Breaking news, AGI updates, and analysis of the intelligence revolution.
📰 News OpenAI just staked its claim on domain-specific AI with GPT-Rosalind — a life sciences model that beats its own general-purpose flagship on real-world biology benchmarks.
📰 News Safety training doesn't remove harmful capabilities from AI models — it packs them tighter. New mechanistic interpretability research reveals why current alignment approaches are fundamentally brittle.
📰 News With half of US data centers delayed by power limits, the largest orbital compute cluster is open for business — and AI infrastructure may be heading to space.
📰 News Four companies. Two-thirds of all AI venture capital. The AI investment tidal wave of Q1 2026 isn't a bubble — it's a tectonic shift.
📰 News ShengShu's $290M raise signals China is investing aggressively in world models as an alternative route to AGI — one that could leapfrog text-only AI.
📰 News OpenAI wants liability shields. Anthropic doesn't. The battle over Illinois SB 3444 could define AI accountability for a generation.
📰 News DeepMind's new hire has 'Philosopher' on his business card. Henry Shevlin will tackle machine consciousness and AGI readiness — because the question of whether AI can be aware just left the seminar room.
📰 News The agentic AI revolution is running into a very old problem: building a demo is easy. Running a production system is hard. Gartner says nearly half these projects won't survive to 2028.
📰 News SoftBank, Sony, Honda, and Japan's government are building a sovereign physical AI platform. The target: autonomous robots and vehicles by 2030.
📰 News Victor reads 500 classified military repositories, cites its sources, and answers soldiers' questions in real time. The US Army just built its first combat AI chatbot.
⚡ Breaking Even as the Pentagon has blacklisted Anthropic as a supply-chain risk, the company is in direct talks with the Trump administration about its next frontier model. The collision of AI safety and national security has entered a new phase.
📰 News OpenAI wants citizen stakes in AGI profits via public wealth funds. It sounds like ownership. It might be dependency.