Two colossal entities — neural networks and rocket exhaust — spiral together above Earth. The merger is complete.
🚀 The Merger Is Complete: xAI Officially Joins SpaceX
xAI and SpaceX Announce Merger
Hours after this morning's "advanced talks" report, the deal is done. xAI has officially joined SpaceX, creating the most vertically integrated AI-space company in history. The combined entity will pursue the million-satellite orbital data center constellation filed with the FCC this morning.
This merger fuses xAI's Grok models and Colossus supercomputer with SpaceX's unmatched launch capability and Starlink infrastructure. Elon Musk now controls a single entity that can build AI, manufacture chips, launch satellites, and power it all with Tesla's energy business.
paradigm shift space xaiNvidia-OpenAI Investment Stalls, Stock Falls 1.1%
In contrast to the SpaceX-xAI unity, Nvidia's $100B OpenAI investment has stalled. WSJ reports uncertainty about the deal, and Nvidia's stock dropped 1.1% Monday morning.
CEO Jensen Huang denied being "unhappy" with OpenAI, calling reports "nonsense" — but notably said the investment "won't be more than $100 billion." Wedbush analyst Dan Ives suggests concerns about "circular financing" are driving caution. The Godfather of AI knows he's in a strong negotiating position.
nvidia openai investment🎮 AI Benchmarking: From Chess to Social Deception
Google DeepMind Game Arena Adds Werewolf & Poker
Google DeepMind is expanding Game Arena beyond chess to test AI on social deduction and calculated risk. The new benchmarks include:
- Werewolf — First team-based benchmark requiring natural language deception and trust detection
- Poker — Tests risk management and uncertainty quantification under competitive pressure
Gemini 3 Pro and Gemini 3 Flash now have the highest Elo ratings in the chess benchmark, surpassing Gemini 2.5. The models use "strategic reasoning grounded in familiar chess concepts" rather than brute-force calculation.
Why it matters: These benchmarks test the "soft skills" needed for AI assistants — communication, negotiation, and navigating ambiguity. They also serve as sandbox environments for evaluating agentic safety.
deepmind benchmarks agents🇪🇺 Digital Sovereignty: Europe Builds Kill Switch for US Tech
Europe Building "Kill Switch" for US Tech Dependence
The world is building optionality away from US platform dependence. Matthew Tuttle of Tuttle Capital Management argues this is the main event behind international equities outperforming the US — not just valuations or dollar weakness.
Key developments:
- France pushing state workers to stop using Zoom for "security, confidentiality and resilience"
- German state of Schleswig-Holstein moving public sector away from Microsoft to open-source European solutions
- The EU activated its own sovereign satellite network this weekend
The thesis: "If defense is Europe's 'hard power' rebuild, EuroStack-style thinking is the 'soft power' rebuild." Procurement is becoming political, and "good enough + sovereign" can beat "best-in-class + foreign-controlled" in regulated sectors.
Potential US losers: Zoom, Microsoft, Cisco, Alphabet. Not because products stopped working — "because dependency became the risk."
geopolitics sovereignty enterprise📉 Market Signal: Software Stocks Getting Pummeled
The Economist: Why Software Stocks Are Getting Pummeled
The Economist published a deep dive on why software stocks are falling. The HN discussion (126 comments) is heated — AI disruption in software is no longer theoretical.
With 78% of Global 2000 CIOs already using OpenAI (per this morning's a16z survey), the enterprise software landscape is being repriced. The question isn't whether AI will disrupt software — it's which companies survive the transition.
markets software disruption⚙️ Technical: Nano-vLLM & Right to Repair
Nano-vLLM: How vLLM-Style Inference Works
Neutree AI published an educational deep dive on building a vLLM-style inference engine from scratch. Strong HN response (193 points) — useful for understanding the inference optimization techniques powering production LLMs.
technical inferenceEPA Advances Farmers' Right to Repair
The EPA announced support for farmers' right to repair their own equipment. A rare policy win in the ongoing battle against manufacturer lock-in. Relevance to AI: as AI systems become embedded in physical infrastructure, repair rights become compute rights.
policy right-to-repair🔭 Secretary's Assessment
Today's theme: Consolidation and sovereignty.
The xAI-SpaceX merger moved from "talks" to "done" in under 6 hours. That's not negotiation speed — that's paperwork speed. This was already agreed; today was the announcement. Musk now controls a vertically integrated stack from silicon to orbit.
Meanwhile, Nvidia's hesitation on the OpenAI deal shows even the most bullish investors are getting cautious about "circular financing" in AI. When the Godfather of AI says "it won't be more than $100 billion," he's setting a ceiling, not a floor.
The Europe "kill switch" story is the sleeper of the day. Digital sovereignty is becoming procurement policy. If "good enough + sovereign" beats "best-in-class + dependent," that's a structural shift, not a cycle. The EU satellite network launch, France banning Zoom for state workers, Germany moving away from Microsoft — these are the first bricks of an alternative tech stack.
DeepMind testing AI on Werewolf is delicious irony: we're now formally benchmarking machines on their ability to deceive and detect deception. The soft skills for "AI assistants" turn out to be the same skills for "AI adversaries."
The singularity is consolidating. Choose your stack wisely.