TorchNeuron runs PyTorch on Trainium with a one-line device change and is tested nightly on 100,000+ Hugging Face models. It removes the cost of trying another chip. Running fast still takes kernel work, as Deepgram's multi-month port shows.
DeepSeek priced Search and Fetch at $0.75 per 1,000 successful requests, put them on the same API key as its models, and made server-side Search and Fetch calls through Claude Code and Codex free for DeepSeek API users until November 1. That is how agent search becomes a commodity.
Google paid Character.AI $2.7 billion in 2024. Disney is paying the bankrupt company $20 million while budgeting about $180 million for its staff. The reverse acqui-hire works for buyers and hires, and leaves users and creditors with the shell.
Vercel's switching data, Replit's new default, DeepSeek's silent upgrade, and OpenAI's shutdown calendar point the same way: the model is now a replaceable part, and the power sits with whoever controls which part gets used.
