John Helton @moof52

The Macintosh computer assembly line at the Apple factory in Cupertino, CA, 1984

If you can afford to buy the Mac Studio M5 Max 128GB and the DGX Spark 128GB together, you will be safe for the next three years. All the models and software will improve massively around December this year and Q1 next year, and this will also solve the subscription problem for all these platforms and censorship. The price will go up massively, and Apple's waiting list will get even longer.

You’ve mostly got it right, with some caveats. Most memory tools keep notes outside the model. RAG, vector databases and memory files store text, search it when you ask something, and paste what they find into the prompt. The model itself never changes. It rereads its notes every time, and if the search misses, it doesn’t know. Living Weights changes the model. When you tell it “I like blue”, it writes its own questions about the fact and trains a small change into its weights until it can answer them. Then it saves that change into the model’s files. It adds nothing to the prompt and looks nothing up. Restart it, and it still knows. Editing facts into weights isn’t new in research. ROME and MEMIT are the best-known papers. What Living Weights adds is doing it live on your own machine, with checks before anything is kept. Training facts into a model usually bleeds into other answers, and that’s where most of the work goes. Before it keeps a lesson, it checks that ordinary answers haven’t changed and that the fact hasn’t leaked onto similar ones. In one test it refused “my dog is called Ginger” because it had started answering “what’s my cat’s name?” with Ginger. Today it handles short personal facts, one at a time. Each one takes 2 to 4 minutes on an M3 Ultra and about 22 GB of memory while it learns. With thinking on, the model can still reason its way back to “I don’t know you”, and we’re fixing that now. A saved fact is hard to remove. It also doesn’t learn from conversations by itself yet. You tell it what to learn with the /learn command. So the coworker comparison works for things it should always know about you. Notes still win for documents and anything that changes often at the moment, but I am solving that. Jarvis on a Spark is where this is heading. A Spark can already serve a model that learned on a Mac, and I’m building learning on the Spark itself as we speak.

It’s absolute insanity. We produce gas here in California, ship it to Arizona, and they sell it for $2 LESS per gallon than we pay here. That extra $2? Democrat policies making gas more expensive here than any other state! Enough is enough, California!!

NY AG debate last night was a romp. Tish James saying she doesn’t need “facts or law” when making prosecution decisions ought to chill all of us to the bone.

Another day, another Econ 101 lesson for Professor Warren. Her “Trumpflation” tracker conveniently ignores the Bidenflation tidal wave that crushed American families, cherry-picks wage growth by refusing to acknowledge positive real wage gains since President Trump came into office, and swaps official medical-care CPI data for premium projections when it suits her narrative. The Harvard Square inflationista is offering a statistical sleight of hand, not a serious assessment of today’s U.S. economy. Here is the full picture, including an apples-to-apples comparison of President Trump’s real economic results measured against the failed policies Senator Warren championed under Biden.

TensorFold 1.0.3 lands with a few cool things. - Living Weights (beta with Nemotron on Metal) and coming to CUDA in 1.0.4 - Nvidia Ampere (beta) for Nemotron, and native Qwen3.8-27B with DFlash2 - Conversation prefix caching - Various bugs fixed and improvements Twelve contributor PRs helped make this happen, so a massive thank you to @niklaslenz_ai and GitHub contributors akol1, jschmied, chaog992, AdrianBinDC and BobClawblaw, and everyone testing, reporting bugs and helping us improve TensorFold. FYI, Living Weights is in very early beta and has sharp edges, but I think I’m on to something. I’d love community feedback to make it better: github.com/ashhart/Tensor… @plotarmordev @petruspennanen @MiaAI_lab @volatilemarkts tensorfold.dev

My condolences to the Ditka family and to all Bear fans on the passing of Coach Mike Ditka. It was truly an honor to play against you and your teams. Now rest in peace and God Bless you Coach

It is wild seeing an LLM learn new things, and it retains them in its weights.

Chad Hurley is Volatile Markets @volatilemarkts

2 days ago

We've been building on open weights long enough to remember what “open” always meant: “frozen”! Download it, run it -it never learns a single thing from you. Today that ended. @ashxhart is shipping Living Weights: an open model that learns while it runs. He could've walked

