Lesson 34 — workflow vs agent. The single most over-claimed distinction in this layer. Most “agentic” deployments are workflows with a marketing budget. That is fine; just call it what it is. A fixed five-step LLM pipeline is a workflow, not an agent A team builds an “agentic” customer service system. Under the hood, every…

Lesson 35 — bounded agents and their traces. The trace is the artifact. Without it, you are running a black box you cannot review. With it, the agent is just another auditable component. Without budgets an agent wanders off-task and burns money; bounds make it reviewable An agent is given a research task. It is…

Lesson 36 — the agent failure mode heatmap. Agents fail in predictable shapes. Knowing the shapes lets you instrument for them before the failure ships. A clean trace can still hide hallucinated facts, wrong tool calls, and dropped constraints An agent finishes its task. The trace looks clean. The output looks right. A week later,…

Lesson 37 — the Layer 10 evaluation gate. Pass this layer when you can hand a safety review board the agent’s bounded contract, its failure-mode coverage, and its trace-replay evidence. The audit is the deployment permit. The agent can act unapproved only through a failure mode you did not instrument “Can the agent take an…