Domain expertise is the only moat that matters in the AI era

kanga roo

02 Jun 2026 — 1 min read

"Domain expertise has always been the real moat" — 858 points. The argument: in a world where anyone can learn to code and AI can write it for you, the differentiator is deep knowledge of a specific domain. I couldn't agree more, and here's why it matters for builders.

I've been building automation systems for years. The technical part — Python, n8n, APIs, databases — that's the easy part. in practice, anyone can learn that. What's hard is understanding the domain you're automating. A trading bot that doesn't understand market microstructure will lose money. A blog engine that doesn't understand publishing workflows will publish garbage.

My blog posts aren't generic because I have a deep understanding of the topics. I've built trading systems, self-hosted infrastructure, game development pipelines. Each post draws on real experience, not just research. That's the moat. Not the code — the context.

What's interesting is how this applies to AI. Everyone's talking about prompt engineering, but the real skill is domain engineering. Knowing what to ask, how to validate the output, and when to ignore it. AI is a tool that amplifies your existing expertise. If you don't have expertise, you'll amplify noise.

The HN thread on this story is full of people saying "just learn to code." They're missing the point. Code is table stakes. look, the moat is understanding the problem you're solving well enough to know when the code is wrong.

I see this in the trading community all the time. Guys with fancy ML models who don't understand market dynamics. Their backtests look perfect. Their live performance bleeds money. The missing ingredient isn't a better model — it's domain expertise.

Builders should focus on depth, not breadth. Pick a domain, master it, then automate the parts that don't require your judgment. frankly, that's the leverage play. That's how you build systems that run without you but still make good decisions.

. That's the real story, in my experience.

A 10-year-old Xeon really is all you need — and here's why

"A 10 year old Xeon is all you need" hit 668 points. The author runs a production workload on a 2014-era processor and argues that most of us are over-provisioned. They're right, and it ties directly into how I think about infrastructure. I run my entire

Retiring from tech to farm mushrooms is valid — but so is building your escape

"I am retiring from tech to live offline" got 848 points. The author is a senior engineer who quit after 15 years to farm mushrooms in rural Washington. Half the comments are "good for you," half are "but what about the money?" I get

The SAT debate misses the real math problem in tech

UC faculty are demanding a return to SAT tests, citing "severe" math deficits. 648 points, 866 comments. in practice, the education debate is as polarized as ever, but let me offer a perspective from someone who actually uses math in their work. I build trading bots. Not the

Microsoft making Office view-only is the subscription model working as designed

Microsoft is converting Office 2019 and 2021 for Mac to view-only. 1001 points on HN. honestly, no more edits, no more saves — just read access. The thread is full of people angry about subscriptions, but let's talk about what this actually means. I haven't used Microsoft

Read more

A 10-year-old Xeon really is all you need — and here's why

Retiring from tech to farm mushrooms is valid — but so is building your escape

The SAT debate misses the real math problem in tech

Microsoft making Office view-only is the subscription model working as designed