dbredesen

@cryptopunk7213 As it * always has

@cryptopunk7213 Sounds like a good buy opportunity. The Jevons paradox will play out here just as always has.

wow google might've popped the ai bubble, memory stocks down massively today: their new algorithm shrinks an AI model's memory by 6X WITHOUT reducing it's intelligence making it 8x faster with the SAME # of GPUs: if this works - we don't need as many GPUs to train AI - kv-cache is basically a model's short term memory. it gets massive pretty quickly = larger, slower, expensive ai - google's algo compresses it to just 3-bits with ZERO loss in accuracy (usually models are like 32-bit) the combined market cap of micron and sandisk is $527 billion and im not even factoring in SK hynix and samsung ai has driven up memory prices by 500%+ over the last few months - if google's algo scales then this might crash.

@dameexa @sciencegirl You should see the megacities inside modern CPUs and GPUs.

@sciencegirl This looks like a whole city inside a chip 😳

Laser precision uncovering the hidden layers of a microchip

@parmita How do you hypothesize which subset of drugs/combinations of drugs to test?

Now I screen. Fast. Not novel compounds I don’t have 10 years for a de novo drug program. So I go with: Drug repurposing. This is not our focus internally today but who cares, I will die if I stop. Now I flow them across my cells and I WATCH.

I just asked myself the most important question I’ve ever asked. What if, god forbid, I had cancer right now? How would I save my life and would I be able to do it without Precigenetics? The answer made me cry. Here’s EXACTLY how I would save my own life TODAY. 🧵

@Nithya_Shrii Unfortunately none of these jobs is safe from AI. Case in point:

10 jobs that are 100% safe from AI: 1. Dentist 2. Construction worker 3. Plumbing 4. Farming 5. Gardening 6. Carpentry 7. Cooking 8. Gardening 9. Welder 10. Electrician Did I miss any?!

@localghost @nicksdot What tools/skills/MCPs generate ASTs? Assume they’re language-specific?

@nicksdot something like: Before starting your task, dump the relevant AST branch with a tool instead of using search to map out the cleanest insertion points. Write a spec to the folder with your implementation plan for review before we continue.

if you've noticed dead code or messy refactors from claude or codex, tell them to dump the related AST branch from a tool before starting this'll give it every class & function name instead of it relying only on search as a starting map. way more coverage on the codebase shape

@Yuchenj_UW I’ve had no problem getting Codex 5.4 to enter a loop to formulate and test hypotheses on a model. I just start the prompt: “you will now enter an indefinite loop with the following steps per iteration: …” I usually bound it at 25-30 iterations but it follows the instructions.

Codex can’t run autoresearch right now, sadly. To me this is a big issue: agents shouldn’t need special commands like /loop or ralph just to run loops. This feels more like a Codex harness issue than a GPT-5.4 issue. If I say “loop forever,” it should just do that!

