This is the origin story of the site you are reading right now.
The idea started with a simple observation: most design leadership content exists in one of two modes. Either it is polished conference keynotes that feel removed from the daily work, or it is LinkedIn posts so compressed they lose all nuance. I wanted a third space, something that shows the texture of real design leadership through actual artifacts, actual decisions, actual conversations. Updated weekly, tied to real work, with a visual identity that reflects craft.
The constraint I set for myself was that the entire thing had to be built and published using AI-native tools. No Figma. No manual coding. No traditional CMS. I wanted to see if the current generation of creative AI could take me from a blank prompt to a live, deployed website with real content, and I wanted to document every step honestly, including the parts that did not work.
The design phase
I started in Claude (the conversational AI, where you and I talk), working through the site concept, content categories, and information architecture before writing a single line of code. This is the part most people skip when they use AI tools for design work, and it is the part that matters most. By the time I had a prompt ready for Claude Design, I had already defined seven content categories (The Reframe, Crit Notes, The Decision, Maker’s Log, Toolbox, Reading List, State of the Craft), a landing page structure with three zones, a content page template, and a visual identity direction rooted in editorial minimalism.
I also had a specific interaction pattern I wanted to carry over from a previous project: a vertical progress bar on the featured content carousel that fills with color from top to bottom as a timer before the content rotates. Small detail, but it signals to the reader that the site is alive and curated.
Claude Design took the prompt and produced a working v1 that was remarkably close to what I had described. The editorial typography, the card grid, the featured content area with the timer bar, the full-bleed illustrations on content pages. It iterated through my feedback in a conversational loop, refining layout, color, and component behavior until we had something I was genuinely pleased with.
The deployment wall
This is where it got interesting. Claude Design can export to several formats, including Canva (for drag-and-drop publishing) and Claude Code (for developer-grade deployment). I tried the Canva path first and hit size limits immediately. The React bundle was too heavy for Canva’s importer, and even after Claude Design tried stripping it down to static HTML and then to flat screenshots, Canva kept rejecting it. That path is better suited for simpler marketing pages.
So I went with Claude Code. The handoff was clean: Claude Design bundles the project files and sends them directly to Claude Code with implementation instructions. But getting from there to a live URL involved a series of small technical steps that someone without development experience would find genuinely frustrating.
I needed to install Node.js (which was not on my machine), then the Vercel CLI (which required sudo permissions on Mac), then create a Vercel account, then link the project, then configure DNS on my domain registrar. Each step was individually straightforward, but the cumulative friction was real. I hit a permissions error, a plugin compatibility issue, a 404 caused by a misnamed index file, and a DNS configuration that required switching my registrar from hosting-mode to basic DNS before I could add the right records.
Pasting a screenshot of a terminal error and getting back the precise fix in seconds — that is the part that felt genuinely new. I was pair-programming with an AI, not on code, but on infrastructure.
Claude (the conversational AI) walked me through every single one of these issues in real time, diagnosing errors from screenshots I shared and giving me the exact commands to run.
The weekly workflow
The site is now live at ldlr.design, deployed on Vercel with automatic SSL. The publishing workflow I have landed on is: I work with Claude throughout my week on real leadership tasks (drafting messages, writing scorecards, developing strategy, creating illustrations). At the end of the week, we pull my recent conversations, identify which ones contain publishable insights, draft the articles together, create illustration prompts, write LinkedIn posts, and deploy. The content is authentic because it comes from actual work, not from sitting down to write a blog post.
The name of the site is Design Outcomes, and the domain is ldlr.design, my initials. When someone asks what it stands for, I get to say Leonardo De La Rocha, which is a better introduction than any bio page.
If you are a design leader thinking about how to share your work and thinking more visibly, the barrier to entry just dropped to almost zero. The tools are here. The interesting question is no longer whether you can build the thing. It is whether you have something worth saying.