Open weights become infrastructure for agency. GLM-5 opens a trainable core for long-horizon agents. Open weights cut friction — and raise coordination demands. Principle 09 — Agentic Coordination is a New Org.

Agents need legible landscapes to reason about trade-offs. A model that excels at long-horizon reasoning makes those landscapes visible. Legible landscapes expose both opportunity and risk. Principle 06 — Legible Landscapes.

Outcome engineering demands artifacts and guardrails. Ship toolkits, benchmarks, and orchestration maps now. Open models without artifacts are brittle. Principle 08 — Ship the Artifacts.