Resisting Cognitive Surrender
Kunal Sheth's working note on the trap of accepting agent output without thinking, and how he tries to keep control: he delegates only where mistakes are cheap, and drives the work by hand on code he cares about. The core worry is that comprehension, not building, is now the bottleneck, and skipping it erodes both quality and his understanding of his own systems.
- Agents make work feel fast and productive, but complexity and technical debt compound quietly, and his familiarity with the codebase erodes with each automated fix.
- Models sound confident even when wrong, and peer pressure to ship more PRs pushes people to stop checking, which he treats as the warning sign to watch for.
- He defines a "vibe limit," the complexity of change he can make confidently with agents, and chooses deliberately where to surrender agency versus where to spend his own time, sometimes driving every edit by hand.
Kunal's note is about "cognitive surrender," the habit of accepting whatever an agent outputs without thinking it through. He admits the workflow is fast, low-effort, and seductive, but says he keeps ending up with bad decisions, accumulating technical debt, and losing familiarity with the systems he works on. Part of the pull is the classic programmer's laziness, part is peer pressure (if colleagues look like they've handed everything off, slowing down makes you look worse), and part is that models sound extremely confident, even when they're wrong. He notes that OSS projects under time pressure seem to balloon in complexity, with agents amplifying the problem, and that the cost of complexity is slow and easy to ignore until it compounds.
The core of his argument is about comprehension. He draws on Lisanne Bainbridge's 1983 paper "The Ironies of Automation," arguing that when agents handle the routine work, you never practice the easy tasks, so you're worst-equipped exactly when the agent gets stuck. He describes a pattern where a bug sends him into one more agent session, the automated fix erodes his understanding further, and his diagnostic judgment decays in a loop. His position is that he remains accountable for what he ships, so surrendering agency leaves him responsible without ownership. He still sees real upside, since agents let him build things he couldn't otherwise, but he's blunt that agents can drive build time toward zero while comprehension remains the bottleneck, and that the two can't both be had for free.
His response is deliberate triage. He defines a "vibe limit," the complexity of change he can make confidently with agents, and plots it against risk. Throwaway or well-sandboxed work can be handed off, while important, fragile, hard-to-check, or poorly isolated systems need his direct attention. He treats this as ongoing recalibration, comparing predictions against outcomes and revisiting the boundary when new models ship, much like leading a team at a faster pace. He also pushes for aggressive simplicity and modularity in anything meant to keep running, reserving throwaway-grade complexity for proofs of concept.