Every reliable process in the world is a protocol: auctions, double-entry bookkeeping, brokerage account transfers, or APIs. Protocols reduce ambiguity but the world remains messy. EDI, SOAP, REST, gRPC, and LSPs were all designed to collapse complex realities into deterministic schemas clients could consume. And yet you've still emailed another team to ask why their "to spec" endpoint wasn't working. It matched the schema, but the meaning stayed in their head.

Agent harnesses are general-purpose executors that run protocols with customizable judgement and permissions. LLM inference provides discretion, tool calls map decisions to action, and your protocol (workflow) can live in plain English. Judgement-dense, high-variance work can now be made less manual with the right engineering. The popularity of agents for everything sounds attractive until you go beyond prototypes. The manual version functioned because people solved problems silently. Watch the project fall flat if you automate before understanding the protocol.

In this session (for senior+ individual contributors) you'll learn:

• A brief history of APIs: Why nobody's APIs are RESTful and how gRPC never scrubbed off the SOAP scum of versioned ambiguity.
• Seeing like a developer (or agent) in a hurry: How to better explain APIs that already exist that you may not be able to change.
• Designing protocols: Using event storming exercises, mapping techniques, and fast prototyping to avoid building the wrong thing.
• Scaling painlessly: How to observe small-scale nits that become regional outages when deployed to millions of clients.
  You'll leave this talk able to avoid classic blunders and build APIs that work in production.