Hiring
“AI-augmented” has followed the same path as “agile” and “full-stack”: a specific idea that became a checkbox. Almost everyone lists it. Very few can describe what changed in their day.
What the claim should translate to
For a developer, it means the AI writes a meaningful share of the first draft and the engineer decides what survives. Speed comes from the tool; correctness comes from the review. Someone forwarding model output unreviewed is not augmented — they’re a relay.
- Developers — Copilot and Cursor in the editor daily, with the judgment to reject what doesn’t hold up.
- Writers — AI for research and structure, a person for the specifics and the point of view.
- Designers — rapid variation and exploration, with human decisions about what ships.
- Assistants — automating their own recurring work so their hours move to judgment tasks.
How to verify it in an interview
Screen shares beat claims. Ask someone to do a small piece of real work in their own setup while you watch. In fifteen minutes you learn more than an hour of discussion: whether the tooling is genuinely part of their workflow, and whether they read what it produces.
Watch someone work for fifteen minutes and you will know more than a resume can tell you in an hour.
Where the multiplier stops
AI assistance compounds on well-understood, well-specified work and does close to nothing for the hard part of engineering, which is deciding what to build and how the pieces should fit. A team that was unclear about its direction does not become clear faster with better autocomplete.
That’s the honest version of the claim: real gains on execution, no gains on judgment. Which is exactly why we hire for judgment first.
Posted by