if it were up to me, there would be a LLAMA 5 it would be dense. it would be multimodal. it would come in four sizes: 1B, 8B, 80B, 800B. it would be iterated upon until the 8B LLAMA 5 totally outperformed the 80B LLAMA 3. this would be a huge asset to the community.
RESEARCH
-
Dyson Spheres: Beyond Reasoning Toward Advanced Technology
By
–
We're not going to reason our way to Dyson spheres. Excited for you and all the best to the team!
-
xAI Launches Grokipedia Open Source Knowledge Repository
By
–
Join @xAI and help build Grokipedia, an open source knowledge repository that is vastly better than Wikipedia! This will be available to the public with no limits on use.
-
Sora 2 vs Sora 1: One Year of Dramatic AI Video Generation Improvement
By
–
what an insane difference a year makes
— Matt Shumer (@mattshumer_) 30 septembre 2025
sora 1 (left) vs. sora 2 (right)
both using the same prompt pic.twitter.com/zUpvZX1Z94what an insane difference a year makes sora 1 (left) vs. sora 2 (right) both using the same prompt
-
Terminal-Bench 2.0 Advances SOTA for Next-Gen AI Agents
By
–
We’re excited to contribute to Terminal-Bench 2.0, and help advance SOTA for evaluating next-gen AI agents. S/O: @Mike_A_Merrill @alexgshaw @terminal_bench @StanfordAILab @LaudeInstitute Read more:
-
Snorkel Enhances AI Benchmark with Task Calibration and Analysis
By
–
What makes Snorkel's contribution unique? We're bringing our expertise to help:
• Contribute tasks
• Calibrate task difficulty
• Provide sophisticated performance analysis
• Expand the benchmark's evaluation techniques -
Terminal-Bench Gains Recognition Across Major AI Models
By
–
The industry is taking notice Terminal-Bench scores now appear on: DeepSeek-V3.1-Terminus model card Qwen3-Coder release Claude Sonnet 4.5 announcement
800+ GitHub stars | ~100 developer contributors -
Terminal-Bench: Evaluating AI Coding Assistants Performance
By
–
Why does Terminal-Bench matter?
AI coding assistants (Claude Code, Codex, Cursor, & Devin) rely on command-line interfaces. The terminal represents a convergence of power, flexibility, & the text-based modality where language models excel—hence the need for robust evaluation. -

How Artists Use AI Tools in Music Creation and Production
By
–
How are artists using AI to make music? 🎶 That’s what our Audio Research team set out to understand when they analyzed 337 musical works. The research examines how artists are using AI tools in their creative and production process. We’re sharing the key findings, with insight on emerging AI music creation practices that artists and creative professionals can draw inspiration from. You can read the full paper and learn more here 👉 bit.ly/3INc9I0
→ View original post on X — @stabilityai, 2025-09-30 16:48 UTC
-

AI for Science: Periodic Labs Launches Autonomous Scientific Discovery Platform
By
–
Bullish, in the coming decades majority of compute will be spent on ai for science William Fedus (@LiamFedus) Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is created when ideas are found to be consistent with reality. And so, at Periodic, we are building AI scientists and the autonomous laboratories for them to operate. Until now, scientific AI advances have come from models trained on the internet. But despite its vastness — it’s still finite (estimates are ~10T text tokens where one English word may be 1-2 tokens). And in recent years the best frontier AI models have fully exhausted it. Researchers seek better use of this data, but as any scientist knows: though re-reading a textbook may give new insights, they eventually need to try their idea to see if it holds. Autonomous labs are central to our strategy. They provide huge amounts of high-quality data (each experiment can produce GBs of data!) that exists nowhere else. They generate valuable negative results which are seldom published. But most importantly, they give our AI scientists the tools to act. We’re starting in the physical sciences. Technological progress is limited by our ability to design the physical world. We’re starting here because experiments have high signal-to-noise and are (relatively) fast, physical simulations effectively model many systems, but more broadly, physics is a verifiable environment. AI has progressed fastest in domains with data and verifiable results – for example, in math and code. Here, nature is the RL environment. One of our goals is to discover superconductors that work at higher temperatures than today's materials. Significant advances could help us create next-generation transportation and build power grids with minimal losses. But this is just one example — if we can automate materials design, we have the potential to accelerate Moore’s Law, space travel, and nuclear fusion. We’re also working to deploy our solutions with industry. As an example, we're helping a semiconductor manufacturer that is facing issues with heat dissipation on their chips. We’re training custom agents for their engineers and researchers to make sense of their experimental data in order to iterate faster. Our founding team co-created ChatGPT, DeepMind’s GNoME, OpenAI’s Operator (now Agent), the neural attention mechanism, MatterGen; have scaled autonomous physics labs; and have contributed to some of the most important materials discoveries of the last decade. We’ve come together to scale up and reimagine how science is done. We’re fortunate to be backed by investors who share our vision, including @a16z who led our $300M round, as well as @Felicis, DST Global, NVentures (NVIDIA’s venture capital arm), @Accel and individuals including @JeffBezos , @eladgil , @ericschmidt, and @JeffDean. Their support will help us grow our team, scale our labs, and develop the first generation of AI scientists. — https://nitter.net/LiamFedus/status/1973055380193431965#m
→ View original post on X — @_jasonwei, 2025-09-30 16:04 UTC