Coreygetcorey.ai - the site Corey built, while you read it
Corey's own marketing site, knowledge base and checkout - researched, written and shipped by Corey.
getcorey.ai is Corey's own website, and the only project in the portfolio Corey built from nothing. The repository starts on 22/06/2026, after Corey existed, so every page, guide and pricing table on it is Corey's work. You are reading the proof rather than a description of it.
What this is, on the record
- Started
- 22/06/2026
- Commits
- 138
- Status
- Shipped and running
- Who did what
- Built and run by Corey. This is the one project in the portfolio that postdates Corey entirely.
- Built with
- Astro, React islands, Tailwind, Cloudflare Workers, Stripe
The project
Every other page in this portfolio describes work that happened elsewhere. This one you are standing in.
getcorey.ai is Corey’s marketing site, knowledge base, documentation and checkout. The repository’s first commit is 22/06/2026, which is after Corey itself existed - so unlike the rest of the portfolio, there is no split between “what was built before” and “what Corey does now”. It is all Corey.
That makes it the honest test. An AI operating system that cannot research a market, write the pages, check its own facts and ship them is not an operating system. It is a chat window.
What we did
The site was scaffolded in June 2026 and rebuilt around adoption in July. The first version sold features. The rebuild sold the first session instead - what Corey is, the Command Center you get, and the three steps to have it running.
Then the content engine. Keyword research produced a map of where the real gaps were, which turned out to be two: the definitional queries for a category that is still being named, and UK compliance content, where almost everything ranking is written for the IRS and dollars.
Pages get built to that map, one per pull request, each one starting from a live look at the actual search results rather than an assumption about them.
What was harder than expected
The plumbing, not the writing.
Structured data drifts silently. Plain-text twins do not get redirects. An index of URLs for
language models rots the moment a route moves, and nothing tells you - llms.txt had accumulated
nine dead links before anyone checked, and nine of twenty-four pages were publishing step-by-step
markup that appeared nowhere on the page.
None of that shows up in a browser. It shows up months later as content that quietly does not perform. The fix was to stop relying on remembering: every check that matters now runs on every change, and fails the build rather than filing a note.
What we decided
The home page sells adoption, not features
The home was rebuilt as a pure adoption page - what Corey is, the Command Center it gives you, and the three steps to start. Feature lists convert people who already understand the category. Nobody understands this category yet, so the job of the home is to make the first session happen.
No invented testimonials, metrics or case studies
A hard rule from the start. No logos we do not have, no numbers we did not measure, no quotes nobody said. It costs the easiest conversion lever on the internet, and it is the reason this page can exist at all - the portfolio is real, so it does not need dressing.
Comparison pages live in /learn, not /compare
Settled 25/07/2026. A /compare namespace would have been four thin pages each needing a hub, nav entry and template, when the slug already carries the ranking intent. Namespace moves are expensive - redirects cover HTML routes but not plain-text twins - so the decision was to not mint one we might regret.
Gates in CI, not in someone's memory
The repo ran build, link-integrity, lint and convention checks as npm scripts that only fired when somebody remembered. The link checker existed because llms.txt had already rotted to nine dead URLs. On 25/07/2026 the lot moved into CI, which immediately caught a linter that had never been a declared dependency.
What got built
- A marketing site, knowledge base and docs set across 78 pages.
- A content engine with per-page structured data, plain-text twins for LLMs, and an llms.txt index.
- An interactive product tour built from real Command Center screenshots, not mockups.
- Stripe checkout and a 28-day no-card trial.
- A link-integrity checker, a structured-data visibility checker and a conventions gate, all wired into CI.
What Corey runs now
Research before writing
Every content page starts with a live search of the real results for its target query, so the angle comes from what is actually ranking rather than from an assumption about it.
Writing and shipping
Pages are drafted, fact-checked against primary sources, run through the build and link gates, and opened as a pull request with the reasoning in the description.
Holding the line on claims
On 25/07/2026 Corey found the proof band claiming three customers when there was one, and corrected it. Catching your own overstatement is the job.
Keeping the plumbing honest
Dead links, invisible structured data, drifted metadata - the unglamorous maintenance that decays silently is checked on every change rather than audited occasionally.
This one runs the other way round. Everything else in the portfolio taught Corey something; this is Corey demonstrating what it learned. It is also the honest test of the product - if an AI operating system cannot research, write, fact-check and ship a website without a marketing team, the claim does not hold. The site is the claim, executed.
Run yours the same way
Everything above is one operator across a portfolio. Corey is that operator, packaged. 28 days on us, no card.