The Era of Ideas: What Comes After Softflation

When building is free but thinking is scarce, the most valuable thing in the world is knowing what to build. I have to be honest: I wasn’t ready for that.

In my last article, I introduced the term Softflation — the systemic devaluation of software caused by AI agents’ ability to generate code at near-zero cost. The response was overwhelming. Hundreds of developers, founders, and product people reached out. But one question came up more than any other, always phrased a little differently but always pointing at the same thing: “Okay, so if software loses its value — what gains value instead?”

I have to be honest: this question kept me up at night. Not because I didn’t have an answer, but because the answer is uncomfortable. It challenges something many of us — myself included — have built our identities around.

The answer is ideas.

Not ideas in the Silicon Valley sense — not pitch decks and napkin sketches and “wouldn’t it be cool if.” I mean something much more specific: the ability to see a problem clearly, to understand what a solution should feel like, and to articulate that vision precisely enough that an AI agent can build it. That’s the new scarce resource. And it changes everything.

The Great Inversion

For fifty years, the software industry operated on a simple hierarchy. At the top sat the idea — often dismissed as the easy part. “Ideas are cheap,” we told each other at meetups. “Execution is everything.” Below the idea came architecture, design, implementation, testing, deployment. Each layer required specialised skills. Each layer cost real money and real time.

Softflation has inverted this hierarchy completely.

BEFORE SOFTFLATION Idea: easy, cheap, everyone has them Architecture: requires senior talent Implementation: expensive, time-consuming Testing & QA: expensive, tedious Deployment: complex, specialised

AFTER SOFTFLATION Idea: scarce, decisive, the real differentiator Architecture: increasingly AI-assisted Implementation: near-zero cost Testing & QA: increasingly automated Deployment: one-click, commoditised

Execution was everything — when execution was hard. Now that AI agents handle implementation at speed and scale that no human team can match, the bottleneck has shifted upstream. The question is no longer “Can we build this?” It’s “Should we build this? And what exactly should it be?”

That’s a fundamentally different skill. And most of us — if we’re being honest — are not very good at it yet.

Why Ideas Were Always Undervalued

There’s a reason the tech industry spent decades dismissing ideas. We had to. The implementation bottleneck was so severe that ideas without execution were genuinely worthless. A brilliant app concept without a team to build it was just a daydream. So we developed a culture that worshipped builders and quietly dismissed thinkers.

This made sense. In a world of scarcity, you value the scarce thing — and that was engineering talent.

But something funny happens when you remove a constraint. You don’t just make the old system faster. You reveal a completely different system underneath. And the system underneath the implementation bottleneck turns out to be this: most software fails not because it was poorly built, but because it solved the wrong problem, or solved the right problem in a way nobody wanted.

“We used to blame failed products on bad engineering. It turns out, most of them had perfectly good engineering. They just had bad ideas — ideas that nobody examined closely because everyone was too busy writing code.”

I see this in my own work, too. Looking back at projects I’ve abandoned, the code was usually fine. What was broken was the thinking. The assumption about who would use it. The belief that a certain pain point was sharp enough for someone to switch tools. These are failures of ideas, not failures of implementation. I just didn’t notice because the implementation was so consuming that I never had time to question the premise.

The Idea Supply Chain

If ideas are the new scarce resource, we need to understand what makes a good one. Not in the abstract motivational-poster sense. In the concrete, operational sense. What does “a good idea” actually look like in the Era of Ideas?

I have to be honest — I struggled with this for weeks. But I think it comes down to three components that I started calling the Idea Supply Chain.

Framework

The Idea Supply Chain

A good idea in the post-Softflation world requires three things: Problem Clarity (understanding the real pain, not the symptom), Solution Taste (knowing what the right solution feels like before it exists), and Articulation Precision (describing it clearly enough that an AI agent — or a team — can build exactly what you mean).

Problem Clarity

This is harder than it sounds. Most people describe problems at the symptom level. “Our team communication is broken.” “Our onboarding takes too long.” “I can’t find what I need.” These are symptoms. The actual problem lives underneath — in misaligned incentives, in missing context, in workflows that evolved through accident rather than design.

AI agents are excellent at building solutions. They are terrible at diagnosing problems. That diagnosis — the moment where you see through the symptom to the structural issue — is a deeply human skill. It requires empathy, domain knowledge, and the kind of pattern recognition that comes from years of paying attention. No prompt can replace it.

