Mike Litman
What a Creative Builder Actually Does
Builder · Writer · Communicator · Product Thinker

Profile 01

Anthropic’s engineering leader named the most undervalued hire in AI. Here is what that looks like when it ships.

Profile 01
Creative builders with product sense

Can spot the right thing to build and prototype it fast. Taste is scarce, typing is not.

Fiona Fung, Head of Engineering, Claude Code / Cowork, Anthropic:
“I don’t care how many lines you can write per hour. I care what you choose to build.”

Profile 02
Deep systems experts for the hard parts

The places where “trust but verify” matters most. Subtly wrong is still wrong.

Infrastructure, security, scale – the parts where implementation precision is the bottleneck. Both profiles are essential. The ratio is what most teams get wrong.

03
November 2025

The person who built Claude Code stopped writing code by hand.

Boris Cherny – creator and head of Claude Code at Anthropic – has not written a single line of code by hand since November 2025. Not as an experiment. Permanently. The shift is not coming. It has already happened. The question is what it means for who you hire next.

Boris Cherny Head of Claude Code, Anthropic Lenny's Newsletter, 2026
04
The bottleneck that moved

Implementation. The code. The typing.

"Coding is largely solved," Cherny said. Not as a boast. As an observation. Claude Code now writes 90–95% of itself. The tool that is changing how the world builds software is mostly built by the software itself.

This is not the end of engineering. It is the end of implementation as the scarce thing. The scarce thing was always something else.

90–95% of Claude Code written by Claude Lenny Rachitsky / Anthropic CPO, 2026
AI handles this now
Writing the code
Debugging the logic
Generating the boilerplate
Refactoring at scale

The bottleneck has moved. Most hiring managers have not moved with it.

AI cannot do this
Knowing what to build
Deciding what is worth building
Recognising when something feels wrong
Caring whether users come back

If implementation is abundant and taste is still scarce, the most valuable hire is the person with taste.

Fiona Fung – Anthropic

Taste is scarce. Typing is not.

This is the engineering leader at the company building the world's most powerful coding tool. She is not worried about whether candidates can write code. She is worried about whether they can tell you what to build.

07
The failure mode

Technically excellent. Nobody asked for it.

The Profile 2 trap is not about bad engineers. It is about teams that can build anything but have not decided what is worth building. You get speed and precision aimed at the wrong target. Elegant architecture serving a product nobody wants. Infrastructure running faster than the product has users.

More AI tools means this trap gets sprung faster. The teams that are purely Profile 2 will ship more, more quickly, in the wrong direction.

Fifteen years in digital agencies taught me this pattern: the brief was always human, the execution was always technical, and the gap between them was always the product person's job to close. When that person was not in the room, the work was usually brilliant and usually wrong.

To be clear: Profile 2 is still essential. The hard parts – infrastructure, security, scale – still need deep systems expertise. This is an argument about the ratio, not a case against systems engineers.

08
The empathy beat

The old filters never measured taste.

LeetCode. GitHub commit counts. CS degrees. These were reasonable proxies when implementation was the bottleneck. The company that could ship fastest won. So you hired for the ability to ship.

They are not broken. They are just measuring the wrong thing now. And because most hiring managers learned to hire using these filters, they keep using them. The job description has not changed because the mental model has not changed.

System lag The filter, not the candidates, is out of date
The reframe

Profile 1 is not “non-technical”. Profile 1 ships.

Creative builders are not people who have ideas and hand them to someone else. They prototype fast, test with real users, and iterate. They use AI as leverage, not as cover. The differentiator is not the creativity. It is the delivery. "Creative" without the "builder" is just a job title with no output attached to it.

Buggy Smart started as a prototype call script. The voice agent has been rebuilt, rerouted, and re-prompted across dozens of iterations. 1,180 venues are on the map because someone kept deciding what "right" looked like and adjusting until it did. That is the job.

Concrete behaviours

What Profile 1 actually does.

01
Talks to users first.
Distinguishes a real problem from an interesting one before any code is written.
02
Decides without certainty.
Moves, tests, adjusts. Call volume cut 175 to 75 to 50 – each a judgement call, not a metric.
03
Ships end-to-end.
Product, design, deployment. Alone if necessary. Every project on the next slide: one person.
What I’ve been building in 2026

I am one. Here is what 2026 looks like so far.

