Essay 05 March 2026 8 min read

The Taste Stack: What I Use to Build

This is not a tutorial. It is a philosophy of tool selection as taste expression. Your stack is a statement about what you value — and mine tells you everything about how I think.

Developers obsess over their tech stacks. React or Vue? PostgreSQL or MongoDB? AWS or Vercel? These are identity questions disguised as technical decisions. The tools you choose say something about what you value, how you think, and what you are optimising for. Developers know this instinctively. They just pretend the decisions are purely rational.

Non-coders building with AI have the same dynamic, but we rarely talk about it. So let me lay mine out — not as a recommendation or a tutorial, but as a window into how taste applies to every decision, including the ones that seem purely functional.

The builder: Claude Code

Claude Code is where everything starts. It is the tool I use to build every product, and choosing it was a taste decision as much as a technical one.

There are other AI coding tools. Cursor, GitHub Copilot, various wrappers and interfaces. I tried several of them. Claude Code stuck because it matched how I think. I do not think in code. I think in outcomes. I think in "I want a page that feels like this" or "I need a system that does this." Claude Code lets me express those thoughts in natural language and translates them into working products.

The key insight is that the tool you choose shapes the output. Not just technically — aesthetically. The conversation style, the defaults, the way the tool interprets ambiguity — all of these influence the final product. Using Claude Code is not the same as using Cursor, even if the underlying task is identical. The tool has a sensibility, and that sensibility interacts with yours.

This is true for every creative tool in history. A Leica produces different photographs than a Hasselblad, even in the same photographer's hands. A Fender produces different music than a Gibson, even played by the same guitarist. The tool is not neutral. The tool is part of the taste.

The host: Netlify

Every product I build goes to Netlify. This is a decision that most people would not think twice about, but it reflects a specific set of values.

I chose Netlify because it is simple, fast, and stays out of the way. There is no lock-in, no complex configuration, no infrastructure to manage. I point it at a directory of files and it serves them. When I need something to go live, it goes live in minutes.

This reflects a deeper principle: I optimise for speed to live, not for scale. I am not building products that need to serve millions of concurrent users. I am building products that need to exist — that need to go from idea to URL as fast as possible. Netlify serves that goal. More complex hosting solutions would give me more capability but slow down the thing I care about most: shipping.

The choice also reflects a taste for independence. Netlify does not own my content or my code. If I wanted to move everything to Cloudflare Pages tomorrow, I could. The products are mine. The hosting is just where they live. That matters to me, even if it would not matter to everyone.

The automator: GitHub Actions

Several of my products update automatically. CultureTerminal pulls from RSS feeds every morning. The Relevance Index rescores over a thousand brands every week. The Pattern generates a daily culture briefing with AI analysis and audio. All of this runs on GitHub Actions — scheduled tasks that execute without me touching anything.

Automation is a taste statement. It says: I care about consistency more than control. I care about systems more than heroics. I want the products to be alive — updating, refreshing, evolving — even when I am not paying attention.

There is a specific pleasure in building something that runs without you. It is the pleasure of a well-designed system, a mechanism that works because it was thought through carefully upfront. Every automated pipeline I build is a bet that the system I designed will produce good results on its own. And that bet only pays off if the curation logic — the rules about what to include, what to surface, what to prioritise — is good enough to run unsupervised.

That is taste embedded in infrastructure. The system runs on my judgment even when I am asleep.

The measurer: Plausible

I use Plausible for analytics, and this is probably the most deliberate taste decision in the stack.

Google Analytics is free, more powerful, and more widely used. So why Plausible? Because Plausible respects the visitor. No cookies, no tracking, no data sold to third parties. A simple dashboard that tells you how many people visited, where they came from, and what they looked at. Nothing more.

This is a values-driven choice, not a features-driven one. I believe products should respect the people who use them. I believe analytics should inform decisions, not feed surveillance capitalism. I believe "less data, better interpreted" beats "more data, barely looked at." Plausible is the analytics tool that reflects those beliefs.

And here is the thing: the simplicity of Plausible has made me a better analyst. When you only have a few metrics, you pay attention to them. When you have hundreds, you look at none. Constraint breeds focus. The tool's limitation is its greatest feature.

The messenger: Buttondown

For newsletters, I use Buttondown. Not Mailchimp, not Substack, not ConvertKit.

Buttondown is small, independent, opinionated, and built by one person. It does emails and it does them well. No social network, no recommendation algorithm, no growth hacking features. Just: write something, send it to people who asked for it.

The choice says something about what I think newsletters should be. They should be a direct line between a writer and their readers. Not a platform that inserts itself into the relationship, gamifies the experience, and optimises for engagement over quality. Substack has become a social network that happens to send emails. Buttondown has stayed an email tool. I prefer the tool.

This is also a taste statement about independence. Substack owns the relationship with your readers. Buttondown lets you own it yourself. If Buttondown disappeared tomorrow, I would still have my subscriber list and could move to another tool. That independence matters — not because I plan to leave, but because the option to leave keeps the relationship honest.

The pattern in the stack

Looking at the full stack — Claude Code, Netlify, GitHub Actions, Plausible, Buttondown — a pattern emerges. Every tool shares certain qualities: simplicity, independence, respect for the user, speed over scale, and a clear point of view about what it is and what it is not.

None of these tools try to do everything. None of them lock you in. None of them are the most powerful option in their category. But all of them are coherent — they know what they are, and they do that thing well.

That coherence is taste expressed through infrastructure. The same principle that makes a good product — clarity of purpose, restraint, knowing what to include and what to leave out — applies to the tools you build with. A bloated toolstack produces bloated products. A focused toolstack produces focused products. The stack is the taste.

Your tools are your values

I am not suggesting everyone should use my stack. The right tools depend on what you are building, what you value, and how you think. That is exactly the point.

The choice of tools is not a neutral decision. It is a creative decision, a taste decision, a statement about your priorities. When you choose a simple tool over a complex one, you are saying you value clarity over capability. When you choose an independent tool over a platform, you are saying you value ownership over convenience. When you choose a tool that respects users over one that maximises data collection, you are saying something about what kind of builder you are.

Most people make these choices unconsciously, defaulting to whatever is most popular or most recommended. But the best builders — the ones whose products feel coherent from top to bottom — make these choices deliberately. They understand that taste starts before the first pixel is designed or the first word is written. It starts with the tools.

Your stack is not just how you build. It is what you believe. Choose accordingly.

Subscribe to Taste Machines

New essays on AI, taste, and building. No spam.