9horsemenStart building
Essays

AI Makes Code Cheap: Why Taste and Domain Expertise Are the New Moat

Published on 15 Sept 2026 by Shri Krishnan

An essay from 9horsemen's founder. For practical help running a business, see the blog. More thinking like this on our podcast, Why Your Software Sucks.

A florist inspects a white hydrangea inside a commercial cooler, with a tablet displaying a temperature alert in the background, illustrating the importance of specific domain expertise over generalised software.

TL;DR:

  • The Core Problem: The industry is distracted by teaching people how to code alongside AI, ignoring that when execution becomes virtually free, the code itself is no longer a competitive advantage.
  • The Structural Reality: Massive enterprise SaaS companies cannot solve hyper-specific business problems because their economic models force them to build generalised, feature-bloated platforms that lack deep domain “taste.”
  • The Solution: The future belongs to operators with elite domain expertise. Your knowledge of the problem, and your taste in solving it, are what AI can’t copy.

Will AI replace programmers? It’s already replacing the typing: AI writes working code for almost nothing. What it can’t replace is knowing which problem is worth solving and what good looks like. That’s taste, and it’s now the moat.

Will AI replace programmers?

AI is taking over the mechanical part of the job, writing code, faster than anyone expected; what’s left is deciding what to build and why.

For the last twenty years, the corporate world worshiped a very specific type of professional: the builder. If you could type Python faster than anyone else, if you could manipulate a financial model with your eyes closed, or if you knew the secret syntax to make a database yield its secrets, you held the keys to the kingdom. We built an entire economic hierarchy around the mechanical execution of tasks.

Then, generative AI arrived and completely commoditised execution.

Right now, the internet is aggressively pushing a panicked narrative to the incoming workforce. The gurus claim that to survive, you must immediately become a “prompt engineer,” learn to code alongside an AI, or memorise a weekly list of a hundred new software tools. It is a massive, collective distraction. It forces people to focus entirely on the mechanical output of the machine, rather than the strategy guiding it.

When a machine can write flawless code in twelve seconds, learning to type code slightly better is a rounding error. When execution becomes free, the only currency left that matters is taste.

What happens when anyone can build software?

The software itself stops being an advantage, because a competitor can build the same thing by Sunday night; what wins is knowing the customer’s problem better.

To understand what “taste” actually means in a business context, look at a local commercial florist.

Three years ago, if a non-technical entrepreneur wanted to build a custom inventory management system for local florists, they were dead on arrival. They would need a technical co-founder, a massive budget, and six months of development time just to get a basic web app off the ground.

Today, that same entrepreneur can sit down with an autonomous coding agent, describe what they want, and have a functional, secure web application deployed by Sunday night.

But here is the brutal reality most people miss: so can everyone else.

When the cost of building software drops to zero, the software itself is no longer a competitive moat. The code is worthless. The founder who wins isn’t the one who writes the most elegant database architecture. The founder who wins is the one who intimately understands that a florist’s highest margin product—white hydrangeas—will rot in exactly 72 hours if the ambient cooler temperature fluctuates above 38 degrees on a delivery truck.

That founder builds a notification trigger specifically for that single, agonising variable. The software doesn’t win because of the code. It wins because of the human taste applied to the problem.

In this new economy, “taste” isn’t about knowing what colour palette looks pretty. It is a deep, observational understanding of human behaviour, psychology, and operational friction. It is the ability to look at a sprawling, chaotic business process and accurately identify the single bottleneck that actually matters.

Why can’t big software companies build for niche problems?

Because they need huge contracts to survive, so they build general platforms for everyone instead of tools shaped around one trade’s problem.

If the future of work relies on hyper-specific domain expertise, you would assume the massive tech companies are building tools to support it. They aren’t.

They are doing the exact opposite. Because venture-backed enterprise SaaS companies require massive, seven-figure contracts to survive, they cannot afford to have taste. They cannot build a tool specifically for the florist’s rotting hydrangeas, or the HR manager’s weirdly specific compliance spreadsheet, or the paralegal’s municipal zoning headache.

Their economic model forces them to build generalised bloatware. They sell massive, blank-canvas platforms and omniscient dashboards that claim to do everything for everyone. But a tool built for everyone is fundamentally useless to the operator trying to solve a hyper-specific, localised nightmare.

Big Tech gives you a generalised AI to write a generic email. They do not give you the scalpel required to carve out the exact operational friction slowly killing your Tuesday afternoons.

Which skills will matter most when AI does the coding?

Judgement: knowing people, business and your field well enough to pick the one good answer out of a pile of average ones.

We are entering a golden age for the strategic generalist. The professionals who will dominate the next decade are the ones aggressively studying art, history, psychology, and the raw, unglamorous realities of how businesses actually run.

They will let the AI handle the mechanical execution. Their job will be to act as ruthless, aggressive curators—sifting through a mountain of average digital output to find the one structural truth that actually moves the needle.

This shift from execution to taste is the foundational philosophy at 9horsemen. We recognise that generalised, bloated software is a dead end. We do not build monolithic dashboards that try to boil the ocean. We build focused products for people who know their work better than any developer does, starting with local service businesses.

Will coding exist in 5 years?

Yes, but as a smaller part of building software: the code will be cheap, and the advantage will sit with the people who understand the problem.

When anyone can generate a full-stack codebase in seconds, code itself ceases to be a competitive advantage. The winners in this next era of software won’t be teams that write syntax by hand—they will be aspiring entrepreneurs and SMEs who possess deep domain insight, uncompromising taste, and an intimate understanding of their end-users’ daily friction.

That’s the bet behind Gallop, our first product, built for one kind of expert: someone who runs a local service business. A barber or a plumber knows their customers and their day better than any developer. Gallop turns that knowledge, described in their own words, into a website, online booking and business app in a few minutes.

If you’re building software of your own, Gallop isn’t that tool, but the same rule applies: your taste and your knowledge of the problem are the moat. I write more about that in the other essays, talk about it on the podcast Why Your Software Sucks, and tell the long version in my story.