The thesis
Most AI-literacy material teaches topics — what a token is, what RAG stands for. Perpenda teaches trade-offs. Every unit lands on a decision a product manager actually has to make: where latency is worth its cost, when retrieval beats a bigger prompt, what a rubric should reward, when "good enough" really is. You read a tight unit, you commit to a call, and your reasoning is graded against rubric criteria — not a multiple-choice key.
There are no streaks, no badges, and no daily nudge. The reward for finishing a unit is being able to make the next decision faster and with less hand-waving.
The path
The units form one curated order — LLM Systems for PMs — built to reach twenty units. Fifteen are published today; the final five are shaped during the closed beta, in response to where readers actually get stuck, rather than planned in full in advance.
Each claim is sourced and tagged by calibration tier — settled, contested, or unsettled — so you always know how much weight a fact can bear. Sources sit after the decision prompt, on purpose: consensus shouldn't prime your answer before you've reasoned to it.
Who it's for
Product managers shipping AI features — and anyone who has to defend an AI build / buy / skip decision to people who will ask hard questions about it. It assumes you're fluent in product work and want to get fluent in the systems, not the other way around.
Who makes it
Perpenda is an independent product, built in Lisbon, Portugal, and released under the GPL-3.0 licence — the app and this site are open source. It isn't backed by a platform vendor or a model provider, which is why a unit can say a tool isn't worth its cost when it isn't.
Questions, corrections, or want a calibration tier challenged? Email hello@perpenda.com — every claim is meant to be contestable.
The details
- Platform
- Android. In closed testing on Google Play; a signed APK is also on GitHub Releases.
- Path
- LLM Systems for PMs — fifteen of twenty units published.
- Grading
- Calibrated against rubric criteria, not multiple choice.
- Licence
- GPL-3.0, source on GitHub.
- Built in
- Lisbon, Portugal.
- Contact
- hello@perpenda.com