There are two kinds of people working in AI right now. Fiona Fung, Head of Engineering at Claude Code (Anthropic), put a name to both of them. She called them Profile 1 and Profile 2, and the distinction is sharper than most hiring frameworks manage to be.
Profile 2 is the one everyone knows how to hire: deep systems expertise, infrastructure thinking, the engineer who can build anything at scale. Profile 1 is harder to describe, which is partly why it is undervalued. Creative builders with product sense. People who can spot the right problem, build the fastest version, and ship it before the organisation has agreed on the brief.
Most companies hire Profile 2 almost exclusively. They have processes for it. Job descriptions, technical interviews, take-home tests. The hiring pipeline for deep systems expertise is well-worn. Profile 1 does not have a job description yet. The title does not exist. Which means the people doing the work are doing it without the formal recognition, and the companies that need them most do not quite know how to find them.
What changed
The shift that makes Profile 1 matter is not new technology. It is a new ratio. For decades, the bottleneck in making things was execution: writing the code, building the infrastructure, implementing the design. The people who could do that were the constraint. You hired them, protected their time, organised everything else around them.
Boris Cherny, Head of Claude Code at Anthropic, has not written code by hand since November 2025. Ninety to ninety-five percent of Claude Code itself is written by Claude. The implementation bottleneck is dissolving. What remains is the harder problem: knowing what to build, knowing when it is right, knowing when to ship it and when to throw it away. Those are not engineering problems. They are judgement problems. And they are precisely what Profile 1 people are trained for, whether they know it or not.
The strategist who spent fifteen years working out what audiences actually want. The editor who can feel when something is not working. The curator who has a model of quality that does not require a rubric. The product thinker who has shipped enough things to know what done looks like. These people now have access to execution capability that previously required a team. The question is whether organisations are set up to use them.
What Profile 1 actually does
The work is not glamorous, and it is not speculative. It is: identify the problem, build the first version, put it in front of a real person, and find out if it works. Then do it again, faster, with what you learned.
Since September 2025, I have shipped more than twenty AI-native products. Voice agents, curation tools, data pipelines, interactive decks, restaurant directories, a health dashboard, a visible bookshelf. Different categories. Same method: spot the right problem, build the fastest version, ship it, find out if it works. None of them required deep systems expertise to start. All of them generated real signal. Several of them are now businesses.
This is the thing Profile 1 does that Profile 2 rarely gets to do: discovery. Finding out whether the idea is right before the engineering investment is sunk. Moving fast enough that failure is cheap and learning is free. Building with enough craft that real people actually use it. The best organisations do this at the start of every project. Most organisations skip it entirely and go straight to the scale phase. That is why they build the wrong thing quickly.
The ratio problem
Profile 2 is still essential. Infrastructure does not build itself. Security, scale and systems reliability all require deep expertise. Nobody sensible is arguing otherwise.
The argument is about the ratio. Most teams are over-indexed on Profile 2 and under-indexed on Profile 1. They have plenty of engineers who can build anything. They have very few people who can tell them what to build, test whether it is right, and ship something good enough to find out. The skill that is hardest to find right now is not technical. It is judgement.
The sequence that most teams skip
Here is the part that hiring managers miss: these two profiles are not alternatives to each other. They are a sequence.
Discover with Profile 1. Build with both. Scale with Profile 2.
Profile 1 is not a permanent replacement. The point is to do the discovery before the engineering. To find the right thing before building it at scale. Most teams skip straight to scale without doing the discovery. That is why they end up with technically excellent products that nobody asked for.
Fiona Fung made this case from inside Anthropic, which is the most technically capable AI organisation on the planet. She is not arguing against engineers. She is arguing for the discovery layer that makes engineering investment worthwhile.
How you know one when you see them
Profile 1 people do not wait for a brief to be perfect before starting. They build something provisional to understand the problem better. They think in feedback loops, not deliverables. They can describe what something feels like before they can describe what it is. They have shipped enough things to know the difference between "this needs more polish" and "this is the wrong idea." They treat the first version as a hypothesis, not a commitment.
The job title that matches this does not exist yet. Head of Building. Chief Discovery Officer. Director of Figuring Out What We Should Actually Be Making. What it is called matters less than whether you have it in the building.
If you are looking for this profile, I would like to talk.
