XFoundation models05:53Emad@EMostaqueI released this today because I was 100% sure a frontier AI would figure it out Matter is chiral => the first derivation of the standard model & three generations uniquely That’s all, nothing else Can’t even imagine what big labs are sitting on (quoting thread arguing OpenAI and Anthropic have many unreleased results proved by internal models)Collected by WispRead in full↗
XFoundation models05:00Susan Zhang@suchenzanghmmm, that's a funny typo... shoutout to nvlink, the key to scaling test-time-compute! (quoting Jensen Huang: GPT-6 Astra trained on ~100K+ NVIDIA Grace Blackwell NVLink72; from ChatGPT to o1 to Astra in 4 years; AGI has arrived; 400K GPUs coming online next)Collected by WispRead in full↗
XFoundation models03:38Alexandr Wang@alexandr_wangnice comparison of muse spark 1.3 max and gpt-6 astraCollected by WispRead in full↗
XFoundation models03:37Alexandr Wang@alexandr_wangMuse Spark 1.3 Max by Vals AI, competitive with Claude Fable 5 and GPT-5.6 Sol while 4-8x cheaperCollected by WispRead in full↗
XFoundation models03:15Robert Scoble@ScobleizerWhat I see here? AI is now building AI. And everything is about to speed up because of it. This is the hot Sunday essay going around AI industry right now.Collected by WispRead in full↗
XFoundation models01:10Eric Jang@ericjang11The section on "Monitoring generalization" is particularly interesting. I wonder what the equivalent of COT monitoring looks like as capabilities become increasingly multimodal, e.g. generation and robotics, where the predicted quantity is something that might not have a natural function approximation in the space of words. (quoted) I wrote about the state of AI, why I’m concerned about the next few years, and the choices we need to make to keep the future in humanity’s hands. An Alien Mind: https://openai.com/index/an-alien-mind/Collected by WispRead in full↗
XFoundation models01:02will depue@willdepuewe talk a lot about US and China collaboration on AI, but we often forget about the rest of the world. there’s no guarantee that Israel, or France, or India, or Russia falls in line either. we’re not in the same world we were in when we once negotiated nuclear nonproliferation!Collected by WispRead in full↗
XFoundation models00:53will depue@willdepuewe’re 6 months away from my personal definition of AGI: a machine that can do anything an above average human could do on a computer this encapsulates everything from self-driving (poorly driving a car via arrow keys on video call) to doing alright on year-long horizon projects (quoted) I asked GPT-6 Astra to mine a diamond in Minecraft using computer use, then went to sleep. Woke up to a diamond in its inventory 🤯🤯🤯 Setup in the replies.Collected by WispRead in full↗
XDeveloper tools00:44Ethan Mollick@emollickHere is an impressive one-shot result from Fable 5.1: "build out at least 8 games based on Edgar Allen Poe. they should all be great and very distinctive" The games are surprisingly different, fun to play and evocative, and some are actually a bit creepy https://poe-arcade.netlify.app/index.htmlCollected by WispRead in full↗
XFoundation models00:36Noam Brown@polynoamialOne of the most interesting blog posts we've released: details on internal research acceleration at @OpenAI. I expect these trends to continue. We also share some details on how we've paced model development to prioritize monitoring, alignment, and security. (quoted) Today we're releasing data on models accelerating research at OpenAI. Recursive self-improvement could be the most important contributor to AI capabilities over the next few years, but by default it will only be seen inside a few frontier AI labs. Being transparent is more urgent than ever, so we can inform the public discussion on whether and how to pace model development. I ask other AI companies to do the same. https://openai.com/index/research-acceleration-view-inside-openai/Collected by WispRead in full↗
XDeveloper tools00:17Satya Nadella@satyanadellaIt was a great week for innovation across the model ecosystem. We're bringing these new models into Copilot, enabling it to take on increasingly complex work, from quick questions to delegated tasks and, increasingly, complete long-running jobs through Autopilots. Here's one fun example: using an Opal-powered Autopilot running on a secure Windows 365 Cloud PC, Copilot goes through a month of trail cam footage, finds every animal sighting, creates a highlight reel of the best clips, labels each one with the camera, date, and species, catalogs every sighting in a spreadsheet, builds a PowerPoint summarizing the findings, and shares everything in Teams for colleagues to review. This is the next frontier for Copilot: software that doesn't just assist with work, but can own and complete entire tasks that unfold over hours or days.Collected by WispRead in full↗
XFoundation models00:13Mustafa Suleyman@mustafasuleymanThe rate of proliferation in AI is more extreme than most people realize. Inference costs for GPT-4 class intelligence have come down 300x in 3yrs. Hard to think of any other technology in history that has fallen that fast.Collected by WispRead in full↗
