Waldo Vanderhaeghen
I'm Waldo Vanderhaeghen, a product manager in Berlin — writing and building at vanderlore.de. Day job is Senior PM at Kleinanzeigen, Germany's largest classifieds marketplace.
The part of product work I keep coming back to is coordination. You have smart people with different context about what's real and what matters: engineers, analysts, designers, stakeholders. Getting them to move together on something that actually ships takes more deliberate effort than any roadmap document captures. I've spent a lot of my career on that part. It's one of the more interesting problems in the work, and genuinely one of the more enjoyable ones.
Outside the day job, I build things. I tackle challenges for friends, projects or businesses that I am involved with. These things start because I get curious about whether they're possible. Most of them are still quietly running somewhere.
I'm also genuinely interested in where AI tooling works versus where it just looks like it works. The gap is smaller than skeptics say and bigger than the demos suggest. Finding the right place for a model, the right place for a hard rule, the right way to know which is which: that's most of what I think about when building. A lot of what's on this site is trying to figure that out in practice.
I write here because it forces me to understand what I did. Explaining something properly is how I know whether I actually understood it. And occasionally it saves someone else a few hours, which feels worth the time too.
On the side, I do product and AI consulting with a few small teams. Currently working with Faust AI and Glastrix.
Common questions
- What do you do at Kleinanzeigen?
- Senior PM for the Retail Media product. Took from inception to launch: a two-sided advertising marketplace where businesses configure their campaigns, and we optimize which ads to surface to Kleinanzeigen users through search and relevancy signals.
- How do you use AI in your work?
- Claude Code is my main tool — I use it for everything from building full applications to automating reporting pipelines. My approach is spec-driven: every task gets a ticket with acceptance criteria before the model touches any code. It prevents drift and keeps outputs auditable. I've written about how I run my agent team and spec-driven development if you want the detail.
- Are you available for consulting?
- Yes, selectively. I work with small teams on product strategy, AI tooling architecture, and building production-grade automation. Current capacity is limited — email me with what you're working on.
- What's the difference between AI that works and AI that just looks like it works?
- Mostly: is the output auditable? Can a human check it? Demo AI produces plausible output in a frictionless flow. Production AI produces output that can be verified, overridden, and traced back to a decision. The hard part is building the harness, not prompting the model. I wrote about why the model doesn't decide.
- What have you built outside of work?
- The main ones: NfcPress — a white-label NFC business card platform for a Berlin print shop, running in production since 2026. A Smoobu-to-Sheets sync for managing short-term rental bookings. An AI invoice pipeline for processing German tax documents from Google Drive. Most are still running.
Get in touch
Got a project idea, a question, or just want to say hi — write me an email or use the form below.