Coding Fast and Slow: Applying Kahneman's Insights to Improve Development Practices and Efficiency
Abstract
Context destruction is invisible. You don’t feel it happen. You don’t know when you’re depleted. You don’t see the OK code you produce. Neither do your reviewers or your pipelines. The first person who notices is your customer. This talk applies Nobel laureate Daniel Kahneman’s System 1/System 2 framework to developer productivity. Through interactive experiments — cognitive puzzles, code reviews under distraction, and the Stroop effect — you’ll experience context depletion firsthand. Then we’ll trace the invisible chain from context destruction through OK code to production, backed by research proving that interruptions reliably degrade work quality and people cannot detect the drop (Foroughi et al., 2014 & 2016). The solution is Context Engineering across four dimensions: Personal (manage your fuel), Organizational (protect your team’s fuel), Technological (automate context protection), and AI (teach AI agents your context so you stop babysitting them). Because you can’t detect depletion — so stop causing it.
Resources
Books
- Thinking, Fast and Slow by Daniel Kahneman
- Attention Span by Gloria Mark
- Flow: The Psychology of Optimal Experience by Csikszentmihalyi
- Outliers: The Story of Success by Malcolm Gladwell
- Take a Nap! Change Your Life by Sara Mednick
- The Power of the Downstate by Sara Mednick
- Liquid Software by Baruch Sadogursky
- DevOps Tools for Java Developers by Baruch Sadogursky
Research Papers
- Foroughi et al. 2014 — “Do Interruptions Affect Quality of Work?”
- Foroughi et al. 2016 — “Are Individuals Sensitive to Changes in Performance when Interrupted?”
- Attention and Capacity Limits in Perception: A Cellular Metabolism Account
- Mental Fatigue During Programming Tasks
Tools
- Reclaim.ai — AI calendar for personal + organizational context protection
- Tessl good-oss-citizen skill — 0%→100% eval scenario
- Tessl.io — Make agents work in real codebases
- Claw Market Map Q1 2026 — AI agent landscape