We collect all the most important AI updates from February.
xAI launches Grok 3

xAI launched Grok3 this week, it is an order of magnitude more capable than Grok 2, with 10x more computing power thanks to xAI's Colossus 100k H100s.
Grok 3 excels in math, science, coding, and general knowledge, with notable performance in image understanding tasks, achieving a 73.2% score on the MMMU benchmark (xAI Blog).
It's seems to be o1 level but it also introduced DeepResearch (they called it DeepSearch) and Search capabilities. The subscription costs $22/month. The most interesting feature of Grok is still the direct access to the twitter feed.
So while G3 is not full o3 level, which didn't launch yet, it signal xAI entering the arena in a competitive way.
https://x.com/adonis_singh/status/1892109817830851060
Microsoft Launches Quantum Chip Majorana 1
https://x.com/MSFTQuantum/status/1892252249797124598
Microsoft’s Majorana 1, unveiled today, is the world’s first quantum processor using topological qubits. Here are three key insights:
- Stable Qubits: It harnesses Majorana zero modes for more stable, error-resistant qubits, advancing quantum computing scalability.
- Extreme Conditions: The processor relies on near-absolute-zero temperatures and magnetic fields to create a rare topological state of matter.
- Scalable Future: Microsoft aims to scale to a million qubits on one chip, building on its 2024 Quantinuum collaboration for reliable quantum systems.
Hyperscaler Capex
https://x.com/kimmonismus/status/1889784565935317169
The hardware AI landscape is accelerating at a breathtaking pace. Tech giants like Microsoft, Google, Meta, and Amazon spending $528 billion on AI infrastructure in 2024 to pursue artificial general intelligence (AGI). Leaders like Dario Amodei and Sam Altman predict AGI could arrive within 1-2 years, promising breakthroughs in health and energy but also warning of job losses, inequality, and societal risks. Despite challenges like electricity and chip shortages, the post calls for urgent global discussions to ensure equitable outcomes as humanity approaches this transformative era.
Scaling up test-time compute with latent reasoning
https://x.com/MatthewBerman/status/1890081482104008920
Full Sources List
Random
- Satya Nadella expect a slow take off, the real problem will be in court, https://x.com/dwarkesh_sp/status/1892605361624146034
- beautiful video from Prime Intellect https://x.com/PrimeIntellect/status/1892282553718992925
- Ethan Mollick, most common words per model https://x.com/emollick/status/1892047167713427687
- Mira Murati launches Thinking Machines https://x.com/thinkymachines/status/1891919141151572094
- Are we living in a super conductor? Yes https://x.com/RBehiel/status/1890886770633621993
- Water consumed by ChatGPT < cows https://x.com/AlecStapp/status/1885192778256179605
- Meta FAIR brain control https://x.com/AIatMeta/status/1890464494972964971
- Small team are the future https://x.com/benln/status/1890413609768956015
- A16z, outsource business is getting disrupted by AI https://x.com/kimberlywtan/status/1890103051354316925
- Only 7 coders in America better than o3 https://x.com/itsandrewgao/status/1889792665086251398
- Predictive LLM that explore 5 moves ahead https://x.com/cadenbuild/status/1889909377886462320
Robotics
- Figure introduces robotics AI that thinks like a human https://x.com/adcock_brett/status/1892577936869327233
- Unitree dance, https://x.com/siyuanhuang95/status/1891760580408573986
GPUs
- ⭐️ Azure introduces Majorana 1 quantum processor https://x.com/MSFTQuantum/status/1892252249797124598
- NVIDIA for training, Groq for inference https://x.com/HarryStebbings/status/1891506929001804212
Memes and Funsies
- small couches https://x.com/inductionheads/status/1891959833030525412
- robot gets angry https://x.com/kimmonismus/status/1892594559424848336
- AI safety in the trenches https://x.com/LouisAnslow/status/1891259852065796507
- Vibe coding https://x.com/omnius_eacc/status/1889708679043158478
Beautiful Visualizations
- Elo score top 9 company history https://x.com/lmarena_ai/status/1892250937907110313
- Beautiful 3d design https://x.com/brandon_xyzw/status/1890974612072095932
- Origami showing NN layers gradient descent https://x.com/imtiazprio/status/1889570723405946896
- Fourier transform visualized https://x.com/BrianRoemmele/status/1889917509719171328
- ⭐️ hyperscaler capex by year https://x.com/kimmonismus/status/1889784565935317169
Research Papers
- Large language diffusion models, a diffusion-based approach that mathes or beat leading autoregressive LLMs in many tasks https://x.com/omarsar0/status/1891568386494300252
- Can LLM encode future outputs? https://x.com/omarsar0/status/1891554210757607686
- ⭐️ Scaling up test-time compute with latent reasoning https://x.com/MatthewBerman/status/1890081482104008920
Tools
- Google introduces AI co-scientist, https://x.com/sundarpichai/status/1892254274895184244
LLMs
- ⭐️ Grok 3 launches https://x.com/adonis_singh/status/1892109817830851060
- GPT 4.5 coming up in the next weeks https://x.com/kimmonismus/status/1889757391341928623
- ⭐️ Ultra scale playbook, a comprehensive guide on training LLMs from 1 to 1000 GPUs https://x.com/Nouamanetazi/status/1892274582503248178
- Perplexity AI releases r1-1776, uncensored DeepSeek r1 model https://x.com/AravSrinivas/status/1891917148869755058, https://x.com/hardmaru/status/1892051306388521474
- Meta FAIR, audiobox aestheitc https://x.com/AIatMeta/status/1889418249466683449
Benchmarks
- SWE-lancer, a new benchmark that tests how much money a model would theoretically make on a set of real-world freelance software engineering task — Hint: Claude is the winner https://x.com/stevenheidel/status/1891988220214120626, https://x.com/miramurati/status/1891918876029616494
- Agent leaderboard top LLM for AI agents https://x.com/omarsar0/status/1890789258782200308
- EnigmaVal, collection of long complex reasoning challenges, https://x.com/DanHendrycks/status/1890091724594393140
- O3 top active competitors https://x.com/alexwei_/status/1889727569421087217
International News
- Speculation China is working on their Manhattan project too https://x.com/zephyr_z9/status/1891453696971600117
- DeepSeek drives 1.3 Trillion China stock rally https://x.com/ns123abc/status/1891013449779405069
AI Agents
- Building AI agents is 5% AI and 100% software engineering https://x.com/n_sri_laasya/status/1890501360543555604
Decentralized AI ⭐
- Asynchronous training where each copy of the model does local computation [...] makes people uncomfortable but it works https://x.com/Ar_Douillard/status/1890881870503493734
- Decentralized AI, high level overview https://x.com/yb_effect/status/1890503208134471836
- Top 10 reads from 0xJeff https://x.com/Defi0xJeff/status/1887724903849455656
- Naphta decentralized AI agent protocol, launches https://x.com/richardblythman/status/1889703891006267724
Blog Posts
AI for Builders
- Operator + Replit IDE https://x.com/gdb/status/1891047216611037229
- Open AI, Best practices for AI builders https://x.com/OpenAIDevs/status/1890146366665769114