The through-line
Platform vendors are shipping inference optimizations (SageMaker, Bedrock, Cloudflare) while tooling layers expand to surface those capabilities—MCP via WebMCP, spend controls via Vercel Enterprise. The constraint is becoming less compute and more observability of what's actually running.
Issue signature
The Wire
- Amazon SageMaker Inference: 2026 year-to-date launches in review
- Spend Management expands to Enterprise Flexible Commitment plans
- WebMCP support now available in mcp-handler
- MilleMiglia: A realistic instance generator for middle-mile logistics
- Saving another 100TB of RAM with math (and Rust)
- Introducing Kimi K3 on Amazon Bedrock
- Should you read the code, is RAG dead, and did Skills kill MCP?
- New experts join Google’s AI & Economy team
- Introducing the Australian Youth Safety Blueprint
- Does AI need an antitrust exemption so it doesn’t kill everyone????