6 3 41 5K 11

If you’re ever at the airport in Chicago, remember that it is named after the Navy’s 1st flying ace in World War II, Lt. Commander Butch O'Hare. - Medal of Honor 🇺🇸 - Navy Cross ⚓️ - 2 Distinguished Flying Cross✈️ - Purple Heart 💜

oMLX 0.7.1.dev1 is out! This dev release adds decision models (Clef, OpenJev), batched prefill for concurrent requests, a refreshed web UI, and faster Qwen3.8, Qwen3.6, GLM-5.3-Flash and DeepSeek V4.1. github.com/jundot/omlx/re… Speedups in this release - Qwen3.6-35B-A3B decode on M5 Max: 156 -> 205 tok/s (+31%). - GLM-5.3-Flash decode on M3 Ultra: 31.3 -> 40.6 tok/s (+30%). The fused decode kernels now run on M3 and M4, not only M5. - Qwen3.8-Flash-Next with expert offload, 32K prefill on M3 Ultra: 396 -> 636 tok/s (+61%). - Qwen3.8-Flash-Next, 8 prompts arriving 0.4 s apart on M5 Max: time to first token 7.41 -> 5.75 s (-22%) with batched prefill. What's new - Decision models. Clef, Clef-Flash and OpenJev answer typed questions with a probability for each option through the new /v1/systemone endpoint. - Batched prefill. Prompts that prefill at the same time now share one forward pass, so MoE models read their expert weights once. - A refreshed web UI with a log table, a settings rail and a customizable dashboard. - Native video input for Qwen3.5, 3.6 and 3.8. - More accurate oQe quantization, with 11-15% lower KL to the original model. - Fewer memory guard rejections in long sessions. Most of the speedups came from contributors again. Thanks to everyone who sent PRs and tested the dev builds!

Every open model ships frozen. What if it could learn and update its own weights as it's being used? I have been working on a bit of engineering called Living Weights for about three months, and I think it's time to release it. The model becomes uniquely yours, no KV stashing, no MD files, no trickery.... just live learning. 😎 Unload the model, move it to another machine, and it still knows what you told it. Could this be the start of recursive self improvement for local AI? It drops in the latest TensorFold release today, with support for @NVIDIAAI Nemotron Lightning. 🤯👀 @abheer @sundeep @NaderLikeLadder Special thanks to @volatilemarkts for the name.

That would be fantastic. It would really lessen the dependency on the web based models. Even lessen the need for those big data centers. Move it all, or almost all, to local ai. 😁

Ash Hart @ashxhart

2 days ago

Every open model ships frozen. What if it could learn and update its own weights as it's being used? I have been working on a bit of engineering called Living Weights for about three months, and I think it's time to release it. The model becomes uniquely yours, no KV

198 141 1K 81K 1K

Very nice “turbo” update to arguably the best open weight image model just dropped! Qwen/Qwen-Image-2.1-Turbo huggingface.co/Qwen/Qwen-Imag…

Qwen @Alibaba_Qwen

2 days ago

🚀 Meet Qwen-Image-2.1-Turbo — create and edit images in just 8 denoising steps! Open weights now available! Built on Qwen-Image-2.1, Turbo is an accelerated checkpoint on the same 7B visual generation architecture. Fewer steps does not mean lower quality: it still generates

198 514 5K 640K 4K

Upset about gas prices??

As I said many times before, TensorFold is a real turning point for local AI. Just try it for yourself, we have a recipe for everyone! All of my TensorFold recipes can be found here: mia-ai.net/models?for=eng…

Abheer Singh @abheer

2 days ago

TensorFold by @ashxhart is one of our newest Inception startups and has been making waves in the local AI community. It's an open-source inference engine for running models on your own hardware. Finally got around to experiencing it for myself. I'm running Qwen3.8 Flash Next in

11 2 48 19K 26

Apple’s OLED MacBook Pro won’t be significantly thinner. It is expected to launch later this month. Source: Mark Gurman

NEW: Apple’s big October will continue on the 27th with a second event to launch the first touch-screen MacBooks, OLED iPad mini and M6 14-inch MacBook Pro & iMac. It follows the smart home launch this coming Tuesday the 13th. bloomberg.com/news/articles/…

Somali scumbag Fahima Mahamud pleads GUILTY to $3million daycare fraud in Minneapolis. Naturally, she tried to flee to London before being arrested by the FBI. Nick Shirley deserves the Pulitzer Prize for journalism. 📽️ = @EndWokeness @nickshirleyy

Actor David Spade raising awareness that a bill was introduced in California to audit where the $20+ billion dollars in missing homeless money went, it passed but then Gavin Newsom VETOED the bill to block the investigation Yes, this really happened. He blocked bills for an audit MULTIPLE TIMES Bipartisan bill AB 2903 (unanimous passed 72-0 in the Assembly, 40-0 in the Senate) would’ve forced annual public reports on where the money went Newsom vetoed it. Gavin Newsom also vetoed similar bills AB 2570 and AB 2093 “The same broader problem with people paying taxes in California — The homeless, they lose $20 billion, but they want more money for it. That's why people get tired of paying taxes and going, what are you doing? Gavin Newsom just vetoed a bill asking for an audit of where the money for homeless goes. He said, not a chance. You're not gonna see that. That's the problem.”