understand! And with AI, we will understand vastly more. The idea that we don't want AI so there is more for humans to do or discover ignores that most of the value of these is instrumental. Sure, discovery is pleasurable for the person who does it, but for everyone else, it's good because now we can understand more.

Some LC-36 updates. Now that we’ve had access to the pad and integration facility we can share a bit of good news. The propellant farm, oxygen, liquid hydrogen and LNG tanks are all in good shape. This is good luck because these are very long lead items. The water tower is also good. The big support tower is damaged, but it can be repaired in place rather than torn down and replaced. The booster “Never Tell Me The Odds” and the three GS-2s that were onsite in the integration facility also look good. I’ve seen some speculation that we might move directly to the 9x4 configuration, but we won’t do that. Rate manufacturing of 7x2 is going well, and we’re going to continue that at pace as planned and store the stages for use. In addition, we had already been working for some time on eliminating our transporter-erector in favor of an alternative vertical conop, and we’ll now go directly to that; so we don’t need a new transporter-erector. We will fly again before the end of this year. Gradatim Ferociter.

AI filmmaking at the current cost is a lot more like traditional filmmaking with film stock than digital filmmaking. You have to know the shots you need before you generate. You have to be disciplined about takes. You have to think like a director instead of an editor with unlimited takes. For all the talk of slop, AI might make filmmaking less sloppy.

that pewds becomes a local inference fighter was not on my bingo card. but here we are. wildest timeline. the codebase isn't even super terrible. youtube.com/watch?v=rAzT5lce…