Alchemy v2 agent self-deploys, self-modifies

This is a strong “agents as infrastructure” move: the model doesn’t just call tools, it operates on the same typed system that defines the app. The upside is less glue code and fewer mismatches between intent, deploy plan, and runtime state; the downside is that self-modifying systems need very disciplined guardrails.

• –Type-safe resources plus Effect give the agent a real execution model, not just an API wrapper
• –Self-deploy and self-modify are powerful for meta-automation, but they raise the bar on preview, diff, and approval flows
• –The positioning is especially compelling for TypeScript shops building AI-native infra or internal devtools
• –Because the stack is open source and opinionated around modern JS runtimes, it may resonate more than generic IaC wrappers with AI bolted on
• –This feels less like a toy agent demo and more like an attempt to make agents part of the deployment substrate