We've interviewed over 200 engineers in the past 12 months and made offers to 14. The strongest signal of production output in 2026 isn't years of experience or framework knowledge, it's something we now call AI literacy.
What AI literacy actually means
It's not "uses Copilot." That's table stakes now. AI-literate engineers do four things differently:
- Prompt with context. They give the AI the goal, the constraints, and the surrounding code. They don't accept the first answer; they iterate.
- Know when to override. They can spot when AI-suggested code is wrong, unsafe, or just stylistically off, and they fix it without thinking.
- Use AI as a thought-partner, not a typing-partner. They debate architecture with Claude or ChatGPT before writing the code.
- Cap their own AI usage. They know when generating code is faster than writing it, and when it isn't.
The skills we now under-value
Memorisation of framework APIs. Esoteric syntax knowledge. Being able to write Big-O analysis on a whiteboard. None of these predict who ships in our shop.
The skills we now over-value
- Code review quality. Reviewing AI-generated code well is the new core skill.
- System thinking. When AI writes 80% of the lines, the 20% that's architecture matters more.
- Writing. Engineers who write clear PR descriptions, design docs, and prompts ship faster.
- Taste. The ability to recognise a good solution from three plausible-looking ones AI offered.
How we interview now
We give candidates an AI assistant during the take-home task. We watch how they use it. The pattern is unmistakable, strong engineers use AI like a senior pair-programmer; weak engineers use it like a copy-paste machine.
Hiring like this yourself? See our dedicated-developer service, we staff web, mobile and design roles. Want to join a team that hires this way? We have open roles.