OpenCode builds provider-agnostic LLM layer

Hot take: this is the kind of plumbing that only looks boring until you need to support real multi-provider behavior at scale, then it becomes the product moat.

• –The core value is normalization: provider-specific edge cases are where most abstraction layers break, so the implementation quality here matters more than the surface API.
• –Shipping both Effect and vanilla APIs is a smart distribution move because it keeps the internal architecture strong without forcing every consumer into one runtime style.
• –If OpenCode gets this right, it can reuse the same provider layer across the repo and expose a credible standalone library later.
• –The post reads like an in-progress product update, not a launch; the main signal is technical direction rather than a finished release.