Principles and practices for keeping engineering judgement, ownership, and craft intact in an AI-enabled world.
Software engineering has always been a human endeavour. The skill, the judgement, and the craft live with the people doing it — and tooling, for every previous generation, has supported that rather than replaced it. AI is powerful enough to change that equation. Human-First Engineering is about defining new ways of working with these tools that keep the skill and craft of software engineers intact, so we stay in control of the engineering and do not deskill a generation in the process.
Where previous tools automated the tedious work, new AI-powered tools can shortcut the very moments engineers learn from — reading code carefully, working through a bug, drafting a solution and realising it is wrong. Those moments of friction are how engineers build the intuition they rely on later.
Automate them at scale and you risk producing a generation of engineers who can assemble systems without ever building the mental models behind them — quietly eroding the pipeline from junior to senior to principal.
And that risk is not evenly distributed. It falls most heavily on the engineers entering the industry today. They are arriving into a job where the visible tasks juniors traditionally learned from — writing small units of code, fixing mechanical bugs, producing first drafts — are precisely the ones AI now does fastest. The opportunity to struggle, and to learn from that struggle, has to be protected deliberately. It will not happen by default.