AI Product Manager

Karo Zieminski Portfolio

AI product manager at a European company building physical AI products, who also builds indie products independently and writes "Product with Attitude," a newsletter on critical AI literacy reaching roughly 21,000 subscribers in 146 countries.

Physical AI ProductsIndie BuildingAI LiteracyEU AI Act

Location

Not listed — probably too busy shipping to update the bio.

Skills

Physical AI ProductsIndie BuildingAI LiteracyEU AI Act

What makes it work

A breakdown of the choices that make this portfolio stand out.

1

Combining day-job AI PM work with independently shipped side products

Building indie products outside her main role gives her a second, faster feedback loop for testing AI product ideas than a single employer's release cadence allows.

2

A named deep-dive on a real launch, not just her own work

Writing a detailed, sourced product deep-dive on how Anthropic shipped Cowork in roughly ten days — including specific technical details like sandboxing and security tradeoffs — demonstrates analytical rigor applied to someone else's decisions.

3

Framing regulatory literacy as part of the AI PM skill set

Explicitly working near EU AI Act compliance for physical AI products treats regulatory constraint as a core part of the job, not an afterthought.

What AI Product Managers can take from this

Specific, actionable tips to apply to your own portfolio — no generic advice.

  • If you build side projects outside your main job, mention them specifically — they demonstrate you can ship without a team's infrastructure behind you.

  • Publish a detailed, well-sourced teardown of someone else's real AI product launch; it's a low-cost way to demonstrate analytical rigor without your own case study.

  • If your work touches AI regulation or physical/hardware AI products, call that out specifically — it's a distinct and less common skill set.

Projects

Karo Zieminski names her day job (physical AI products in Europe) but doesn't show it. The portfolio is her side work: a 159,002-word experiment she ran and wrote up, and two small tools she built for the Substack writers who read her.

ResearchintermediatePublished August 27, 2026

Does Pangram Work on Substack? A 159,002-Word Experiment

Pangram AI detectorPangram APISubstackBlind testing

After Substack rolled out the Pangram AI detector in July 2026, she ran everything she wrote, said in meetings, or got back from AI tools through it for a month. She tested four authorship scenarios, from pure AI output to AI drafts edited five to seven times, through Substack's integration, Pangram's own tool and its API.

A clear, bounded test of a live AI feature, with its limits stated up front ('a deep single-author study, not a benchmark'). That is the evaluation discipline AI PMs need before trusting a model's output.

  • 38,911 input words and 120,091 AI output words tested

  • 79% of her own human writing flagged as AI

  • Frames the finding with Goodhart's law: writers adding typos to look 'more human'

Web AppintermediateLive in 2026

StackShelf

Web appSubstackLLM-discoverable profile pages

A marketplace for Substack writers to showcase their products and books. Each creator gets a 'Shelf-in-Bio' page in a bento-grid layout that pulls together products hosted anywhere (Gumroad, Amazon, Ko-fi). The pages are designed to be discoverable by ChatGPT and Perplexity.

She spotted a distribution gap in her own community and shipped a product for it, with a working payment model built on the newsletter.

  • Shelf-in-Bio page with a custom URL per creator

  • Bi-weekly featured creators, chosen from submissions

  • Affiliate 'handshake' badge for creators open to cross-promotion

ToolbeginnerLive in 2026

LinkSwap

Web appMagic-link authSubstack

A backlink exchange for newsletter writers. You post an article that needs links, other members volunteer by topic, each side places a link to the other's post, and both confirm. Nothing is posted automatically.

It shows tight scoping: one tedious workflow (finding, pitching and tracking link swaps) cut down to a volunteer board and a two-sided confirmation.

  • No bidding or negotiation: volunteering is the whole handshake

  • Title fetched automatically from a pasted post URL

  • Built for search engines and AI answer engines alike

Ready to build your portfolio?

Follow the AI Product Manager roadmap — skills, projects, and timeline to get hired.

AI Product Manager Roadmap