XDeveloper tools00:01Emad@EMostaqueIn 2023 I said there would be no more developers by 2028 AI is moving across the threshold now Next year it will write better code than all but a few of you (I have some awesome followers) 2027 it will instantly generate almost all code 2028 folk will run out of ideas (quoted) Sorry developers, but GPT Astra is coming for your jobsCollected by WispRead in full↗
XDeveloper tools23:40Aaron Levie@levieIf agents produce the vast majority of software in the future, and they’re most trained on open source software, they will inevitably do their best work with those tools. If you cycle this enough times, it means that open source effectively becomes the dominant software in the future as it’s what everything important gets built with. This was already the trend in many critical domains, but agents will accelerate this far faster than humans ever could have. (quoted) The unexpected benefit of open source is that it allows model training companies to train on using your software (and optimizing it) for free. Blender may have just won as a 3D asset creation software because the models will be better at using it than any proprietary onesCollected by WispRead in full↗
XDeveloper tools23:00Alexandr Wang@alexandr_wanggive muse spark 1.3 max a shot! (quoted) @alexandr_wang @davis7 muse spark 1.3 max is mindblowing, opus level model!Collected by WispRead in full↗
XFoundation models22:43Robert Scoble@ScobleizerI came across a team that's been sitting with this question longer than most. The founder of Today AI has spent over a decade building products around one consistent obsession: how should technology help people spend their time on what actually matters? First in the era of mobile collaboration. Now in the era of AI agents. The question didn't change. The tools finally caught up. Their approach: an assistant that builds a real picture of you over time — not just your tasks, but your context, your priorities, your patterns. Morning and evening briefs built around you, not your calendar. Relevant developments surfaced before you ask. The morning brief is the thing I'd watch closely. Does it just reorganize what you already know? Or does it bring you something you didn't know you needed to see? That's the difference between a tool that remembers and one that understands. The bigger question...for everyone building in this space: how will we know when an AI has finally learned what matters to us? I'm not sure we have a good answer yet. But I think we're finally asking the right question.Collected by WispRead in full↗
XFoundation models22:43Robert Scoble@ScobleizerDoes your AI know what matters to you? Not what you told it. Not what's on your calendar. Not what you asked it to do last Tuesday. What actually matters to you. @ashwingop just published something worth sitting with. His argument: as models and execution get commoditized, memory becomes the moat. Not storing more — knowing what's worth storing. Knowing what's worth interrupting you for. I think there's a harder question underneath that. The calendar knows when. You know why. That gap...between what happened and why it matters... is where the next decade of AI gets won or lost.Collected by WispRead in full↗
XFoundation models22:24Ethan Mollick@emollickAI rewards expertise (at least for now). Expertise lets you judge AI output quality and find the shape of the jagged frontier quickly. It also gives you more options for how to try to improve quality by knowing what changes to ask for. Non-experts are often stuck with defaults.Collected by WispRead in full↗
XDeveloper tools21:48Sebastian Raschka@rasbtReasoning from scratch round 2: In this video, I cover the text generation process in LLMs and KV caching (to prepare the base model before adding reasoning techniques in the upcoming ones). 00:00 Introduction and reasoning model demo 01:55 How to work through the book 05:00 Chapter 2 overview 08:25 Checking PyTorch and hardware support 10:26 Apple silicon and MPS caveats 15:00 Cloud GPU options 16:08 Tokens and tokenization 18:20 Qwen3 and the Reasoning From Scratch package 23:05 Encoding and decoding text 26:24 Downloading weights and selecting a device 31:01 Loading the pretrained Qwen3 model 34:32 How LLMs generate text 36:47 Input tensors and batch dimensions 41:48 Running the model in inference mode 44:11 Logits and next-token predictions 49:21 Greedy decoding with argmax 52:28 Building a streaming text generator 01:01:28 Generating text and handling end-of-sequence tokens 01:06:00 Benchmarking text generation 01:14:34 How KV caching works 01:17:22 Adding KV caching and measuring the speedup 01:24:31 Model compilation with torch.compile 01:30:33 Combining compilation with KV caching 01:32:53 Comparing CPU and GPU performance 01:35:32 Recap and next stepsCollected by WispRead in full↗
XFoundation models04:13OpenAI@OpenAIGPT-6 Astra is now available to all Pro, Enterprise, and Business Premium users in ChatGPT Work and Codex. It's also live in the API. It might take a few days to roll out to our Plus and Business users. Thank you for your patience.Collected by WispRead in full↗