What happens when everyone can create everything?

Let's take a look at software:

20 years ago → The Physical Era

The barrier was capital. You needed $250k for racking, servers, and infrastructure just to validate a proof of concept. The moat was your (or an investor’s) bank account.

10 years ago → The Cloud Era

AWS collapsed the infrastructure cost, and the barrier shifted to talent. You needed $50k and an engineering team or a devshop. The moat was your technical capability.

Today → The Generative Era

The cost of initial creation has collapsed to the price of a subscription. With AI-assisted IDEs, a single builder can deploy a highly functional POC in a weekend for a just a few hundred dollars and some cloud credits.

The economics of product innovation have fundamentally inverted.

The immediate impact: Velocity

This collapse allows for unprecedented speed. We have moved from "permissioned innovation" (eg. needing investors to start) to "permissionless abundance."

Having lived through these eras building product, the shift in the last 12 months is incredible.

• Then: Building a robust iOS backend took several cycles of dedicated engineering time (and possibly significantly more depending on complexity).

• Now: In the past few weeks, I recreated that same architecture in a matter of days. On another project, I stood up a PWA (Progressive Web App) with React in a single week, getting it into the hands of my initial test users and iterating live.

To be clear: this does not mean high-quality, senior engineering is obsolete. Absolutely not. What has changed is the order of operations. Previously, you needed experienced engineers just to enter the game. Now, you can rapidly move past the boilerplate required to validate the idea. Senior engineering is still required for architecture, security, and scaling - but you can incur significant portions of that cost after you know you’re building the right thing.

The distribution paradox

'Easier to build' means 'harder to be noticed'. When the barrier to entry hits zero, the market is flooded with 'average' software. The challenge has shifted entirely from creation to distribution.

While we are focused on product development, the same goes for other similar areas - like content and media where the marginal cost of creation is effectively zero. (Note: I didn't say good content). When everyone can create, what separates winners from the noise?

Taste is differentiation

In a world of infinite supply, taste becomes the mechanism to power effective distribution.

By definition, AI models regress to the mean. They output the average of the internet (and significant parts of the physical world too). Taste is the intersection of deep user empathy and an opinionated execution. It's not aesthetics - it's decision-making. Here are two attributes needed to break through the noise:

• Specificity: Solving a user need so granuarly that the product market fit is the marketing. Users feel so understood that not using this isn’t a consideration.

• Application: Refusing the 'average' default of the model to inject a distinct point of view. This applies to the workflow, the design, the brand, the marketing.

Together, the deep user empathy and a opinionated viewpoint creates brandable differentiation. Stripe was built by developers who knew the pain of payments. Linear was built by designers and engineers who knew the pain of project management. Figma was built by engineers and graphics experts who knew design required collaboration.

Their taste emerged from lived experience. They knew which defaults to reject because they'd been frustrated by them. They knew which features to prioritize because they understood the job-to-be-done intuitively, not only through user research.

Domain expertise is the input to taste. It's what makes your deliberate choices credible rather than arbitrary. Taste without domain expertise is just opinion. Taste with domain expertise becomes conviction backed by pattern recognition.

Horizontal platforms vs niches

'Specificity' does not mean 'small market'. It refers to the precision of the fit, not the scope of the problem. A horizontal platform can be specific if it solves a workflow problem other tools ignore. Specificity means users evangelize because they feel seen. Horizontal platforms also have significant power of workflow and data proximity, as well as distribution.

Again, taste and opinionated design still reign supreme. Validated examples include project management (Linear), and design (Figma). Both are quite vocal of their worldview [1] - in fact, Figma has a great overview where they interview the Linear team. [2]

AI-amplified taste

AI democratizes execution, which makes curation more valuable. A builder with strong taste can now explore the solution space faster - generating different variations, testing workflows at speed, tightening the iteration cycle until the fit is precise.

But AI outputs are only as good as the taste that guides them. In a world of abundant supply, curation - the ability to identify and refine what's exceptional - becomes the foundation for distribution. Curation creates the signal that distribution amplifies. Without differentiation, you're just amplifying noise.

How taste drives distribution

Differentiation compounds through three mechanisms:

• Word-of-mouth efficiency: Products with taste are remarkable - worth talking about. Users become evangelists because the product reflects their identity.

• Signal in the noise: In a sea of sameness, a distinct point of view acts as a filter. The right users self-select in; the wrong users self-select out.

• Premium positioning: Taste enables pricing power, which funds go-to-market. Commoditized products compete on price; differentiated products compete on fit.

Taste doesn't eliminate the need for distribution - it makes distribution possible. You can't growth-hack your way out of being unremarkable or undesirable.

In short

In the Physical Era, we paid for metal. In the Cloud Era, we paid for code. In the Generative Era, taste becomes the currency of success.

Differentiation is no longer about how you build; it is about who you build for, and why they care. Taste is the thousands of micro-decisions that prove you understand.

What else?

1. Evolution of Average: AI is a snapshot of the past. As soon as AI learns to replicate the current "Good Taste," that style becomes ubiquitous and, by definition, "Average.” Understanding why choices resonate - not just what looks good today - is necessary.

2. The supply paradox: As AI makes building easier, the variance in product quality and product options should increase. But instead, we're seeing convergence toward AI-generated mediocrity. Again, reinforcing that it isn't about outsourcing decisions to the model, but rather just using it as a tool to get more done.

3. Evals to systematize taste: In a world where AI generates infinite variations, the bottleneck becomes knowing what 'good' looks like. Evals - formalized testing protocols that programmatically measure outcomes - are how you operationalize taste. Can taste become a reproducible system that maintains differentiation even as more people touch the product? How does that evolve over time as taste continues to be refined?

4. The talent inversion: As code becomes abundant, the bottleneck shifts from implementation to differentiation. The ‘soft skills’ of tech become a hard currency: product, marketing, and UX have a huge role to play in the differentiation, distribution, and adoption of the products of tomorrow.

5. In-housing of SaaS: Only a few years ago, the default was Buy. Even for niche internal tools, building was too expensive, so we paid for "good enough" SaaS subscriptions. In 2025, the default for utilities shifts to Build. For companies: Are you building in the commodity layer (CRUD utilities typically built on client’s proprietary data), or the premium layer (where reliability, compliance, integration are key). Again, taste will be the key differentiator to building and positioning product.