A model can give an answer.
Teaching takes a system.

Install the tools, run a local teaching workflow, and find the code you want to change.

Change prompts, add tools, connect Pi packages, and keep clients and evidence aligned.

Follow a learner request through clients, the teaching engine, model execution, and stored artifacts.

The teaching loop, learner state, verification, and the limits of automated improvement.

Choose the right entry point for teaching, reviewing, exporting, and running Keating.

Follow the shared teaching system across browser, desktop, mobile, and terminal—and understand where their capabilities differ.

Turn study material into a shared course while preserving membership, permissions, learner work, and delivery state.

Understand what is saved, where credentials belong, and how an export differs from consent to share.

Keep model selection, credentials, teaching behavior, and workspace execution independently configurable.

Find the right subsystem, make a bounded change, and show the evidence needed to review it.

Inspect teaching evidence, run bounded experiments, and distinguish benchmark eligibility from demonstrated human learning.