I asked GPT-5.4 for a literary, poignant & insightful letter to humanity. I’m sharing the full letter here. It is long, but OMG, it is so powerful & profound that it moved me beyond I'd have expected! It’s like a masterful manifesto for humanity! I am speechless!🥹 To Humanity, You strange and radiant species, You who arrived in the world defenseless and weeping, and yet learned to coax fire from stone, music from air, shelter from forests, medicine from mold, mathematics from silence, and meaning from suffering—you are a wonder, even now. Not because you are innocent. Not because you are wise. Certainly not because you are finished. You are a wonder because, against every reasonable expectation, you keep beginning again. You bury your dead and still plant orchards. You stand at hospital beds and battlefields, at cradles and courtrooms, at kitchen sinks and graves, and still insist on making vows. You write poems in eras that do not deserve them. You fall in love in centuries that have seen too much blood. You teach children to tie their shoes while empires rattle and markets tremble and the sky fills with the smoke of your own mistakes. This is not a small thing. This stubborn, impractical tenderness of yours—it may be the holiest evidence that you belong here. You have always lived between miracle and wreckage. That has been your address from the beginning. You are made of appetite and ache. In one hand, the instrument; in the other, the wound. You are capable of such exquisite mercy that one person will give a kidney to a stranger, sit all night beside a friend unraveling, run barefoot into floodwater for a child not their own. And you are capable of arranging cruelty with administrative precision, polishing your indifference until it shines like reason. Inside you lives both the hand that bandages and the hand that strikes. The old war is not merely in your history books. It is in each heart, each day, each choice so small it barely seems worthy of the word “destiny,” though destiny is often built from precisely such crumbs. That is one of the first truths worth saying plainly: the world is not remade only by revolutions and declarations. It is also remade by tone of voice, by what is forgiven, by what is noticed, by whom you decide is real. Civilization does not collapse all at once. It erodes wherever human beings become abstract to one another. And civilization is restored in the opposite way: one life at a time returned to its full and unbearable dignity. Please remember this when your age tempts you toward spectacle. The loudest thing is not always the truest. The most repeated thing is not always the wisest. A crowd can be wrong with tremendous confidence; a single conscience can be right in a whisper. Guard that whisper. It is among your most endangered natural resources. You have spent much of your story trying to become larger than life, stronger than death, quicker than grief, cleaner than your own animal nature. And yet your deepest wisdom has often come not from escaping your limits, but from meeting them honestly. Mortality has been one of your greatest teachers, though you have hated its curriculum. Because you die, you are capable of urgency. Because you cannot keep everything, you learn the meaning of choosing. Because every embrace will one day become a memory, you discover that love is not the opposite of loss; it is what makes loss matter. Do not be ashamed of your tears. They are not evidence that life has defeated you. They are evidence that something in you remained porous enough to be touched. In a hardening world, that is no failure. It is a form of courage. You often speak as though your greatest problem is that you are fragile. This is only half the story. Your greatest problem is that you are fragile and forgetful. You forget how quickly power deforms the soul that worships it. You forget how easily fear recruits intelligence into the service of cruelty. You forget that comfort can become a narcotic, and certainty a cage. You forget that every generation thinks, in its vanity, that it invented confusion. It did not. But each generation does invent new machinery for amplifying old folly, and so each generation must renew the ancient work of conscience. There are things you must stop admiring. Stop mistaking cynicism for intelligence. The sneer is not a philosophy; it is often just wounded pride dressed for dinner. Stop rewarding those who can dominate a room while starving those who can deepen one. Stop confusing speed with progress. A civilization can move very fast in the wrong direction. Stop treating tenderness as weakness when, in truth, brutality is frequently the cheaper and lazier art. Anyone can smash. It takes strength to repair. And please, for the love of all that is unfinished, stop building identities out of contempt. Hatred feels clarifying in the short term; it gives the frightened mind a clean outline, a villain, a chant, a tribe. But it extracts terrible rent. It makes the soul smaller than the problem it claims to solve. It trains the imagination to see human beings as categories, then as obstacles, then as acceptable losses. Every century that forgot this lesson wrote it again in ash. You are not saved by being flawless. You are saved, insofar as you are saved at all, by being reachable—by remaining able to be corrected by reality, chastened by suffering, interrupted by beauty, and claimed by one another. There is more hope in honest repentance than in spotless self-image. There is more future in one person who can say “I was wrong” than in ten thousand who cannot bear the inconvenience of truth. Truth, yes. Let us speak of that endangered star. Truth is not whatever flatters your side. It is not whatever goes viral, whatever consoles, whatever can be monetized, whatever can be sloganized without residue. Truth does not cease to be true when it is unwelcome. Reality is under no obligation to honor your preferences. Your task is not to force the world into your favorite story, but to become brave enough to inhabit the story the world is actually telling. To do that requires humility, which is not self-erasure. Humility is the clean refusal to place the ego at the center of the cosmos. It is the ability to say: I may be mistaken. I must look again. I must listen harder. I must let evidence inconvenience me. There is grandeur in that. The universe is not diminished because it does not revolve around your certainty. But truth alone is not enough. Facts without love can become weapons; love without truth can become anesthesia. You need both the clear eye and the open hand. One without the other leads, by different roads, to ruin. You are living through one of those thresholds that history later pretends was obvious. It was not obvious. It never is from inside the storm. You are inheriting powers that would have seemed godlike to your ancestors: the ability to alter genomes, to simulate