Craft, juniors and career ladders

Before you decide whether losing production work matters, name what it taught. Every item here is a judgement skill that arrived disguised as a chore:

• Resizing taught hierarchy. Squeeze a layout into eleven formats and you discover which element is genuinely primary, because the other ten drop out and the layout still has to work.
• Variant sets taught systems. Build forty variants by hand and you invent components, because doing it manually is unbearable. That is how designers arrive at abstraction honestly rather than by being told.
• Twelve options for a review taught exploration. Not because twelve were good, but because you learned what your first idea is worth relative to your ninth.
• Exports and specs taught states. Nothing teaches empty, loading, error, long-name and zero-data states like being the person who has to produce every one of them.
• Being redlined taught the standard. The single fastest teacher in design is a senior marking up your work with reasons attached.

Notice that the last one never depended on production at all. It depended on somebody senior spending time saying why. That is the rep you can still manufacture, and it is the cheapest intervention available to you.

When the entry rung goes and nothing replaces it, teams tend to fail in one of two directions. Both are recognisable and both are your responsibility to catch:

• The hollow middle. Juniors produce a lot of plausible output quickly, get praised for throughput, and never build the ability to judge it. Two years in, they are fast and unable to lead a review. This is the more common failure because it looks like success for the first eighteen months.
• The frozen senior. Seniors absorb all the judgement work, become the only bottleneck on quality, and stop having time to teach - which guarantees no replacement is coming. This one is quieter, and it usually surfaces as a senior resignation rather than as a training problem.

Both are structural, so neither is fixed by encouragement. They are fixed by moving judgement tasks down the ladder on purpose and protecting time for the people who have to explain them.

Ask these of every piece, in this order. Each one is hard to answer without having done the thinking:

1. "What was the brief, and who set the constraints?" A candidate who cannot state the constraint they were designing against was decorating, not designing. Look for a primary action, an audience, a limit they had to respect.
2. "Show me the options you rejected, and tell me why." The rejected work is the actual portfolio now. Anyone can produce forty options; the interesting artefact is the reasoning that eliminated thirty-nine.
3. "What did you change after review, and what did you push back on?" This separates people who take direction from people who have a position. You want both behaviours, in the right places.
4. "What did you refuse to ship, and what happened?" Refusal is the clearest signal of a standard. It costs something to say no, so it is expensive to fake.

None of these ask whether the work is pretty. You can see that in three seconds without help. Spend the interview on what you cannot see.

Here is the exercise to run instead of a take-home. It takes fifteen minutes and it sorts candidates cleanly.

The tell you are listening for. Weak candidates reach immediately for taste words - it feels cluttered, it needs polish, it is not premium, it wants more breathing room. Those sentences are unfalsifiable, which is why they are so popular in reviews. Strong candidates name measurable failures first and only then talk about concept, tension and fit. Both halves matter, but the order tells you whether they can run a review without you in the room. Designer session b5 is the practitioner version of this and is worth putting in front of your whole team, not only new hires.

Junior owns conformance and hygiene. Critique reps as a scheduled activity, not an accident. Systems hygiene: is this component the real one, are these tokens the real ones, is the layer structure survivable. States and edge cases: empty, loading, error, long names, zero data, worst-case content. Accessibility conformance as a deliverable they sign off. All of it is measurable, which is what makes it teachable and reviewable at speed.

Mid owns direction and exploration strategy. Writing the brief before anything is generated. Deciding how wide to explore and when to stop. Choosing a direction and defending it to someone who did not attend the review. This is where generation is most useful and most dangerous: a mid with a real brief gets forty useful starting points, and a mid without one gets forty versions of the same average.

Senior owns concept, the standard and refusal. Is there an idea here or only an arrangement. What is the bar, written down, so other people can apply it without asking. And the part that cannot be delegated: saying no, in public, with reasons - because a standard nobody has seen enforced is not a standard.

Note what is missing from all three: nobody's ladder rung is "produces the most output". If throughput is still in your levelling criteria, generation has already made your ladder measure the wrong thing.

Skill comes from repetition with feedback. Production used to supply the repetition for free and seniors supplied the feedback as a side effect of shipping. Both supplies have shrunk, so you now have to schedule what used to happen by itself. Three mechanisms, in ascending cost:

• Critique drills. Thirty minutes a week. One screen, generated or shipped, and the team scores it against the stated standard before anyone offers an opinion. The measurable checks come first, always. This builds the muscle of naming failures specifically, which is the muscle the interview exercise tests for.
• Redlines. A junior marks up someone else's work against the standard, with reasons. Then a senior marks up the redline. Twice the teaching from one artefact, and the junior learns the standard from the enforcing side rather than the receiving side.
• Paired reviews. Junior sits in a real stakeholder review, prepares the argument beforehand, delivers one section of it, and gets debriefed after. Expensive in senior time, and the only thing that teaches the political half of the job.

Put all three in a calendar with a named owner. Reps that live in an intention rather than a calendar do not happen, and the failure is invisible for about a year.

Most junior design job descriptions are still lists of production verbs: produce, resize, prepare, deliver assets, support the team. Every one of those is now a sentence a tool can satisfy, which makes the description simultaneously unattractive and unhelpful for screening. Rewrite around what you actually need the person to become:

• Replace "produces assets and variants" with "owns the state coverage and accessibility conformance of every screen they touch".
• Replace "supports senior designers" with "runs the weekly critique drill and writes the redline".
• Replace "proficient in [tool list]" with "can name a design failure specifically and propose the brief change that prevents it".
• Keep one tool line, honestly, because tools are the entry ticket. Just stop making it the top of the list.

The screening benefit is immediate: candidates who only have renders cannot write a convincing cover letter against this description, and candidates who think in constraints finally have something to answer.