Different categories. Same method: spot the right problem, build the fastest version, ship it, find out if it works.

Buggy Smart
1,180 venues mapped
First Order London
Voice agent for restaurants
With Moshi
AI phone answering
Queue Index
London queue data
The System
125+ registrations
EVERYWEAR
Fashion intelligence
The Visible Shelf
295 books
Oishii London
84 restaurants
Trove
Cultural intelligence
Taste Brief Course
Live + taking signups
Litman System
Design system v0.1
Modern Retro
96 AI brand films
Health Dashboard
Personal health tracking
CultureTerminal
Daily culture intelligence
First Out
London tube exit finder
Little London
Kids activities guide

No engineering team. No agency. Built with Claude Code, AI tools, and product instinct.

The evidence

“Everyone codes. Our product manager codes, our engineering manager codes, our designer codes.”

Boris Cherny – Head of Claude Code, Anthropic

“I don’t care how many lines you can write per hour. I care what you choose to build and how you know it’s right.”

Fiona Fung – Head of Engineering, Claude Code / Cowork, Anthropic

The company building the most powerful coding tool in the world is structured around product sense, not raw output. That is not coincidence. That is a design decision.

13
The imbalance

Most teams are over-indexed on Profile 2. AI is making it worse.

One person with product sense shipped 20+ products in less than a year. Now imagine what happens when a team gets the ratio right – Profile 1 judgement amplified by Profile 2 depth. That is what the winning teams are building towards.

The standard hiring instinct when AI arrives: hire more engineers, the people who understand the tools. But when implementation is no longer the bottleneck, more implementation capacity does not solve the problem. It accelerates the wrong direction.

The teams winning right now are not the ones with the most AI engineers. They are the ones where someone in the room has the taste to know what is worth building, and the authority to stop what is not.

Profile 1 is not a permanent replacement. The sequence matters: discover with Profile 1, build with both, scale with Profile 2. Most teams skip straight to scale without doing the discovery. That is why they build the wrong thing quickly.

The pattern

The companies that win hired Profile 1 before they had a name for what they were hiring.

The pattern is consistent: tiny headcount, clear product point of view, someone at the centre who can build and judge simultaneously. Anthropic structures its entire Claude Code team this way. "Everyone codes – our PM, our EM, our designer." Not because they have to. Because they hired people for whom that is natural. You do not need to call it Profile 1. You need to hire for taste, then build the systems around it.

What the old CV says
Senior Software Engineer, 7 years
Java, Python, Kubernetes
Led team of 8
Computer Science, First Class

The CV looks different because the job is different. If your screening process cannot see it, you will keep hiring the wrong profile.

What Profile 1 looks like
20+ live products in less than 4 months
Users came back. Some became businesses.
Shipped alone, end-to-end, every time
No degree in this. No category for this.
16
The naming problem

The job has arrived before the job title.

Builder. Creative Technologist. Product Engineer. Technical Product Lead. None of these quite fit. The role exists. The language does not.

Boris Cherny put it clearly: "The title software engineer is going to start to go away and be replaced by 'builder'." Most JDs are still written for Profile 2 using Profile 2 language. You cannot find the person you need if the job ad describes someone else.

Boris Cherny Lenny's Newsletter, 2026
What to look for

How you’ll know Profile 1 when you meet them.

01
The JD won’t find them.
It is written for Profile 2. Look for what they built, who used it, and whether they shipped without being asked.
02
Ask what surprised them.
“Tell me about the last thing you shipped that surprised you.” Profile 1 has an immediate answer. Profile 2 tells you something that worked.
03
Look at live work.
Not listed credentials. What exists in the world because of them? If the answer is nothing, keep looking.
My verdict

“The creativity and ideas from Profile 1 with the systems integration and rigour from Profile 2.”

The scarcest resource in AI is not the code. It is the judgement about what the code should do – and the rigour to make sure it does it right.

I have been building that judgement since January. 20+ products in 2026. No engineering team. If you are looking for this, I would like to talk.

mikelitman.me · hello@mikelitman.me

Continue from slide ?