Deadlink Labs
Build to Understand.
Featured
IN PROGRESSFeatured on the bench · LOG 001
Building the Deadlink Labs website with AI, in public
Yes, AI can build a website. But the interesting part happens way before the first line of code. Start prompting too soon, and you get a website that could belong to anyone. Think first. Plan it. Then prompt.
$ npm run dev
astro v5.18.2 ready in 415 ms
┃ Local http://localhost:4321/
watching for file changes...Client work
Work I built for other businesses. Each one has its own log entry.
-
They'd already done the research. They just couldn't tell who was worth writing to.Dozens of prospects researched, then scattered across notes, bookmarks and an inbox. Two days of building with Claude Code turned it into a console that says who's worth an email this week.
-
Nobody fills in the form. I automated it away with n8n and Claude.Field technicians never finished their service reports, and in a regulated plant an unfinished report is an audit finding. A two-minute voice note goes in, a finished report comes out.
-
Uruguay Outfitters: a hunting and fishing lodge websiteThey sell a week away to people who have never seen the place, so the photographs have to do the selling. A fast, photo-heavy site the team can update without calling me.
-
Crehana: automating an e-learning content operationNobody could say how many courses were in production without asking several people and piecing it together. Three years later a dashboard answered it, and more than $22,000 a year in tools and manual work was gone.
Recent
Products
Throwback
Before the lab: Dig through the artifacts that survived.
Who runs this
Marcelo Brouard, Buenos Aires. 20+ years turning messy operations into systems that run themselves: post-production teams, pipeline automation, data and dashboards, AI workflows, and the occasional website.