Your brain has two systems. System 1 is fast, intuitive, and wrong more often than you'd think. System 2 does the slow, deliberate work: debugging, architecture, catching what System 1 missed. Every context switch burns System 2 fuel, and every distraction drains the tank further, which is why developers write code for only 52 productive minutes per day.

AI agents make this worse: they generate at System 1 speed and System 1 quality while demanding System 2 oversight for every line. Every alt-tab between your agent's output and your own reasoning is a context destruction event.

The human problem and the AI problem share a root cause, and context engineering fixes both.

- Personal: flow protection, cognitive load, and why a 20-minute nap is backed by more science than your daily standup meeting.
- Organizational: why back-to-office mandates are context destruction policies.
- Technological: why your AI agent spent 4 hours on a build failure caused by an AWS outage, and how domain knowledge plugins took it from 20% to 95% success.

12 deliveries over 2.5 years — my most battle-tested talk. Includes Devoxx Poland, DevNexus, GeeCON, BaselOne, Dev2Next, KCDC, UberConf, JConf.Dev, Shift, and Voxxed Days Amsterdam 2026. Consistently high ratings. Interactive format — audience participates in cognitive science experiments (Stroop test, bat-and-ball riddle, attention puzzles). The 2026 version adds a substantial context engineering module that ties behavioral science to AI tooling. Works as 45-min session or keynote. Solo talk.