Earlier this week I wrote that the dinosaur caught up, and that the magic went with it. Here’s what that actually looks like.
https://consulting.waltercolindres.com is live. I want to show it off, but I also want to show how it gets built, because the how is the more interesting story.
It’s not one magic prompt. It’s a pipeline.
It starts with me. I run Claude Code through Paseo and hand it the raw material: metrics, data, and what I want the page to say. Claude builds the page from layout templates we designed together earlier. Same structure, same polish, every time.
Then the visuals. I pass the same content to ChatGPT, which generates images in the illustrative style I developed for my work. Once the assets land in the same folder as the content, the page is complete.
Claude opens a pull request on GitHub. That kicks off a publishing flow through Vercel and spins up a preview. When it’s ready, I get a ping.
Then it’s my turn again. I proofread. I check that it looks the way I want. If it’s right, I promote it to production.
Notice where I show up in that flow. The beginning and the end. I decide what gets said and why. I decide when it’s good enough to ship. Everything in between is production, and production is exactly what I wanted off my plate.
That’s the trade I was wrestling with. I gave up the hunt for the bug and the pixel. What I kept is the part that was always mine. What the site says, who it’s for, which case studies earn a spot, what “good” looks like. That comes from 15 years across QSR, eCommerce, and consumer digital.
It’s also what I’m offering. If you’re trying to figure out how AI fits into your product, design, or content workflow, I’ve been living in that question every day.
I’m still tuning the pipeline. Next up is shrinking the gap between “here’s the data” and “here’s the preview” even further.