Claude Code speed vs growth: a dev's AI dilemma

The account is a first-person case study, not a product announcement, and its value lies in the gap it describes between output and absorption. Delegating implementation to agents shortened delivery time, but the developer notes that the weight of deciding what to settle before implementation actually increased — requirements, design trade-offs, future changeability — all pre-AI software engineering basics.

The twist is that those same decisions were the ones he was handing to AI each day, because a recommendation in minutes beat his own 30 minutes of thought. The forgetting incident following an AI-iterated design made that trade-off visible: understanding in the moment did not become durable knowledge. Efficiency also failed to open learning time, since more work was accepted and reviewing generated code became the bottleneck.

His response is deliberately small — one AI-posed design question each morning, answered from memory and then critiqued — treating AI as an amplifier whose effect scales with what the person brings. He is explicit that he does not yet know how much this raises his skill or whether the learning sticks.