Project 02 · This site · another AI journey
Built in a single Claude session from a briefing document. No developer, no framework, no code written by hand.
About the project
The Practitioner AI series started as six LinkedIn posts. Post 1 went live on April 11, 2026 — a first post ever, from someone who had never published original content on LinkedIn. By the time the metrics came in — 1,299 impressions, 780 members reached, senior executives and CEOs in the audience — it was clear the series was reaching the right people.
But LinkedIn posts have a ceiling. The real story — the hours of prompt refinement, the state machine solution, the screenshot technique, the $6.79 domain, the afternoon deployment — does not fit in a format built for scrolling. Readers who wanted more had nowhere to go. This site is where they go.
What happened next illustrates the AI methodology more clearly than anything else on this site. Claude was asked to review the entire Instant Pot Doc conversation and extract what a website would need — the story arc, the key moments, the Behind the Build narratives, the brand palette, the hosting instructions, the technical constraints. Claude authored a structured briefing document from that conversation. That document was then handed to a second Claude session — which built this entire site from it.
Claude wrote the requirements. Claude built the site. David approved, directed, and deployed.
One pattern emerged from this project that is now used deliberately. A single AI conversation has limits — not just technical ones, but practical ones. As a chat grows longer it carries more context, more history, more accumulated decisions. At some point the right move is to take the output of that conversation and hand it to a new one with a clean, specific mandate.
One chat to think. One chat to build. One chat to publish.
The sprawl stays in the original conversation. The new chat gets only what it needs. When a conversation gets heavy — summarize, brief, and spawn. The output of one chat becomes the input of the next. That is a workflow pattern any executive can use today.
No developer. No framework. No code written by hand. One briefing document. One afternoon.
Behind the Build
Three narrative stories about the decisions and moments that shaped this site.
01
The Briefing Document
Writing requirements before writing a single line of code
Before I asked Claude to build this site, I wrote a briefing document. Not a vague description — a structured brief with every detail I could anticipate: brand palette with exact hex codes, font names, section order, full content for each section, hosting instructions, and an explicit note that I am not a coder and everything must be deployable by drag and drop.
I also prepared a second document — narrative stories, already written, ready to drop into the Behind the Build section. Claude was given everything it needed before the first message was sent. The entire site was produced in a single session.
The lesson held from the Instant Pot Doc build: the quality of what AI produces is directly proportional to the quality of what you give it. Write the brief before you open the chat.
The executives who got bad software were the ones who could not articulate what they wanted. That rule does not change when your builder is a language model.
02
Briefing Doc to Live Site
What one session with Claude actually produced
The entire site you are reading was built in a single Claude session. I uploaded two documents — a full briefing and a snippets file — and within that conversation Claude produced a complete, deployable website. Then I uploaded the Instant Pot Doc logo mid-session and asked if it could brand the project card. Claude converted it to base64 and embedded it directly in the HTML so the image travels with the file — no external assets, no image folder, no dependencies.
When I asked whether the website itself should be listed as an AI project, Claude said yes immediately — and added a third project card, new Behind the Build snippets, and updated the deployment instructions, all within the same session.
From first message to a file ready for Netlify Drop: one conversation, one afternoon.
Treat AI sessions like working sessions, not search queries. Bring everything you need, stay in the conversation, iterate in real time. The output compounds within a session in a way it never does across separate chats.
03
Drill Down, Not Scroll Down
The UX decision that restructured everything
The original site was a single page with flat sections — Projects, Practitioner AI, Behind the Build — all stacked together. As the project list grew, stories from different projects were going to pile up with no way to tell which project they belonged to.
I described the problem to Claude and asked for two UI options with a recommendation. Within minutes Claude came back with a visual diagram of both approaches and a clear recommendation: project hub pages. Each project card links to a dedicated page with everything about that project in one narrative arc. The reasoning was specific — senior executives make one decision early, is this worth my time, and if they click a card they have already self-selected. Give them the complete story without making them work for it.
The restructure went from decision to three linked HTML files in a single session. Index page, Instant Pot Doc hub, davidsevans.com hub — all built, linked, and ready to deploy as a folder.
AI is not just a builder. It is a thinking partner. Describe the problem, ask for options, ask for a recommendation with reasoning. The answer you get back is often faster and more structured than what you would have worked out alone.