Claude Code Didn’t Just Make Developers Faster — It Has Changed Who Gets Paid
I did finance at university and I thought it would be fun to weave some economic theory into the wider AI discussion about developers.
One of the key things you learn in economics 101 is that.
Value Flows to Scarcity
Let’s use Magic cards to explain.
This card is worth about $19 at the time of writing this article.
But if you click here on TCGPlayer, you can see that $19 is just the beginning.
Wow, a card that is $5,393??? Why is that? Remember, these cards all do EXACTLY THE SAME THING IN THE GAME. There is literally no mechanical difference to them.
It’s exactly the same card, but it’s just Japanese and then it’s also a borderless treatment. Our Chocobo is now black too which is rare.
Basically the odds of getting one of these is way higher.
But $5,000 isn’t even the most that this card gets to.
For that we have this special card. The Golden Chocobo.
Zoom in on the bottom left of the card.
That XX/77 means that there is only 77 of these printed in the world.
These cards are going for north of $70,000 USD in auction. And this is the best example of economics in action.
So let’s remember that value flows to scarcity.
Made with the help of Gemini
So then what is the impact of AI on coders?
So AI has been released and that has clearly had an impact on coding and coders. So here is a funny question. If you were a Chocobo card, which one would you be?
Taken from the website Scryfall
Would you be the basic $19 card? Maybe you are the nice blue one or maybe you are that golden numbered to 77 one.
The reason this is a hard question to answer is that, we don’t really have a frame of reference for value in coding. In Magic it’s easy, we have print runs, but in coding, the only thing we really have is salary, but that doesn’t capture pure value intrinsically like a print run does. There is average engineers that are paid lots and then there are great engineers who aren’t paid as much as they should be.
Zooming out from you for a second, the first thing we have to do is get a grasp as to what exactly a programmer is and what traits they have.
Once we have established this, then we can see what AI has actually done to us poor Chocobos.
Converting you the programmer to a Trading card
To help transform you into a Chocobo I have researched some existing literature on coding and what makes developers valuable.
I found two really interesting theories.
The first is Anders Ericsson’s research on expert performance and the second is cognitive load theory developed by John Sweller.
While they were great, they were incredibly wordy. So I’m going to try and compress it down into four clear areas.
The four key areas of a programmer (generated by Gemini)
Let’s put some definitions to the 4 quadrants.
- Syntactic work. Remembering API structures, language quirks, implementation patterns. Basically, stuff you can Google or that you used to StackOverflow.
- Algorithmic reasoning. Manipulating logic, understanding complexity, optimizing within constraints.
- Problem decomposition. Taking some vague business requirement and translating it into something you can actually build. This is where you figure out what the hell you’re even supposed to make.
- Systems thinking. Understanding how all the pieces interact. Designing for things that won’t break when the next developer touches them in six months.
When you put it on a card, it looks like this.
I love how good Gemini is at making things
Question: Which of the above areas do you think coding has changed?
I think that Cursor/Claude Code has crushed the cognitive cost of the first two buckets while leaving the last two almost entirely untouched.
I find this interesting because it isn’t a uniform productivity boost across the board. It’s a selective elimination of certain kinds of work. As a result, based on our Chocobo example, this is going to have ramifications for value and who is going to get paid.
That’s because of the golden rule: “value flows to scarcity”
When something becomes scarce it becomes more valuable (Generated by Gemini)
So AI has come along and eliminated scarcity in something according to my theory, which is that structural and algorithmic work.
By definition this then makes the remaining scarce capabilities disproportionately more valuable.
Think about it like this. The bottleneck in software development has historically been implementation capacity. There were never enough skilled developers to build everything organisations wanted. That scarcity created premium wages for anyone who could write functional code.
So then what happens when Claude Code allows one developer to produce what previously required three? Well, the bottleneck to making software hasn’t completely disappeared, it has moved. (And while we are at it, I believe the bar will raise higher, as people expect more output per dollar spent on software, meaning that more complicated systems will get built, which will cause a new scarcity.)
This bottleneck has to move, that’s how economics works
We’ve had a disrupting event happen and now the bottleneck will shift upstream to problem definition. It will also migrate downstream to integration and maintenance.
This sets the stage for who I think the new scarce resources are, aka the new golden Chocobos.
I’m not really sure how this trading card game works but this card is rare
So who are they?
The people who can articulate what should be built, and the people who can ensure it actually works in production, become the new scarce resources.
Going to the card above, they are going to be good at problem decomposition and systems thinking. An easy nickname for them is going to be Jedis. I believe that money will start flowing to the Jedis.
But the good news, I think there is going to be different subclasses of Jedi that you can benefit from this money flow.
Jedi #1: The architects and strategic thinkers.
