Blueprint
Marketing Website
A site whose whole job is to tell people what you do.
A website whose entire job is to tell people what you do. Services, proof, who you are, how to reach you. Sometimes called a brochure site.
What defines it is what it does not have. No accounts, no logins, nothing that changes from one visitor to the next.
“Static” is the most misleading word in this category. It does not mean frozen. It means each page is assembled once, when the site builds, instead of being rebuilt from scratch for every visitor who arrives. The content still changes whenever you want it to. It can pull from an API, rebuild on a schedule, or rebuild the moment someone edits a page, and a deploy is automatic and takes under a minute. What you give up is per-visitor variation. What you get for it is a site with no database to breach, no server to patch, and nothing sitting between the visitor and the page.
How I orchestrate the build
Editor
Where I sit. Claude Code runs inside it, so the work and the code are in one window.
Orchestration
Does the actual building. I describe the change, it writes it, I check the result.
Building from my phone
Same project, away from the desk. Useful more often than it sounds.
Input (I don't type)
I talk, it types. Dictation that keeps up with a train of thought.
What it's built with
Framework
Builds the pages ahead of time and ships zero JavaScript unless a page needs it.
Styling
Styles written next to the markup, so there is no separate stylesheet drifting out of sync.
Images
Resizes and converts to modern formats at build time, so nobody downloads a 4MB photo.
Structured content
Posts, services, case studies. Each type gets a defined shape, and a missing required field or the wrong kind of value fails the build instead of silently saving. This site runs eight of them.
SEO
Generates the sitemap and the tags search engines read. You write them, so you know what they say.
Form handling
The bit of code that catches a form submission. Serverless: it exists for the 50ms it runs, then it is gone.
Bot protection
Sits on the form and blocks bots without making a human click traffic lights.
Where the files live
The working copy
The real one. A directory in my projects folder on this computer, where the work actually happens.
Version control + what the deploy watches
Every version of every file, forever, and the ability to go back to any of them. Pushing here is what triggers a deploy. It is version control, not a backup.
Updates
Watches every package the site is built on. Once a week it bundles the updates into one change, the site builds on a clean machine, and if the build passes it merges and deploys by itself. If it fails, nothing ships and the live site stays as it was.
The backup
Every Sunday, a full copy of every repo, uncommitted work included, goes to cold storage that has nothing to do with GitHub. That is the actual backup.
Big files
Video and large images. A code repo is the wrong home for a 500MB file, so those live here and the site points at them.
Where it's hosted and how it deploys
Integrations
Form email
Delivers the form submission to an inbox. An API rather than SMTP, so you can see whether it actually arrived.
Analytics
Traffic numbers without following anyone around, which is also why there is no cookie banner.
Legal policies
Privacy policy and terms that update themselves as the laws change, rather than a PDF from 2019.
The questions I get most often
These come from people building sites on WordPress, on other builders, and from people who have not built one with AI yet. I came from WordPress, so a lot of my answers are framed against it. None of this is gospel. It is how I do it and why, so take what fits what you do. Most have the same shape: something you currently pay for, maintain, or worry about stops existing here, and the fair question is where it went.
- Q What about a page builder?
- Astro and Claude are my builder. I describe the change and it lands in the code, so what ships is a real page instead of a builder's markup wrapped around the content. It also means nothing I build is locked to a plugin that can stop being maintained.
- Q Do you really feel in control without a builder?
- More than I did with one. Every file is readable, every change is in version control, and nothing is hidden behind a settings screen. What a builder gave me was control over pixels by dragging them, and that turned out to be worth less to me than it felt like. A marketing site is HTML, CSS, and content, and what these tools write is clean and it works. If you would trust an agent to build software with logins and a database, a marketing site is the easy case. I care far more about the site working and making the client money than about every pixel being placed by hand.
- Q How do you know what the agent wrote is any good?
- The same way I judge any marketing website. Does it load fast, do the forms go through, and is the SEO right? I can check all three without opening a file, and to me they are the whole job of the site. I do not read the code line by line, and for a site whose job is to tell people what you do, I do not need to. The build catches some of the rest: content that does not match its defined shape fails before anything ships.
- Q Isn't static too limited?
- Not for what I build with it. Static describes when the page is assembled, not whether it can change. It can pull from an API, rebuild on a schedule, or rebuild the moment you edit something. What I give up is showing a different page to each visitor, which a marketing site does not need.
- Q What about bookings, payments, or logins?
- Forms need nothing extra, that is the stack above. For bookings and payments I usually add one small function, or an embed from a service that already does it well, and the site stays static. Logins are where I stop calling it a marketing website. Once people have accounts and their own data, it is a web app, and that has its own blueprint.
