Enterprise
Agent Platforms
Courses 1–3 taught you to build one harness, secure it, and customize the model. Course 4 teaches you to operate hundreds of agents as managed infrastructure — with cost controls, provider routing, governance, and the delivery layer that turns a harness into a platform. The thesis: the single harness is a prototype; the platform is the product.
Agents consume up to 1,000x more tokens than chat on identical tasks (Stanford-MIT). 62% of agent inference bills are re-sent context. Fewer than 1 in 3 organizations can attribute AI spend to outcomes. 40% of agent projects will be cancelled by 2027 — for cost, not capability.
| Layer | What it is | This course |
|---|---|---|
| 1. Provider | Frontier, open-weights, self-hosted | E01–E02, E08 |
| 2. Gateway | Routing, fallback, rate limits, budget | E01, E03–E04 |
| 3. Harness | Pi-derived agent (Courses 1–3) | E00, E06–E07 |
| 4. Application | Slack, email, Jira, CI/CD, ERP | E09–E12 |
This course index is live. Module content is in production — in production.
Prerequisite: Course 1 or equivalent production harness experience. Status: in production.