AI Strategy
Briefs and field notes on the build-out of AI — compute, power, capex, and the enterprises deploying it. Written for operators and investors.
'Any One Model Can Go Away': The Case for a Model-Independent AI Stack
Four Amazon Nova models reach end of life in September, and Satya Nadella is telling companies to keep context, memory, and evaluations outside any single model. What a model-independent AI stack looks like, the six assets to keep, and the seven steps that build a tested path from one model to the next.
The Library
Claude Mythos Preview: Inside the Model Anthropic Won't Release
Anthropic built its most capable model yet, then decided not to release it. Mythos Preview cracked Firefox at 84%, beat GPT-5.4 and Gemini 3.1 Pro across the board, and triggered Project Glasswing, a 12-partner defensive cybersecurity program. Here is what the benchmarks show and why Anthropic's own researchers call it the riskiest model they have ever trained.
The Rise of Neoclouds: How AI-Specialized Clouds Became a $35 Billion Market
A new class of GPU-first cloud providers has exploded into a $35 billion market in barely two years. Here is how neoclouds work, who the key players are, and what comes next.
The Agentic Enterprise ROI Tipping Point
Enterprise AI has moved from productivity pilots to revenue systems. March 2026 data shows 74% of enterprises reporting measurable ROI in year one.
NVIDIA GTC 2026 and the Agentic AI Era
Jensen Huang previews Vera Rubin and Feynman architectures, NemoClaw agent platform, and the shift from generative to agentic AI.
The 2026 CapEx Shift: Training Clusters to Inference Grids
$600B+ in hyperscaler capex, the training-to-inference pivot, and how the network topology is changing.
AI Infrastructure Outlook (Dec 2025)
How AI becomes infrastructure: chips, data centers, and power — plus what the next quarter could break.