Solution Taste

This is the hardest to explain, maybe because in German we would call it Fingerspitzengefühl — a sensitivity in the fingertips, an intuitive feel for what’s right. There is no English word that captures it quite as well.

Solution Taste is the ability to know what a good solution feels like before you’ve built it. It’s the instinct that tells you a three-step flow is better than a five-step flow. That this feature should be automatic, not configurable. That the user doesn’t need a dashboard — they need a notification.

This is the skill that the best product designers have always had. In the Era of Ideas, it becomes the most valuable professional skill in technology. Not coding. Not project management. Taste.

A story

A friend of mine — a product designer in Berlin — told me about a project where her team used AI to generate twelve different prototypes for a settings page. Twelve fully functional versions, each with different information architecture, different interaction patterns, different visual hierarchy. The AI produced them in a single afternoon.

Her job was to look at all twelve and say: “This one. And change the second section.”

That choice — that act of taste — took her six minutes. But it was the most valuable six minutes in the entire project. Without it, the team would have shipped something mediocre. With it, they shipped something users actually loved.

Six minutes. But six minutes backed by fifteen years of caring about how software feels.

Articulation Precision

The third link in the chain is the one that trips up most people. You can have perfect Problem Clarity and exquisite Solution Taste, and still fail because you can’t describe what you want precisely enough.

In the old world, this gap was filled by engineers who could read between the lines of a vague specification. They would ask clarifying questions, make assumptions, iterate. In the Era of Ideas, your specification goes to an AI agent — and AI agents are simultaneously more capable and more literal than any human engineer. They’ll build exactly what you describe. Which is a problem if what you describe isn’t quite what you mean.

This is why “prompt engineering” evolved into “context engineering” and is now becoming something closer to intent architecture — the discipline of expressing what you want with enough precision and context that an AI system can make good decisions about the ambiguous parts.

The Death of “I Just Need a Developer”

For years, I heard the same sentence from non-technical founders. Everyone has heard it. You’re at a dinner, someone finds out you work in tech, and within two minutes: “I have this amazing idea, I just need a developer.”

We used to roll our eyes at this. The gap between “idea” and “working product” was so vast that the sentence was almost comical. It was like saying, “I have a great idea for a building, I just need an architect, a structural engineer, a construction crew, and twelve months.”

Here’s the uncomfortable part: those people were right. They were just early.

In 2026, “I have an idea and I just need a developer” has become “I have an idea and I need an AI agent.” And the AI agent is available, right now, for the cost of a subscription. The dinner-party dreamers of 2019 were, in a strange way, seeing the future more clearly than we were. They just didn’t have the tools yet.

But — and this is crucial — the people who actually succeed with those tools are not the dinner-party dreamers. They’re the ones with Problem Clarity, Solution Taste, and Articulation Precision. The idea was never the naive part. The quality of the idea was always what mattered. And now, finally, idea quality is the only thing that matters.

“The dinner-party dreamers of 2019 were accidentally right. ‘I just need a developer’ became literally true. What they didn’t understand is that the developer was never the hard part. The hard part was having an idea worth building.”

What This Means for Careers

I have to be honest: the career implications are significant, and not everyone will like hearing them.

If your professional identity is built on the ability to write code — specifically, on the scarcity of that ability — the ground is shifting beneath you. This doesn’t mean developers become irrelevant. Far from it. But it does mean the type of developer who thrives is changing.

The developers who will flourish in the Era of Ideas are the ones who were always a little frustrated by pure implementation. The ones who cared about why they were building something, not just how. The ones who pushed back on specifications they thought were wrong. The ones who said, “Before I build this, can we talk about whether we should?”

Those developers were sometimes seen as difficult. Slow. Not “shipping-focused” enough. In the Era of Ideas, they’re the most valuable people in the room.

For non-developers, the opportunity is enormous. Domain experts — doctors, teachers, lawyers, logistics managers, anyone who deeply understands a problem space — suddenly have direct access to building tools for their own domains. The nurse who knows exactly how patient handoff documentation should work doesn’t need to convince a product team anymore. She can describe it to an AI agent and have a working prototype by Friday.

But here’s the catch: that same nurse needs to learn Articulation Precision. She needs to develop the skill of translating her domain expertise into something an AI agent can act on. That’s a new literacy — as fundamental, eventually, as email or spreadsheets.

The New Competitive Landscape

In the Softflation article, I described how I released an open source tool only to discover multiple identical projects had appeared in the same window. That experience taught me something about the Era of Ideas that took a while to fully articulate.