Those who can specify what should be built see their leverage increase. A skilled product thinker who previously couldn’t code can now build functional prototypes. A technical architect who spent 60% of time on implementation details can now spend 90% on system design.
I think I fall into this camp. Because coding is taking care of the syntactic work and the algorithm work for me, it allows my problem decomposition and systems thinking to shine.
I also think this is why my Medium articles frustrate some purist developers. I think they probably have higher syntactic skills than I do and so get frustrated at the things I say. It’s not me being antagonistic on purpose, it’s just different personality styles being expressed.
Jedi #2: The integrators and operators.
As Yoda once said.
To the Jedi who can integrate Salesforce to anything goes the spoils.
AI makes an avalanche of stuff. Code just flies out of the system. So now integrators are more valuable than ever.
That’s because you can write code quickly, but it still must be deployed, monitored, maintained, and secured. DevOps, SRE, platform engineering. These roles become more critical as code volume increases. The people who understand production systems, who can debug mysterious failures, who maintain the connective tissue between services. Their work cannot be easily automated because it requires real-time contextual judgement.
Also you can easily see this in practice if you have ever tried to merge a massive AI commit to production. You basically need the integrator to help you out.
Jedi #3: The domain experts.
I think this is sort of like another view of Jedi #1. Basically this person is able to articulate themselves in an easier way and that allows their systems thinking to shine because they aren’t bogged down by their lack of syntactic knowledge.
Software increasingly eats domains that require specialised knowledge. Healthcare, finance, legal, logistics. A nurse who can use Claude Code to build clinical workflow tools captures value that previously would have gone to a developer who needed months to understand the clinical context.
I believe that these 3 roles are now more scarce and so the money flows will start to shift towards them.
BUT BUT BUT — I can hear you saying
I have sort of trivialised development and overglossed engineering. So bear with me and read this next chunk.
Money will still flow to the 10x engineer. That’s because they almost exist on a super category of their own. In fact this is obvious when you read the literature on what makes a good developer actually good.
The tech industry has loved talking about “10x developers” for decades. Someone who produces an order of magnitude more value than average.
But I wanted to research what these traditional 10x developers were and there’s actually a lot of information deep diving on this. I think this research is even more relevant today given the rise of AI but it was also fascinating to me how much absolutely nothing has really changed in what makes a good programmer over the last 50 years.
I’ll do my best to cohesively explain it so here goes.
Large-scale studies at IBM and Bell Labs, along with management research synthesised in Peopleware by Tom DeMarco and Tim Lister, consistently observed 5×–28× performance differences between programmers given the same tasks, tools, and experience levels.
Factors that didn’t correlate to your success:
- typing speed
- memorised syntax
- raw IQ beyond a basic competence threshold
- years of experience alone.
Basically you hit this level of good enough and then the Lance Armstrongs of the coding world would power ahead of you. So if they weren’t typing faster or memorising things better than you, what were they doing?
These are the super Jedi.
They were better at decomposing problems, thinking about causality, good at tracking how a systems state would work in different scenarios and then they also had the ability to basically simulate things in their mind. E.G you ask for a new feature and they could imagine how it would affect everything before they even implemented it.
This conclusion aligns with earlier systems thinking captured in The Mythical Man-Month by Fred Brooks, which emphasised that software productivity is dominated by conceptual complexity rather than mechanical effort.
In short, the consistent scientific finding is that extreme performance in software comes primarily from how developers think about systems, not how fast they type within them.
The Metacognitive Premium
Perhaps the most consequential shift is the increasing premium on metacognition. The ability to think about your own thinking.
A developer with strong metacognitive skills knows when to trust Claude Code’s output and when to verify. They know when their prompt was ambiguous. They know when they’re operating at the edge of their competence and need to slow down.
The Dunning-Kruger effect becomes particularly relevant here. Less skilled developers may overestimate their ability to evaluate AI-generated code, confidently accepting solutions they don’t fully understand.
This creates a quality problem that the market will eventually price in. But not before significant damage accumulates.
What This Actually Means
The fundamental insight is this: I believe that Claude Code and their successors are redistribution mechanisms.
They transfer economic value from those who could execute* to those who can conceive. From those who could implement to those who can integrate. (* excluding Super Jedi)
Now it’s not about the things that you memorised or the crazy niche error that you got, it’s about problem solving, systems thinking and problem solving. But the thing is, it’s not really today is it, it’s something that has always separated the best from the rest. It’s just that people like myself now have a seat at the table because we aren’t gatekept from the syntactic and algorithmic complications that meant we could never even articulate ourselves in the first place.
Written by Chris Dunlop
I’ll help you code with AI. Cursor tips & Business Strategy. I run a company doing AI for the All Blacks, Olympic Team & the Stock Exchange www.cubdigital.co.nz