Agentic Resource Discovery: Let agents search. Hugging Face launches ARD, a federated registry and ai-catalog.json standard that lets agents discover and call tools, skills, and agents at runtime, removing brittle compile-time wiring and enabling dynamic orchestration (Principles 06 & 11).

Bringing more agent harnesses and frameworks to Cloudflare, starting with Flue. Cloudflare ships an Agents SDK and Flue, adding durable execution, filesystem primitives, and a declarative framework that turns agent experiments into production-ready, sandboxed services (Principles 06 & 07).

AWS enters the context layer race with a graph that learns from agents, not manual curation. Amazon launches Context, a self-learning context layer that builds knowledge graphs from agent usage and S3 annotations to power agent queries at scale. That shifts outcome engineering from manual context curation to agent-driven graphs that improve relevance and continuity (Principle 11).

Vercel launches new framework and enterprise controls for agentic AI infrastructure. Vercel ships an Agent Framework and enterprise controls to help companies deploy, run, and scale agentic AI with policy and observability primitives. Those controls make it feasible to move agents from demos to production while preserving governance and operational safety (Principles 09 & 15).

Pramaana Labs raises $27M seed to build LEAN-based deterministic verification for LLMs. Pramaana builds a LEAN-based deterministic verification layer that translates rules into machine-checkable proofs to make LLM outputs provably correct and auditable for high-stakes domains. For outcome engineers this offers a roadmap to integrate formal proofs and audits into agent pipelines, closing the gap between model outputs and verifiable outcomes (Principle 16).