Further Learning Paths

Skills decay without use, and AI capability shifts fast enough that standing still means sliding backward. But unfocused "keep learning about AI" advice wastes time in a different way than not learning at all. So every path below follows one rule from the program: learn with a workflow you own in mind. Read the chapter, watch the talk, then apply it to the thing on your registry — or it did not happen.

All of these live on the Resources page, which stays available after the cohort ends.

Everyone

Co-Intelligence (Ethan Mollick) — the book-length version of the working stance this program taught. Read it in the month after the capstone, while the habits are forming.
Anthropic Academy — free, self-paced courses that extend Phase 1 skills. Ideal for new joiners who missed the cohort; pair each new hire with your Phase 1 worksheets and the Academy, and they can reach the team's baseline without a facilitator.
How I use LLMs (Karpathy) — rewatch three months in. It reads differently once you have your own daily practice to compare against.

Builders and workflow owners

Building Effective Agents (Anthropic engineering) — reread before each new agentic workflow; your second agent should be designed, not improvised.
Claude Code and MCP documentation — the working references for anything you connect or automate; treat them like the standard library docs of your new toolchain.
The Claude Code talk (Boris Cherny) — where the tool's design philosophy comes from, useful when you are deciding how far to push automation.

Leaders and sponsors

Prediction Machines (Agrawal, Gans & Goldfarb) — the economics: what gets cheap when prediction gets cheap, and where judgment gains value. The frame for your next year of AI investment decisions.
The Coming Wave (Suleyman) — the governance stakes at societal scale; context for why your Phase 3 controls are not bureaucracy.
OpenAI DevDay keynote and provider announcements — watch the industry's direction once a quarter, then ask the only question that matters: which of our registry workflows does this change?

Process and people owners

Working with AI (Davenport & Miller) — thirty real cases of jobs redesigned around AI; mine it for patterns that match your departments.
Human + Machine (Daugherty & Wilson) — the "missing middle" framing for deciding what stays human, what automates, and what becomes collaboration.
The practitioner interviews — revisit the one for any function about to start its second workflow; their second-workflow mistakes are listed so you can skip them.

A cadence that survives contact with the calendar

One item per person per month, chosen from their path, discussed for ten minutes in the champions cadence: what did you apply, to which workflow, what changed. That is the whole system. It is small enough to survive busy quarters — and after a year it compounds into a team that nobody has to train from scratch again.

Definition of done

Every workflow owner and sponsor has picked their path and their first item, and the monthly learning slot exists in the champions cadence.