There are eight episodes of Produktland out now.
I end every single one with the same question: what should product managers try on Monday, and what should they leave alone?
Here are all eight answers in one place. Plus the things the guests recommended along the way that are worth opening.
01 · Hurry slowly Natalia Dornonville de la Cour · Novo Nordisk
“Spend the energy early in the life cycle, on discovery and on validating, instead of automating a workflow nobody knows is the right one. A bad workflow with AI on top is still a bad workflow.”
Don’t: turn AI into a pure efficiency exercise. It’s a qualification job, not an efficiency job. That distinction is her sharpest, and it sorts most AI initiatives in an instant.
02 · Build the prototype in a single HTML file Kasper Juunge · computerlove.tech
“Open Claude Cowork. Take a concrete problem you have as a product manager. Ask it to build a prototype in a single HTML file. Send it to a user before Friday.”
Not a repo. Not a stack. One file.
Don’t: stop talking to users just because you can now build everything yourself.
03 · Run the feature factory test Ulrik Plejdrup Christensen · ADRÆT
“If your development team runs out of backlog because AI cleared it, you don’t have an AI problem. You have a feature factory.”
It’s the fastest organizational diagnosis in the entire podcast. A team that runs dry is fulfilling orders instead of solving business problems. AI just exposed it.
Ulrik also has the best name for a trap I’ve heard all year: the glitter paper problem. Pretty AI output makes us lower our critical guard. It looks right, so we don’t check it.
04 · Play guitar Dennis Green-Lieber · Propane
“Find two hours to experiment with AI tools, without it having to be useful for anything. Cursor, Codex, the terminal. Doesn’t matter. You can’t know what works for you until you’ve tried it yourself.”
Don’t: take other people’s TikTok snippets at face value. And don’t let your organization decide what you get to learn.
05 · Ask, damn it Bjarne Tveskov · former LEGO designer
“If you’re getting bad results from your LLM, you usually haven’t prompted well enough, so tell it that, and ask again. Ask it to find your blind spots. Use it as devil’s advocate.”
My favorite tip of the whole season, and it takes zero setup. Bjarne also delivers the line I’ve quoted most since: “You’re 90% done in no time, and then all you’re missing is the last 90%.” AI always adds. It can generate, but it can’t reduce.
Don’t: make ten prototypes on Monday morning just because you can. Make one good one instead.
06 · Let the model interview you Mikkel Malmberg · indie maker
“Talk to it. Ask it questions. Use interview me about this change before it writes code.”
It’s a concrete prompting move: instead of describing the task and hoping, you ask the agent to grill you until it actually gets it. Mikkel also uses skills like Grill Me and ask questions if underspecified to make it automatic.
Don’t: hand your developers a “finished” solution instead of a prototype of the interaction. And don’t think you’re done learning because it feels easy. His advice for anyone getting started: build the first 20 crappy programs. Getting started is easy. Getting good is hard.
07 · Care about the organization, not the tool Benjamin Christian Munk · Munk AI
“Ask about ownership, operations, support, version control, security and access before anything vibe-coded goes to production.”
Benjamin’s numbers say it all: 84% of executive teams say AI strategy is important. 5% have one. And Copilot for everyone is not a competitive advantage. That’s personal AI, which your competitors can buy tomorrow. Structural AI is the kind that changes how you work.
Don’t: hire vibe coders without having the guardrails in place. They can succeed, but not alone.
08 · Give someone the task, and no other tasks Christian Pedersen · WasteHero
“Put one or two dedicated people on exploring only this, phrased loosely, the way you’d brief an agent: here’s what I want to achieve, find the path yourself. Pick them based on appetite, not title.”
Within a day, they’d found the first problem. At WasteHero it was the product documentation, the agent couldn’t read what lived in slides and in people’s heads, so they moved it to markdown in GitHub.
Don’t: stack AI on top of everything people already do. If the job still requires roadmap, stakeholders and backlog “oh and also AI”, nothing happens. Take something away, and make the room.
📚 Things from the episodes worth opening
The guests recommended plenty along the way. The best of it:
Marty Cagan, Inspired · Kasper’s recommendation, and he was emphatic: read it. Twice, preferably. Christian points to Transformed for anyone about to start pilot teams.
State of PM 2026 · Dennis’ team has produced a report on the state of product management. Open source and free.
lovetobuild.net · Bjarne’s blog. He has deliberately left the platform silos to get back to his own voice, and you can feel it. Start there if you want to understand “I don’t want to read anything you didn’t bother to write”.
Verbos Podcast · Kasper’s own podcast with Jonas. If you want the developer’s angle on all of it.
👉 Your tip this week
Take Bjarne’s. It’s free, it takes zero setup, and you can use it in the next prompt you write.
Next time your AI hands you something mediocre, don’t rewrite the prompt in silence. Say it out loud, to the model. Ask it to find your blind spots. Ask it to play devil’s advocate.
Ask, damn it.
👉 Listen to Produktland · the 18 pieces of advice, distilled
More episodes coming. 🙏🏻
Best, Martin