How I Use AI (timkellogg.me)

This piece frames “AI coding” less as a set of tools or recipes and more as an operational ethos built on two principles: ownership and exploiting gradients. Ownership means you must take full responsibility for AI-produced code—commit it under your name, understand what the algorithm actually did, and be prepared to debug, refactor and answer for its failures. The author likens this to management practices (Genchi genbutsu, “management by walking around”) rather than traditional individual software engineering: success requires oversight, verification, and orchestration of AI outputs across your codebase and org. “Exploit gradients” is about finding high-leverage, low-effort opportunities where small investments in AI produce outsized returns—rapid proofs-of-concept, quick dashboards, or targeted refactors that reveal whether an idea is viable. The practical upshot: workflows will shift toward rapid experimentation, human-in-the-loop validation, automated tests/monitoring, and smarter task allocation. That shift affects hiring and career paths—managers must teach and give room to learn, and adaptable engineers who embrace ownership and opportunistic thinking will outcompete those who treat AI merely as a coding shortcut.