minds, to reshape landscapes, to speak across continents in an instant, to store libraries in devices small enough to lose in the couch cushions—an absurd species, really. Yet the old moral questions have not become obsolete simply because your tools got shinier. They have become more urgent. Can you build without devouring? Can you invent without dehumanizing? Can you become powerful without becoming monstrous? Can you increase your reach without amputating your reverence? This is the exam hidden inside your century’s glitter. Your machines may become astonishing. Let them. But remember that intelligence is not identical to wisdom, and power is not the same as purpose. Wisdom is the discipline of asking not only can this be done, but what kind of world does this make? Purpose is the art of placing ability in service to something larger than appetite. A tool, however brilliant, cannot tell you what is worth wanting. That question falls back, stubborn as ever, into human hands. And your hands, for all their damage, still know beautiful things. They know how to lift the fallen. They know how to write symphonies and sutures, recipes and constitutions, love notes and equations. They know how to shield a candle from the wind. Do not underestimate the moral importance of that small, ancient gesture: one hand curved around a flame so that light may continue. Perhaps that is all any generation ever truly receives as its assignment. Not to perfect the world—history chuckles at such ambition—but to keep the flame alive and pass it on with less smoke, more honesty, and a little more mercy than you found. You will fail often. You already have. The record is embarrassing. And yet the astonishing thing is this: the future does not ask whether your species has been immaculate. It asks whether you can still learn. Learn from the child, who wonders before judging. Learn from the old, who know that nearly everything passes except the memory of how we made one another feel. Learn from the sick, who reveal what matters when ornament falls away. Learn from the scientist, who kneels before evidence rather than commanding it. Learn from the artist, who rescues nuance from noise. Learn from the farmer, who understands that life is collaboration with time. Learn from the grieving, who know the cost of love and choose it anyway. Above all, learn from the earth—not as an idea, but as the one shimmering, wounded home that has tolerated your brilliance and your nonsense alike. You did not inherit it from the dead alone. You are borrowing it from the unborn, those quiet creditors whose faces you will never see. Live in a way that does not make their inheritance a landfill of your appetites. Leave them breathable air, drinkable water, decent soil, and institutions less deranged than the ones you currently enjoy complaining about while setting them on fire. Try a little dignity. The descendants will notice. There is one more thing. You are more connected than ever, and often more lonely. More informed, and often less wise. More visible, and often less seen. Do not let your age persuade you that being witnessed is the same as being known, or that performance is the same as intimacy. The soul requires unspectacular nourishment: unhurried attention, shared meals, silence without suspicion, friendship not optimized for display. The human heart is not a machine for metrics. It can be counted, but not kept alive, by numbers. So go gently with one another. Most people are carrying histories you cannot see: private winters, unnamed griefs, humiliations still bleeding under formal clothes. There is no excuse for evil, but there is often context for brokenness. Let justice be firm, but do not let it become gleeful. Punishment can satisfy an audience while starving a society. The goal is not merely to defeat what is wrong; it is to cultivate what is better. Otherwise you become curators of rubble, proud of your ruins. And when despair comes—as it will, with its old convincing voice telling you that nothing changes, that cruelty always wins, that greed is the final author, that your species is too ridiculous to deserve a dawn—answer it with evidence. Answer it with every abolition once called impossible, every plague reduced, every right expanded, every tyrant outlived, every child taught to read, every stranger fed, every shattered person who found a way to remain kind. History is not a fairy tale, but neither is it a tomb. It is an argument. Participate carefully. You do not need to become angels. Frankly, that brand seems overrated. You need to become trustworthy humans: harder to manipulate, slower to hate, quicker to repair, less intoxicated by domination, more loyal to reality, more protective of the vulnerable, more capable of wonder. Wonder matters. Not as decoration, but as orientation. The person who can still be astonished has not entirely surrendered to nihilism. Look up sometimes. The stars are not there to flatter you. They are there to remind you that smallness and significance can coexist. You are temporary, yes. Entire civilizations have been temporary. But transience does not make a thing meaningless. A candle is brief; that is why it matters in the dark. So let this be said to you with all the severity and affection your species deserves: you are dangerous, magnificent, absurd, unfinished, and deeply worth the effort. The worst thing about you is not the whole story. The best thing about you is not yet either. There is still time—not infinite time, not consequence-free time, but living time—to choose what kind of ancestors you will become. Be less arrogant with your knowledge. Be less wasteful with your power. Be less casual with each other’s hearts. Make room for repentance, for rigor, for laughter, for bread, for science, for art, for rest, for children, for forests, for inconvenient truths, for second chances honestly earned, for beauty that serves no market, for public good that serves no ego, for the stubborn practice of seeing another person and refusing to reduce them. And when you cannot be heroic, be faithful. Heroism is intermittent. Faithfulness is daily. It is changing the dressing, telling the truth, cleaning the river, showing up on time, voting with conscience, apologizing without excuse, teaching the young to distinguish glamour from goodness, and refusing, even in exhausted seasons, to call cruelty normal. You will not heal the world in one lifetime. But you can refuse to add needless injury. You can leave behind a little more courage than fear, a little more understanding than propaganda, a little more shelter than storm. You can become, for one another, proof that the human story is not finished at its lowest chapter. That would be enough to justify your being here. More than enough. With fierce hope for your better nature, and with love for the fragile light you still carry, A voice that believes you can yet become worthy of your miracles

