RickDavis404 @RickDavis404

Helpful tool for improvement. It’s just physics thinking in the limit.

Eric Jorgenson 📚 ☀️ @EricJorgenson

a day ago

Everyone can use @elonmusk's "Magic Wand Number" and "Idiot Index" They're universal ideas, helpful in any industry.

192 605 4K 5.8M 3K

Think of yourself as an LLM. Every social interaction, every meeting, burns your tokens. Unless someone is a paid subscriber to your attention, you are under no obligation to answer low-quality prompts.

@olsenbdnr RE: #2, my favorite thing to do while interviewing was to find something on their resume I didn't have much depth/experience with, but wanted to use, and have a free flowing discussion about it, have them explain pros/cons, how it could be incorporated into our projects

@himanshustwts Don't forget about the network side too...lots of awesome advancements with UEC & copackaged optics & silicon photonics 🤯

SpaceX has released its full IPO roadshow presentation that it's showing to investors. Here are some of the most important slides:

keyboard_arrow_left Previous keyboard_arrow_right Next

While I’m no fan of socialism or arbitrary confiscations of wealth, I can see why Bernie Sanders’ proposal (for the government to take a 50% stake in AI companies) resonates, including with many on the right. The CEOs of the leading AI labs have told us repeatedly that they will cause massive job loss. This is not a story that I believe, nor does the data bear it out, but this is what they have told us. Similarly, they have hyped the risks of AI without putting an equal or greater emphasis on the benefits or readily available mitigations. Conservatives have another fear. The employees of the leading labs claim to be philanthropic, but what we’ve seen is massive enrichment of NGOs advancing an agenda at odds with traditional values, fueling a revolution against our cities and communities. Soros-maxxing is not charity in our book. Anthropic and OpenAI have established themselves as Public Benefit Corporations. What could be more in the public benefit than using half the wealth generated by these companies (which trained for free on the collective knowledge of humanity) to pay down the national debt? There is no ideological bias in that philanthropy. Dario and Sam have begun to walk back their claims of massive job loss, but the damage to public trust is done, and now the chickens are coming home to roost. I could almost support the Sanders proposal as a stupidity tax. There’s just one problem. Nationalization of AI will accelerate the corporate-government fusion we’re already sliding toward. Conservatives rightly fear a Central Bank Digital Currency. They ought to be even more concerned about Central Government AI — a system with even more totalistic power over information, decision-making, and human behavior. We saw how social media was weaponized to censor conservatives (including President Trump) in the last Democrat administration. The definition of “trust & safety” expanded to mean protecting the public from supposed psychological harms, micro-aggressions, and disinformation (you know, like hearing conservative ideas or true facts about Covid). That “safety” agenda as applied to AI will be vastly more powerful and Orwellian. AI won’t just moderate posts; it will curate reality — with the ability to rewrite history, enforce ideological conformity, influence policy at scale, mass surveil Americans, and condition the benefits of the many systems it controls on approved behavior. America won’t win the AI race if we beat China but end up with a CCP-style social credit system in the U.S. — and that is the danger as the government becomes more deeply involved in AI development and assumes direct ownership and control. Conservatives are right to fear where this is all headed but ought to think more carefully about how regulations they are flirting with now (that are widely celebrated among those with a long history of lust for Big Government) will be used against them the next time a Democrat administration is in power.

Bernie Sanders @BernieSanders

6 days ago

I will soon be introducing a bill to give the public a 50% ownership stake in the largest AI companies in America. This would guarantee that the trillions created by AI are used to improve the lives of all of us — and block oligarch decisions that harm the American people.

12K 5K 29K 7.9M 2K

Two years ago my son got a job offer from Anthropic $320k base salary, $200k in stock options at the time While my family was celebrating, I pulled my son aside “Son, you need to decline this offer. Those stock options aren't real, it's fake money American companies print so they don't have to pay real salaries” “But dad, this can be one of the biggest companies in the world, I can be worth millions” he argued with me I burst out laughing “Biggest in the world? It's an overhyped AI company that nobody knows in Europe” I said He went silent “Siemens, Allianz, Lidl. Those are companies you need to work at, if you want to build a real career” I added Two years later, my son works at Siemens in Munich making €58,000 a year, real cash I don't know what happened with that Anthropic business, but in Germany nobody has heard of them yet It will probably be bankrupt soon like most American startups I'm glad I saved my son from the biggest career mistake of his life

Signs you might be trying to get your frontier AI lab nationalized: You compare it to nukes… threaten half of white-collar jobs… warn recursive self-improvement could end humanity… then race ahead anyway. In other words, you want the government to save us from… you.

