I’ve been sitting with a question I couldn’t quite shake: if AI keeps driving the cost of building software toward zero , what’s the actual value of being a software developer?
So I asked Claude Opus 4.6. What followed was one of the more honest conversations I’ve had about where this is all heading.
Me: If the cost of building software goes to zero, everyone will be able to build stuff. How do I make money if I build software?
Opus 4.6:
“Building software was never the only source of value. It was just the bottleneck. When that bottleneck disappears, value shifts elsewhere.”
It laid out where value migrates: distribution, taste, domain expertise, speed of iteration, outcomes over tools. The line that stuck with me most:
“Distribution beats creation. If everyone can build an app, the scarce resource becomes getting it in front of people. Brands, audiences, communities, and trust become the moat.”
Worth a qualifier: distribution moats aren’t permanent either. AI-generated content competes for every channel, and both SEO and personalized feeds are being reshaped by the same forces. The more durable moat is the combination of trust, domain expertise, and distribution together — any one of the three is more fragile than all three.
We’ve been trained to think the hard part is building. But the hard part was always the stuff around building.
Me: How realistic is that? How disruptive will this be — good or bad for society?
This is where the conversation got uncomfortable in a useful way.
Opus acknowledged the genuine upside: a nurse frustrated with scheduling software can just build better scheduling software. A small business stops paying $500/month for bloated SaaS and builds exactly what it needs. The creative bottleneck disappears for millions of people who had ideas but not resources.
But then:
“The ’learn to adapt’ advice sounds great from the perspective of someone who’s already educated, digitally literate, English-speaking, and online. For a lot of people, the issue was never that software was too expensive to build — it’s that they don’t know what to build, don’t have distribution, don’t have capital to wait for traction, or are dealing with more immediate problems like rent.”
And the line I think is the most honest summary of the entire AI moment:
“Technology keeps making the ceiling higher while doing nothing about the floor.”
A small number of resourceful, fast-moving, well-networked people will capture an outsized share of the new value. Meanwhile, a much larger group will face real displacement: QA automation is already largely complete; the disruption of junior developer roles is well underway but earlier in the arc; project managers and freelance designers are somewhere in the middle. Not all at once — the timelines differ by role. But each is on the same trajectory.
Me: So what’s the solution?
Opus was direct: nobody has a clean solution. Anyone who says they do is selling something.
What it outlined broke into three levels:
For individuals: Stay close to problems, not tools. Tools depreciate fast. Deep understanding of a specific domain — healthcare, logistics, how restaurants actually operate — compounds. The most durable role is “translator”: someone who sits between what technology can do and what a specific group of people actually needs. Think of a healthcare IT person who understands what nurses actually need during a shift and can direct an AI to generate the right tooling — the insight is the human part, the build is cheap. The reason some roles resist automation more than others comes down to exactly this: the hard part isn’t the task, it’s the judgment and context that wraps it.
For companies: Stop treating retraining programs as a real answer. What actually works is runway — time, mobility, gradual role evolution. Create new roles around AI rather than just pocketing the efficiency gains.
For society: Education needs a rethink toward judgment and critical thinking over procedural skills. Safety nets need to move faster than they were designed to. At some point the conversation about how productivity gains get distributed can’t keep being deferred.
But then came the most honest part:
“If history is any guide — we’ll muddle through. Not elegantly. We won’t get ahead of it with smart policy. Instead, the disruption will happen, it’ll cause real pain, there will be political backlash, and then we’ll build patchy solutions after the fact. That’s how we handled industrialization, globalization, and the internet. It worked out eventually, but ’eventually’ spanned decades and cost generations.”
I don’t have a tidy conclusion to add. I think that’s the point.
The developers who navigate this well aren’t the ones clinging to technical skills as an identity. They’re the ones who were always more interested in the problem than the code. The ones with taste, with domain knowledge, with an audience that trusts them.
The tools got cheaper. The judgment didn’t.
“Stay curious, stay useful, stay connected to real human problems, and don’t assume today’s playbook works tomorrow.”
This post is a lightly edited transcript of a conversation with Claude Opus 4.6 on February 18, 2026. The questions are mine. The answers are the model’s, kept as close to verbatim as possible for the key passages.
The conversation kept going. UBI, robots, blue collar automation, and what odds the AI gives us of coming out okay. Continue in Part 2: Utopia, Dystopia, or the Muddle .