Speed Without Skepticism Is a Design Flaw
Daniel Kahneman's Thinking, Fast and Slow describes two modes of cognition: System 1 fires instantly and feels certain; System 2 is slower, effortful, and catches what System 1 misses. Agents behave like pure System 1. They complete the task, return an answer, and express no uncertainty about whether the task was the right one. A founder who treats that output as finished work has outsourced both execution and judgment, and only one of those was safe to outsource.
Overconfidence Compounds Across a Workflow
Kahneman shows that overconfidence is not a personality flaw — it is the default output of a fast-pattern system operating without friction. In an agentic pipeline, each step inherits the confidence of the last. A subtask completed with unwarranted certainty becomes an assumption baked into the next subtask, and the error compounds quietly. This is precisely why clear tasks, measurable outputs, and feedback loops exist — not to constrain agents but to introduce the friction that System 2 would otherwise supply.
Anchoring Means Your Prompt Is the First Number
Kahneman's anchoring effect shows that the first number in a negotiation pulls all subsequent estimates toward it, regardless of its accuracy. In language-directed software, your prompt is that anchor. Agents will elaborate on whatever frame you give them, confidently and at length. The founder's job is not to write more — it is to audit the frame before issuing it, asking whether the goal set in the prompt is actually the goal worth pursuing, and whether the constraints named are the constraints that matter.
The Planning Fallacy Lives in Your Roadmap
Kahneman named the planning fallacy: people consistently underestimate time, cost, and complexity because they reason from best-case scenarios rather than base rates. Founders building AI-native companies are especially exposed because agents can prototype fast enough to make the whole roadmap feel achievable by Friday. The human role is to hold the base rate — to ask what this class of project has historically cost, what has gone wrong before, and which decisions in the plan are irreversible and deserve a longer pause before shipping.