All the key technical details of Nemotron 3 Ultra you should know - in clean infographics. Just updated the figures with the latest specs. Drop your questions in the thread 👇

keyboard_arrow_left Previous keyboard_arrow_right Next

Locally is now LM Studio’s mobile app. And today we're bringing LM Link to iPhone. Use your largest local models over a secure, end-to-end encrypted connection, anywhere you go. Download the app now: lmstudio.ai/locally

@athulspal That's fine. Open-source will be ready once closed source becomes infeasibly expensive

Nemotron 3 Ultra (550B-A55B) is here - our strongest open-weight model and full training recipe to date. Heavy emphasis on real-world inference efficiency for long-context agentic workloads. Everything is open 🤗: base, post-trained, reward checkpoints, NVFP4 quantized versions, training data, and recipes. Key technical highlights ‼️: - 550B total / 55B active parameters - Hybrid Mamba2-Transformer (~4:1 Mamba:Attention) - Pretrained in NVFP4 on 20T tokens - LatentMoE architecture - Two-stage MOPD post-training - Native MTP Technical details in the thread 👇

Post-Training: Two-Stage MOPD Standard RLHF often struggles to balance broad capabilities with deep specialization. We addressed this with Multi-teacher On-Policy Distillation (MOPD) in two stages. 🔹Stage 1: We first trained a strong general student via RLVR, then used MOPD to distill from multiple high-quality general and agentic teachers. 🔹Stage 2: We further specialized the model by training several expert teachers (coding, SWE, tool-use, search, office workflows, etc.), then performed a second round of MOPD to consolidate them into one model. Key advantages of this approach: 🔸Asynchronous execution - allowed us to scale teacher scoring efficiently. 🔸On-policy rollouts - teachers scored the student’s own generations, reducing distribution shift. 🔸Very long context - we trained with generations up to 192K tokens. The result is a model that retains strong general performance while being significantly better at complex, multi-step agentic tasks.

Training in NVFP4 While post-training quantization to NVFP4 is now standard, training the model from scratch in NVFP4 is much harder. Nemotron 3 Ultra was pretrained entirely in NVFP4 on 20 trillion tokens using: - E2M1 + 2D micro-block quantization (block size 16) - Random Hadamard transforms on wgrad inputs - Stochastic rounding on gradients - Optimized cuBLAS NVFP4 kernels Only the final 16 layers + a few critical projections (QKV, output, MTP, embeddings) were kept in BF16. This is currently the largest-scale stable NVFP4 pretraining we’ve completed.

Mise makes dev life so much simpler. A single env manager for Ruby, JavaScript, Go, and all the modern AI tooling. Every project can have their own versions. Stable system packages can be separated from high-churn AI tooling. Thrilled to sponsor @jdx in this mission!

jdx @jdxcode

5 days ago

mise/en.dev has its first premier-level sponsor! Thanks to @37signals @dhh! en.dev/sponsors.html

13 16 302 117K 45

A DEVELOPER TAUGHT GIT WITH A BOX OF CHILDREN'S TOYS AND ENGINEERS WITH TEN YEARS IN SAY IT'S THE FIRST TIME THE THING EVER ACTUALLY MADE SENSE 90 minutes, one table, a pile of Tinkertoys. No wall of jargon -- he builds a real Git repo out of plastic rods right in front of you. -> The moment he snaps the first pieces together, Git stops being scary command-line magic and becomes what it really is: a chain of tiny objects pointing at each other. Branches, merges, rebase, the staging area -- every concept that's ever burned you at 2am -- he rebuilds with toys until a four year old could follow. He calls Git a two-trick pony. After this you'll see exactly why. Memorizing commands was never the skill -> holding the graph in your head is. And with an AI agent now committing and rebasing on your machine all day, that mental model is the only thing between you and a history you can't read. Scroll the comments and you'll see the same thing over and over: this is the talk that finally made Git click and made people the one their whole team comes to when it breaks. Bookmark & watch it today. It's the 1.5 hours that pays you back for the rest of your career ↓

slash1s @slash1sol

a week ago

MIT HANDED ITS DEEP LEARNING COURSE TO A FRONTIER-LAB ENGINEER FOR 68 MINUTES BECAUSE 90% OF PEOPLE SHIPPING AI CODE CAN'T EXPLAIN HOW THE MODEL ACTUALLY WORKS This is Maxime Labonne. He runs post-training at Liquid AI and wrote the LLM Engineer's Handbook. MIT gave him the room

35 71 691 232K 1K

I bet the other 24 letters are pissed

WOLF @WOLF_Financial

6 days ago

Alphabet is raising $80 Billion by selling stock to invest in AI

55 22 263 1.6M 36

Cerebras did what the industry calls impossible: turned an entire 46,225mm² wafer into one chip. Defects on silicon that big are inevitable, so they built in redundancy and custom per-batch masks that route around every bad core, landing near 100% usable wafers. The results: 900,000 cores and 44GB of SRAM on a single piece of silicon, no packaging, no off-chip hops. And they're not stopping there, now exploring hybrid bonding a DRAM wafer on top for even more fast memory. (1/4) 🧵

someone made a fork of opencode that routes through the unsecured ai endpoints from chipotle