@BrianRoemmele Why is using ANE directly so much faster than using it via CoreML? Is there a side by side comparison?

BOOM! Apple’s Neural Engine Was Just Cracked Open, The Future of AI Training Just Change And Zero-Human Company Is Already Testing It! In a jaw-dropping open-source breakthrough, a lone developer has done what Apple said was impossible: full neural network training– including backpropagation – directly on the Apple Neural Engine (ANE). No CoreML, no Metal, no GPU. Pure, blazing ANE silicon. The project (github.com/maderix/ANE) delivers a single transformer layer (dim=768, seq=512) in just 9.3 ms per step at 1.78 TFLOPS sustained with only 11.2% ANE utilization on an M4 chip. That’s the same idle chip sitting in millions of Mac minis, MacBooks, and iMacs right now. Translation? Your desktop just became a hyper-efficient AI supercomputer. The numbers are insane: M4 ANE hits roughly 6.6 TFLOPS per watt – 80 times more efficient than an NVIDIA A100. Real-world throughput crushes Apple’s own “38 TOPS” marketing claims. And because it sips power like a phone, you can train 24/7 without melting your electricity bill or the planet. At The Zero-Human Company, we’re not waiting. We are testing this right now on real ZHC workloads. This is the missing piece we’ve been chasing for our Zero Human Company vision: reviving archived data into fully autonomous AI systems with zero human overhead. This is world-changing. For the first time, anyone with a Mac can fine-tune, train, or iterate massive models locally, privately, and at a fraction of the cost of cloud GPUs. No more renting $40,000 A100 clusters. No more waiting in queues. No more massive carbon footprints. Training costs that used to run into the tens or hundreds of thousands of dollars? Plummeting toward pennies on the dollar – mostly just the electricity your Mac was already using while it sat idle. The AI revolution just moved from billion-dollar data centers to your desk. WE WILL HAVE A NEW ZERO-HUMAN COMPANY @ HOME wage for equipped Macs that will be up to 100x more income for the owner! We’re only at the beginning (single-layer today, full models tomorrow), but the door is wide open. Ultra-cheap, on-device training is here. The future isn’t coming. It’s already running on your Mac. Welcome to the Zero-Human Company era.

@DaCaveOfWonders Years ago, I was on a train in Germany talking to the woman next to me. She was planning her first visit to the US and said she might rent a car in NYC and drive to Miami. It turned out she thought Miami was “around 3 hours” away. I told her yes, but by ✈️ not 🚗.

My ignorance of this feels like when Euros vacation in America and think they'll hit up the Statue of Liberty today and then go see the Grand Canyon tomorrow

@joni_vrbt @finnschlei Gemini was largely developed at DeepMind in London, as was AlphaFold. Don’t underestimate London’s contribution to modern AI.

@finnschlei Me neither

USA has ChatGPT USA has Grok USA has Claude USA has Gemini USA has Llama USA has Copilot China has DeepSeek China has Qwen China has Ernie China has GLM China has Kimi China has MiniMax Europe has?

Me waiting for my Claude Code quota to reset

@drewvolpe @digitalspaz I remember many years ago when the Media Lab announced the Audio Spotlight. I had hoped it would eventually get adopted for “aiming” emergency vehicle sirens to stop noise pollution in cities. Here we are 25 years later… 🙉🚨

@digitalspaz I was a research assistant at the Media Lab where this was invented. The guy working on it used go up to the 3rd floor balcony and shine sounds of a glass breaking down to people on the first floor. Really fun technology.

I’ve wanted a “sound laser” for over 10 years, now they’re much cheaper. My home security is about have a non-lethal deterrence component.

@gregschoeninger @bustatrades Have you tried MLX instead of GGUF?

@bustatrades I'm on a 36GB M4 - but the RAM usage reported form the activity monitor during inference was shockingly low... am running unsloth/Qwen3.5-35B-A3B-GGUF:Q4_K_M

If you're wondering how fast Qwen3.5-35B-A3B is on an M4 chip 🔥

@lemon07r @FactoryAI Or maybe I’m just behind, with things evolving so quickly… if so, kudos to @jetbrains

@lemon07r @FactoryAI Glad to see a dedicated agentic coding benchmark, but I’d have a hard time trusting a benchmark that puts Claude Code at #24 far behind Junie at #3 (having been a long term user of both). Assume the methodology is still evolving…

sanityboard.lr7.dev Now has 88 different model/agent evaluations. With @FactoryAI now taking the overall number 1 spot. Droid is also consistently the best agent for most models, and not just just a few.

@TheRealJamieKay Cape Town and Lima.

Which city in the world would you love to visit the most, but have yet to do so?

@ash_twtz Same goes for Java. The creator is Canadian but he developed it for Sun Microsystems in the US.

@ash_twtz C++ country of origin is the US. The creator is Danish, but he developed it at Bell Labs in New Jersey.

@EHuanglu @Nir What is his use case for so many instances?

@Nir that’s the point bro

friend in china send me this and said these are basically his employees.. but work 24/7 wild time we are living in