I have always despised when job postings listed “the ability to multi-task” as a desired skill for a candidate. People, much like single-processor computers, imitate multi-tasking via fast context switching and managing the interrupt triggers. This thought from a very long time ago, led me to wonder how many parallels naturally occur between computer patterns and how people think and a new variation occurred to me recently. What if Daniel Kahneman’s dual-process model (System 1 and System 2), introduced in his book Thinking, Fast and Slow can be applied to AI trends such as Agentic Coding? Super brief summary of some ideas from Thinking Fast and Slow: The 2 systems… System 1 — Fast • Automatic, intuitive, effortless • Runs constantly, like breathing • Completes "bread and…", reads hostility in a face, swerves at danger • Generates impressions and feelings that become your beliefs System 2 — Slow • Deliberate, effortful, logical • Lazy — avoids work whenever possible • Multiplies 17 × 24, fills tax forms, checks a math proof • It is supposed to monitor System 1, but usually just endorses whatever System 1 suggests The danger: System 1 writes most of the story of your life, and you never notice. 3 cool concepts from the book (there are more, read the book!)... WYSIATI — "What You See Is All There Is" System 1 builds a coherent story from the information in front of you and never asks "what am I missing?" Anchoring The first number you see drags every later estimate toward it. In Kahneman's famous experiment, UN delegates spun a rigged roulette wheel (10 or 65), then estimated how many African nations were in the UN. The wheel shifted their answers by 20 percentage points — even though everyone knew it was random. Availability Heuristic We judge likelihood by how easily examples come to mind, not by actual frequency. Plane crashes get more coverage than car crashes, so we overestimate flying risk. System 1 issues are what most AI products and implementations solve today and are well-suited to. AI is the most powerful tool ever created for performing System 1 level tasks, especially related to programming, where parallel structure, compliance with standards are deeply related to maintainability and error detection. AI is amazing at cross-reference, documenting patterns, and creating task lists to identify gaps between a written system (code), best practices, and company policies. System 2 is the part that still requires a human in the loop. The coding activity of building software has become cheap. The important part now is to ask questions that require taste, discernment, and a desire to improve our chosen field so we can counteract these problem and trends. Kahnneman’s book presents the 3 concepts above that are traps of System 1 approaches, and are only countered by System 2 approaches. WYSIATI -- To counteract "What You See Is All There Is" in application development, ask these questions first: -We can build any application. Should we build this particular thing? -If we build this, what else could we do with it that would be beneficial? -What negative effects could occur if we build it? Anchoring –You will have multiple iteration cycles on any product you decide to build. Every iteration will be connected to the first iteration you build. To preempt anchoring in software development, ask this before you build a new product: -Is this use of our time and attention in building this application going to give us the most return for our investment compared to other things we could have built instead? -How can I feel certain enough about my choice? -If you are already building, ask, “Have we proven this idea won’t work yet?” Availability Heuristic – Building because you saw someone else succeed, or you want to replace the paid product you use. Just because you are aware of Uber or Microsoft Office tools, does not mean you should copy those products. Ask this instead: -What gives my idea the strongest possibility of success? -What advantages do the current leaders in that market have that I cannot resolve? (Microsoft has an immense and deep distribution advantage, decades of business experience, and nearly limitless capital. Uber already has international name recognition and large user base). Kahneman recommends these practical ways to pick the right times to invoke System 2 thinking, and we can use AI technology to smash through System 1 tasks in historically unprecedented speed, as well as to gather information that will support us in the System 2 realms. • Slow down important choices • Use checklists and reference classes (gathered with research and analysis driven by AI) • Run premortems (imagine it failed — why?) • Adopt a shared vocabulary so colleagues can name a bias out loud What other ideas about how humans have good ideas and cognitive errors can be applied to current technology trends?