We've entered a new coding paradigm. Agents can now generate full features at a pace that outstrips traditional engineering workflows, but the systems we use to review, understand, and ship software still assume humans wrote every line.
For decades, the software development lifecycle was paced by human comprehension. Teams had time and mental bandwidth to align on implementation, intent, and quality before code reached production.
AI-assisted coding changes that balance. Code generation has accelerated, but the diff still does not carry the reasoning, tradeoffs, or intent that shaped it.
According to data from Cursor, PRs are 3x bigger than they were 18 months ago! Consequently, developers resort to rubber-stamping agent-generated code with little understanding or accountability. As ownership erodes, codebases accumulate hidden risks, and poor-quality software inevitably slips into production.
And human cognition isn't the only system under strain.
