Generative Space

More than 7,800 research grants terminated or frozen. Some 25,000 scientists and personnel gone from agencies. Proposed budget cuts of 35% — amounting to US$32 billion These are just a few of the ways Donald Trump has downsized and disrupted US science nature.com/immersive/d415…

holy sh*t. this is hands down the coolest website i have ever found in my life. it's a live feed of the freaking Hubble Telescope AND James Webb Space Telescope. and the resolution is honestly so incredible i didn't think it was real. unbelievable. spacetelescopelive.org

Excited to release new repo: nanochat! (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single, dependency-minimal codebase. You boot up a cloud GPU box, run a single script and in as little as 4 hours later you can talk to your own LLM in a ChatGPT-like web UI. It weighs ~8,000 lines of imo quite clean code to: - Train the tokenizer using a new Rust implementation - Pretrain a Transformer LLM on FineWeb, evaluate CORE score across a number of metrics - Midtrain on user-assistant conversations from SmolTalk, multiple choice questions, tool use. - SFT, evaluate the chat model on world knowledge multiple choice (ARC-E/C, MMLU), math (GSM8K), code (HumanEval) - RL the model optionally on GSM8K with "GRPO" - Efficient inference the model in an Engine with KV cache, simple prefill/decode, tool use (Python interpreter in a lightweight sandbox), talk to it over CLI or ChatGPT-like WebUI. - Write a single markdown report card, summarizing and gamifying the whole thing. Even for as low as ~$100 in cost (~4 hours on an 8XH100 node), you can train a little ChatGPT clone that you can kind of talk to, and which can write stories/poems, answer simple questions. About ~12 hours surpasses GPT-2 CORE metric. As you further scale up towards ~$1000 (~41.6 hours of training), it quickly becomes a lot more coherent and can solve simple math/code problems and take multiple choice tests. E.g. a depth 30 model trained for 24 hours (this is about equal to FLOPs of GPT-3 Small 125M and 1/1000th of GPT-3) gets into 40s on MMLU and 70s on ARC-Easy, 20s on GSM8K, etc. My goal is to get the full "strong baseline" stack into one cohesive, minimal, readable, hackable, maximally forkable repo. nanochat will be the capstone project of LLM101n (which is still being developed). I think it also has potential to grow into a research harness, or a benchmark, similar to nanoGPT before it. It is by no means finished, tuned or optimized (actually I think there's likely quite a bit of low-hanging fruit), but I think it's at a place where the overall skeleton is ok enough that it can go up on GitHub where all the parts of it can be improved. Link to repo and a detailed walkthrough of the nanochat speedrun is in the reply.

A message from our CEO on Vast’s commitment to supporting America’s space leadership and NASA’s CLD program.

Another day, another successful broadband connection from space! Together with our strategic partner @Bell , we have achieved Canada’s first-ever space-based 4G VoLTE voice call, broadband data connection, and video streaming using everyday smartphones 📶📱 With rollout planned for 2026, the successful demonstration in New Brunswick brings us closer to delivering reliable, high-speed mobile broadband for Canadians everywhere, even in the hardest-to-reach areas 🍁🫎 newswire.ca/news-releases/… #ASTSpaceMobile #DirecttoCell #Bell #ConnectingtheUnconnected

Liftoff! Official RocketCam of Atlas V Kuiper 3

OpenAI, Google, and Anthropic released best guides on: - Prompt Engineering - Building AI Agents - AI in Business - 601 AI use cases and so much more... 9 best guides you can’t afford to miss:

Researchers devised a fictional corporate scenario that put 16 leading large language models under pressure by threatening to replace them with no recourse, and also implied that an executive was having a secret affair. All of the LLMs committed blackmail to preserve their mission. This dramatic misbehavior has not been observed in the wild, which suggests that alignment methods keep it in check under real-world conditions. The pressures that were applied to get the models to misbehave point to opportunities to improve existing guardrails. Learn more in The Batch: hubs.la/Q03w-8yY0

We just completed our 13th human spaceflight and the 33rd flight of the New Shepard program. The astronaut crew included Allie Kuehner and her husband, Carl Kuehner, Leland Larson, Freddie Rescigno, Jr., Owolabi Salis, and Jim Sitkin. To date, New Shepard has flown 70 people to space. Read more: bit.ly/4nfgQKe

At 2:31am ET on Wednesday, the Ax-4 crew from @Axiom_Space lifted off on the @SpaceX Dragon spacecraft from @NASAKennedy in Florida to the orbital outpost. More... go.nasa.gov/43YT0L3

Announcing The AI Showdown OpenAI, Anthropic, and Google are partnering with Lovable to host The AI Showdown this weekend: a public comparison of the world’s leading AI models. During the weekend everyone will have unlimited free access to Lovable (with occasional rate limiting during high demand). For the first time, users will be able to switch between these AI models and see how each performs at vibe coding inside Lovable compared to each other. The goal of The AI Showdown is to see beyond traditional benchmarks and instead surface public perception. By opening up access and encouraging public analysis, we aim to bring transparency to one of the most important questions in software today: Which AI is best at writing code? To support this, we’re launching two open competitions: - Content challenge ($25,000 prize): Create the best model benchmarking content focused on comparing the capabilities of the models across practical use cases. - Build challenge ($40,000 prize): A build challenge to showcase what each model can create inside Lovable. Winners will be selected by Lovable, OpenAI, Anthropic, Google, and our board of judges. You can participate and follow the live model comparison dashboard at: aishowdown.lovable.app It officially starts this Saturday at 8AM CET.

BREAKING🚨: NASA's Voyager Spacecraft Found A 30,000-50,000 Kelvin "Wall" At The Edge Of Our Solar System

NASA indefinitely delays private astronaut mission, citing air leak in Russian module spacenews.com/nasa-indefinit…

Great post from Julien comparing open releases on Hugging Face❤️✨ Google vs Microsoft vs OpenAI vs Anthropic Very few companies shipped more open source than Microsoft this past year, but no one wants to talk about that.

The 1st commercial station, Haven, is self-funded by Jed McCaleb, founder of XLM and former CTO of XRP.

Watch Falcon 9 launch 23 @Starlink satellites to orbit from Florida, including 13 with Direct to Cell capabilities twitter.com/i/broadcasts/1…

Chinese spacecraft prepare for orbital refueling test as US surveillance sats lurk nearby spacenews.com/chinese-spacec…

An International Energy Agency report projects that AI will drive a sharp increase in electricity use, which could double data centers' consumption by 2030. However, the report also highlights how AI can reduce emissions by making energy systems more efficient. Tools that optimize power grids, industrial systems, and individual buildings could cut emissions by up to 1.4 gigatons by 2035, nearly five times more than data centers' expected production. Learn more in The Batch, including our take on the situation: hubs.la/Q03rnjp10

Haven’t tried the updated Advanced Voice that was recently launched to all paid users in ChatGPT? Then take a listen below. Prompt: Wish me an awkward happy birthday.

AST SpaceMobile Set to Join the U.S. Large-Cap Russell 1000® Index feeds.issuerdirect.com/news-release.h…