When building is free, being first to build stops being an advantage. By the time you’ve published, three other people have published something similar. The code is roughly equivalent. The features overlap. The READMEs even look alike (because they were probably all AI-generated too).

So what is the advantage?

I think it’s what I’d call idea depth. Let me explain with an example. Imagine ten people all decide to build a project management tool for small creative agencies. All ten use AI agents. All ten produce functional applications within a week. On the surface, they’re interchangeable.

But one of the ten builders spent three years working at a creative agency. She knows that the real problem isn’t task tracking — it’s scope creep. She knows that creative teams don’t think in sprints, they think in moods. She knows that the most important feature isn’t the Kanban board, it’s the ability to flag when a project has silently shifted from what was agreed to something nobody budgeted for.

That builder’s tool will look different. Not because the code is better — AI agents wrote all ten codebases. But because the idea was deeper. The problem was understood at a level the other nine builders never reached.

The new moat

Idea Depth

In the Era of Ideas, competitive advantage comes not from building faster or cheaper, but from understanding a problem more deeply than anyone else. The depth of your insight determines the quality of your product — and depth cannot be prompted into existence.

What We Lose (And What We Must Protect)

I don’t want to paint this as a purely optimistic story. That would not be honest, and I think we have enough optimistic AI narratives already.

The Era of Ideas has real losses. The craft of programming — the deep satisfaction of solving an elegant problem with elegant code — is being hollowed out. Not eliminated, but reduced to a niche. Like letterpress printing or hand-stitched bookbinding: still valued by some, no longer the primary means of production.

I feel this loss personally. I started coding as a teenager, and there was a particular joy in the act of building — the logic, the debugging, the moment when something finally worked. That joy doesn’t disappear in the Era of Ideas, but it becomes optional. You can still code by hand, the way you can still develop film photographs. But the world has moved on.

What worries me more is the loss of deep technical understanding. When you write code by hand, you develop an intuition for how systems work, where they break, what’s expensive and what’s cheap. Developers who grew up in the implementation era carry this intuition like muscle memory. Developers who grow up in the Era of Ideas may never develop it — and that could lead to a generation of builders who can describe what they want but can’t evaluate whether what they got is any good.

We have a word for this in German: Halbwissen. Half-knowledge. Knowing enough to be dangerous, not enough to be safe. I think the biggest risk of the Era of Ideas is an epidemic of Halbwissen — millions of people building software they don’t truly understand, guided by AI agents they can’t truly evaluate.

“The biggest risk of the Era of Ideas is not that we build too much software. It’s that we build it without understanding — and we lose the ability to tell the difference between software that works and software that merely appears to work.”

Five Principles for the Era of Ideas

I’ve been thinking about how to navigate this new landscape — not just theoretically, but practically, in my own work. Here’s where I’ve landed, at least for now.

Go deep, not wide. In a world of infinite breadth — where anyone can build anything — depth is the differentiator. Understand one problem better than anyone else. Live inside it. Talk to the people who have it. That depth becomes your unfair advantage, and no AI agent can replicate it.

Develop your taste. Taste — Fingerspitzengefühl — is a skill, not a talent. You develop it by paying attention. By using lots of software and noticing what feels right and what doesn’t. By studying design decisions in products you admire. Taste is just pattern recognition refined through exposure and reflection.

Learn to articulate precisely. Practice describing what you want as if you were briefing a brilliant but very literal junior engineer. This is the new core skill — and like any skill, it improves with practice. Write specifications. Describe your vision in words before you describe it in prompts.

Stay technical enough to evaluate. You don’t need to write code. But you need to understand enough about how software works to know when an AI agent has given you something good and when it’s given you something that will break at scale. This is the antidote to Halbwissen.

Build community, not just software. My open source experience taught me this. The code is the least defensible part of any project. What’s defensible is the community of people who trust your judgment, rely on your documentation, and believe in your vision. Invest there.

The Bottom Line

The Era of Ideas is not a future state — it’s where we are right now. The builders who thrive will be the ones who understand that the most valuable code is the code you choose not to write, the most valuable feature is the one you decide not to build, and the most valuable skill is knowing the difference. Software is free now. Thinking never will be.

Next

[embed]After the Era of Ideas: Five Roads Into a Post-Scarcity World What happens when AI doesn’t just build your software — but also tells you what software to build? I spent a long time…medium.com