AI Infrastructure
Where models actually run, and what breaks when a provider pulls the plug.
4 articles
The model you build on is rented. Price the risk.
Anthropic's two newest models went dark worldwide for 19 days on an export-control order — zero notice, and the provider didn't choose the timing either. Most teams price that risk at zero. Here is what it actually costs, and the seam that makes a swap a config change.
If you don't host it, you rent it: reading Odyssey's $1.45B against Fable's blackout
Odyssey's $1.45B raise and Fable 5's overnight shutdown are the same story. How to architect for model-provider dependency: gateways, version pinning, and naming your no-fallback bets.
Google's data agents and MCP Toolbox: what they're for and how to set one up
What Google's data agents and MCP Toolbox actually do, which databases you can connect, the realistic outcomes — and a hands-on setup: install, tools.yaml, and identity-scoped auth.
Charge AI Bots for Content: AWS WAF Starts Billing Crawlers
AWS WAF can now meter and charge AI bots per request instead of blocking them. What's GA, how pay-per-crawl works, and where the edge tollbooth breaks.