Part 4/12:
Definition: Curating and maintaining relevant tokens—the relevant context—for long-running tasks and agents.
Impact: This discipline acknowledges that models degrade as context grows, so it's about providing the right tokens—documents, instructions, system prompts—that guide the AI effectively over long sessions. Think of it as building a rich environment where the agent can operate without constantly fetching additional info, significantly boosting efficiency.
3. Intent Engineering
Definition: Encoding organizational purpose, goals, values, and decision boundaries into infrastructure that guides autonomous agents.