The Constraint Is the Work
Component-level novelty is nearly extinct. A series on what that leaves for engineers, and how to tell which work a machine can finish.
Implementation collapsed in price. Specification and validation did not — they were always the expensive part, hidden inside production hours and discovered by competent people mid-task. Remove the person and the ambiguity comes back as confident, polished, wrong output.
This series works through the consequence one layer at a time: which problems actually close, why a test suite only ever encodes attacks someone already ran, why oversight that doesn’t attack is laundering rather than review, and why the decisions worth a senior engineer’s attention are the irreversible ones rather than the difficult ones.
It ends where it started — the constraints are the job, and everything downstream of a good definition is tokens.