- Q Can you sell products on it?
- Digital products, yes. Stripe Checkout takes the money and a small serverless function delivers the file, and the site stays static, because there is no stock to count and nothing to ship. Physical products are a different animal and get their own blueprint.
- Q How do you edit it?
- The same way I build it. I open the project, describe the change, and it ships. That is section one of this page, and it is why there is no editing interface in my stack by default. For a client who will genuinely edit their own site, I add a CMS. That is its own question below.
- Q Can a client edit it themselves?
- Yes, and then a CMS goes in. For me it is a real decision rather than a checkbox. I have done it with headless WordPress and with Sanity. The trade is the same either way: you gain a place to type, and you take on another system to log into, keep current, and pay for. The test I use is frequency. Most of my clients touch their site once or twice a year and are better off sending me the change, which ships the same day. If you are genuinely going to be in there every week, or you have a team of editors or someone writing posts, I want to know before the build starts, because it changes how the whole thing is put together and it is far cheaper to plan for than to retrofit.
- Q What about custom post types, fields, taxonomies, relationships, filters, and facets?
- I do all of it, with less ceremony. A post type is a content collection. Custom fields are that collection's schema, and they are typed, so a missing required field or the wrong kind of value fails the build instead of quietly saving blank. A taxonomy is just a field on the content, which means no term tables and no orphaned categories nobody can delete. Relationships are a reference by slug: this page pulls every tool you see from the tools collection that way.
- Q So filtering and faceted search still work?
- Better, in my experience, and this is the part people do not expect. Everything is already in memory when the site builds, so a filter is an operation on an array rather than a query. My tools page sorts 92 items by type and category with no database behind it. Nothing runs on a server while a visitor waits, so filtering cannot slow the site down or fall over under traffic. In WordPress each of those is a database query, which is a large part of why you end up buying a caching plugin to undo them.
- Q What about SEO? Don't you need Yoast or Rank Math?
- I do not use either, and it is worth knowing what those plugins actually do. Almost all of it is writing tags into the head of the page: the title, the meta description, the Open Graph tags that control how a link looks when it is shared, the canonical URL, the robots directives, and the sitemap. Those are the output. The plugin is a form for typing them, plus a traffic-light score that grades your writing. I write the tags directly, so there is no form in between and nothing to configure.
- Q What about schema markup?
- I generate it for every page. This site emits structured data for the organization, the person, breadcrumbs, case studies, videos, the podcast, and the community offer, which is what search engines and increasingly what AI assistants read to understand who you are. That is code rather than a plugin setting, so it cannot be half-configured and it cannot silently stop working after an update.
- Q Will it actually rank?
- I cover the technical side completely: fast pages, clean markup, correct tags, valid schema, and a sitemap. That clears the bar that plugins are sold to clear. What no plugin has ever done is give you something worth ranking for, and that part is content and proof rather than configuration. A green light on a dashboard is the easy half.
- Q What about security?
- Almost nothing to attack. No server, no database, no login, no plugins to keep patched. The pages are files. The risk moves to my accounts (GitHub, Cloudflare, and the domain), so those are what I protect. The one piece of code that does run when someone visits is the small function that catches a form, so that is the part I review: the bot check runs first, everything a visitor types is escaped before it goes into an email, and the recipient is fixed in the code rather than taken from the form. This is the single biggest reason I build this way, and it is why a security plugin like Wordfence would have nothing to do on my sites.
- Q What about updates?
- There is no plugin ecosystem in my setup to keep in step, so nothing breaks on a Tuesday because two vendors shipped on the same day. The site does still sit on packages, Astro and a few dozen libraries, and those get security fixes like any other software. GitHub's Dependabot watches them and bundles the week's updates into one change, and the site has to build on a clean machine before anything merges. If it builds, it deploys on its own. If it fails, nothing ships, and the site that is already live stays live.
- Q What about backups?
- Git is not the backup, even though it gets called one. Git is version control: every version of every file, and a way back to any of them. The backup is separate. Every site lives in three places: the folder on my machine, GitHub, and a weekly copy in Backblaze B2 that includes work I have not committed yet. There is also no database to back up, which is most of what a WordPress backup plugin is protecting.
- Q What about caching?
- I do not think about it anymore. Caching exists to avoid rebuilding a page for every visitor, and my pages were already built before anyone asked. They sit on a CDN close to the reader, so there is no work left to skip.
- Q What happens if you're gone?
- Someone else picks it up, and that is a big part of why I build this way. It is a standard Astro project in a git repository, not a custom platform or a pile of plugin settings, so any developer, or any agent, can open it and read it. The repo carries its own documentation of how it is put together and why, kept current every time the code changes.