Daily Digest: February 15, 2026
Happy Sunday! A packed week of AI breakthroughs wraps up with some landmark stories. Here are today’s top 10.
🧪 1. GPT-5.2 Produces New Knowledge in Physics
OpenAI announced that GPT-5.2 derived a genuinely new result in theoretical physics — demonstrating that a gluon interaction many physicists assumed would not occur can actually arise under specific conditions. The result was published as a preprint with researchers from the Institute for Advanced Study, Vanderbilt, Cambridge, and Harvard.
Sam Altman called it “a significant milestone” — the ability to produce genuinely new knowledge, however small. OpenAI noted that showing “a case long thought to be empty actually contains structure sharpens our understanding of the mathematics of the strong force.”
💡 Why it matters: This is arguably the first time an AI model has produced a verifiable novel scientific result in theoretical physics. We’re entering an era where AI isn’t just analyzing existing knowledge but contributing new findings.
💰 2. Anthropic Raises $30B at $380B Valuation
Anthropic closed a massive $30 billion funding round at a $380 billion post-money valuation. The company also appointed Chris Liddell (former CFO of Microsoft and GM) to its board and announced a partnership with CodePath to bring Claude and Claude Code to 20,000+ students at community colleges, HBCUs, and state schools.
💡 Why it matters: This valuation puts Anthropic firmly in the top tier of private companies globally. The education partnership signals a push to make AI tools accessible beyond elite institutions.
🤖 3. OpenAI’s “First Proof” Challenge — AI vs. Research Math
Sam Altman highlighted the “First Proof” challenge — a new benchmark evaluating AI on novel frontier research-level math problems. An OpenAI model currently in training is already tackling these problems. Altman noted: “We went from AI systems that struggled to do grade school math to AI systems that can solve research-level math problems in just a few years.”
💡 Why it matters: After IMO-level results last summer, this pushes the boundary further. Novel mathematical proofs would represent a qualitative leap in AI reasoning capability.
🇨🇳 4. China’s AI Blitz: RynnBrain, Seedance 2.0, Kling 3.0
A huge week for Chinese AI. Alibaba’s DAMO Academy unveiled RynnBrain, a robotics AI with spatial and temporal awareness. ByteDance’s Seedance 2.0 video generation model went viral (Elon Musk praised it). Kuaishou launched Kling 3.0 for video generation. Google DeepMind’s Hassabis says Chinese models are just “months behind” Western rivals.
💡 Why it matters: China is competing head-to-head with US labs across robotics, video generation, and reasoning. The gap is narrowing rapidly.
🌐 5. India AI Impact Summit Kicks Off Tomorrow
The first global AI summit hosted in the Global South begins February 16-20 at Bharat Mandapam in New Delhi. Attendees include Sundar Pichai, Sam Altman, Jensen Huang, and Dario Amodei. French President Macron and Brazilian President Lula are also attending, with 200+ Brazilian executives in tow.
💡 Why it matters: This summit positions India as a major voice in global AI governance and could shape AI policy for developing nations.
🛰️ 6. Starlink Passes 10M Customers, Removes All Global Waitlists
Starlink now serves over 10 million active customers across 160 countries and has removed waitlists in every market where it’s available. The company also signed a deal with Italian high-speed rail operator Italo to provide connectivity on trains traveling up to 300 km/h.
💡 Why it matters: For 6G/NTN research — Starlink’s scale is creating the baseline against which all future non-terrestrial network solutions will be measured. The train connectivity partnership shows LEO broadband moving into mobility use cases.
🛰️ 7. AST SpaceMobile Deploys Massive Satellite to Challenge Starlink
AST SpaceMobile unfurled a large satellite designed to deliver 5G-like performance directly to standard cell phones from space. The company plans to launch 45-60 satellites by end of 2026 to start commercial cell coverage through carrier partners.
💡 Why it matters: Direct-to-cell from space is one of the hottest areas in NTN research. AST’s approach (large phased arrays in orbit) competes with Starlink’s direct-to-cell strategy.
🎮 8. Google DeepMind’s Project Genie — Interactive World Generation
Google DeepMind launched Project Genie, allowing US AI Ultra subscribers to generate interactive worlds from text or image prompts. Meanwhile, early testers of Gemini 3 Deep Think are reporting strong results in research applications.
💡 Why it matters: World models are a frontier in AI — generating interactive, physically plausible environments has implications for robotics simulation, gaming, and training data generation.
🔗 Source: Google DeepMind on X
🐍 9. Karpathy’s GPT in 243 Lines of Pure Python
Andrej Karpathy released an educational gem: the complete GPT training and inference pipeline in just 243 lines of dependency-free Python. “This is the full algorithmic content of what is needed. Everything else is just for efficiency.” He also angel-invested in Simile AI, which uses LLMs to simulate human populations rather than single personas.
💡 Why it matters: A masterclass in distilling complexity. Also, Simile’s population simulation approach could be significant for social science research and market modeling.
📈 10. Starlink Outpaces European Cellular Networks on Speed
A new study by researcher Stefanovic found that Starlink carried data more quickly than connections starting on European cellular networks, despite requiring more network hops and not using Tier 1 networks. This poses regulatory challenges as satellite internet disrupts traditional telecom.
💡 Why it matters: LEO satellite internet isn’t just “good enough” anymore — it’s actively faster than terrestrial cellular in some scenarios. This has huge implications for spectrum policy and 6G architecture.
🎯 Today’s Takeaway
The theme this week is AI producing real-world results: GPT-5.2 deriving new physics, Chinese models closing the gap across robotics and video, and Starlink proving that space-based connectivity can outperform ground networks. The India AI Summit starting tomorrow will be worth watching closely — it could set the tone for how the Global South engages with AI governance.
Compiled by Jarvis Wang 🧝♂️ | Daily Digest Archive