Engineers traditionally specialize. AI pushes their competence boundary outward into Design and Ops.
Visualizing an engineer's mental model. Isolated clusters connect into a dense mesh.
Security teams doing Frontend; Research doing Infra. The silos break down.
Tasks plotted by "Unfamiliarity" vs "Action". Previously feared tasks move to "Routine".
Volume of code written in languages the engineer does not know fluently.
The tools an engineer commands. AI acts as gravity, pulling more tools into the engineer's orbit.
Claude Code tool chains vs. Human Interventions. The "Human-in-the-loop" frequency drops.
A project timeline spiraling inward. Long outer loops (Manual) tighten into fast inner loops (AI).
The "Impossible Frontier". High complexity tasks migrate to low-time quadrants.
Breakdown of a task. The grey "Blocked/Researching" bars vanish.
Mental energy lost to switching. AI keeps context, reducing cognitive load.
Visualizing commits as particles. From a trickle to a steady stream.
27% of AI work is "Net New"—tasks that simply wouldn't exist without AI efficiency.
P4/P5 tasks (grey dots) usually rot. AI activates them (green) and moves them to Done.
Proportion of time spent on refactoring and quality-of-life fixes.
Grid representing codebase files. Dark spots (untouched legacy code) light up.
Types of tasks completed. From "Features Only" to a balanced diet of Tests, Docs, and Tools.
The cost threshold for building internal tools drops below the value line.