Author
Margus Veeber
Head of Web at Pipedrive · Founder, VeeberMedia · Host, Web Unpacked
I lead the web org at Pipedrive, a sales CRM used by more than 100,000 businesses, and I have run VeeberMedia as my own studio since 2015. Ten years of client and product work before that, from early stage to enterprise.
Everything on this shelf comes out of that work. The stacks are the ones I pick when a site actually has to ship, the skills are the ones running in my own repos, and the prompts are the ones that survived a project with visitors on it. Nothing gets published here that I have not run myself.
I host Web Unpacked, a podcast about how web teams really operate, and I post on LinkedIn most weeks about what is changing in web work now that AI does a share of it. Based near Munich, writing in English for a global audience.
Roles
PipedriveHead of WebHead of Web2021 - now
VeeberMediaFounderFounder2015 - now
Web UnpackedHostHostongoingNear Munich, Germany · writing in English, globally
Where the entries come from
The house rule is that nothing is published until it has run on a real project with real visitors. These are the projects.
The web org at Pipedriveday jobA marketing site at global SaaS scale, where a bad call costs real pipeline. Migrations, performance work, and how a web team is run come from here.
Four sites of my ownNext.js · VercelAll four run on the stack this library recommends, and I moved every one of them there myself. When an entry says a tool holds up, it is holding up on these.
SaaS-Welt
Selge
Template-WeltWeb growth audits and a small number of advisor engagements. Nothing sharpens a checklist like reviewing a site you did not build.
Web UnpackedpodcastConversations with the people running web at other companies. Half the team stacks on this site start as a question I asked on the show.
Writes about
Published on The AI Web
6 postsPlus every one of the 68 entries in the library, which he writes and tests himself.
Six things that make a website readable by AI assistants
What ChatGPT, Claude and Perplexity actually fetch from a page, and the six things that decide whether they can use it: the crawler rules, HTML that carries the content, headings that answer, a summary and takeaways, a real FAQ, and structured data with an author behind it.
Do you still need a CMS when the assistant edits files?
It depends on who edits and how often. Alone, I run four sites on files in a repository and the assistant is the editor. With a team, a CMS earns its place: not every typo should have to go through an AI tool.
Can I let Claude or ChatGPT edit my website? What works today
Webflow, Framer, Shopify and Lovable say yes. Squarespace says no. Every website builder, CMS and host in the library, checked against the vendor's own pages, with what your assistant can actually change.
How to move a website off Webflow to Claude Code, GitHub and Vercel
I moved four of my own sites off Webflow this spring, the biggest at about 200 pages in five evenings. The steps, what the assistant did on its own, what it skipped, and who should stay where they are.
What goes in llms.txt, and does anything read it?
A plain-text index of your site written for AI assistants. What the format asks for, what this site puts in its own, and the measured evidence that almost nothing fetches it.
“Give it more context” means three different things
The most common piece of AI advice is right and too vague to act on. Splitting context into how you want the work done, what is true today, and what stays true is what makes it usable.
How the shelf works
What goes in, what stays out, and why the whole thing is free.
Start with the stack
The tools he picks when a site has to ship, with a verdict on each one.