September 2026 · Design notes

Clue ladders: designing a daily puzzle about real packages

Condary is a daily game about the packages that make up the Conda and Python ecosystems. Each puzzle is a ladder: five clues about one real package, from obscure origin story down to a mercy clue. The design constraint that shapes everything: the clues are about real packages, so the clues have to be fact-checked like real documentation.

Game Condary — Season 0 Team Anaconda Incubation Sources Distro-DB · project docs · maintainer history

The ladder

Five clues per package, ordered as a single information gradient — not category-themed, not five trivia facts in a row. The top clue is a deep-cut flash of knowledge: package archaeology that should feel impressive to know, not required. The bottom clue is a mercy: the line every tutorial starts with. The design goal is that the modal solve lands at clue two or three — enough of a climb to be satisfying, enough of a descent to never be humiliating.

An abridged example — pandas, Season 0:

Clue 1 · 1000 points

I began in 2008 as a skunkworks project at a quantitative hedge fund, so a researcher could reproduce econometric work after moving a front-office stack onto Python.

Clue 2 · 800

When two labeled objects do arithmetic, I take the union of their indexes and fill the holes with missing values rather than silently dropping labels.

Clue 5 · 200 (the mercy)

Most tutorials start with import mystery as pd. I aim to be the fundamental high-level building block for practical data analysis in Python.

Notice what clue 2 does: it's the alignment quirk — a behavior you'd know from using the library even if you've never read its history. A good ladder mixes three kinds of knowledge: origin story, behavioral signature, and everyday API. Clue 3 in the full ladder is literally a code snippet. If you use the thing, you can climb without ever knowing the lore.

Fact-checking the fun

Every ladder is drafted against records from Distro-DB — package summaries, upstream URLs, install names — and then verified against primary sources: official docs, project history pages, maintainer posts, PyPI metadata, GitHub advisories, and the occasional piece of conference lore. Each clue's notes record what was deliberately left out and why.

The pandas clue 1 again: the hedge-fund origin stays, but Wes McKinney's name, the firm, and the year are kept out of the sentence — people who know that story should still get to feel clever at 1000 points. And one widely repeated etymology, the "panel data" backronym, is not used, because it doesn't appear on the project's current official about page. Famous-but-unverified loses to boring-but-true.

The puzzles are fiction-adjacent, but the facts aren't. If a clue wouldn't survive being read by the package's maintainers, it doesn't ship.

Season 0 and the single-file constraint

Season 0 is twenty packages, one per day, advancing at midnight UTC. Like its sibling Quizaconda, the entire game ships as a single self-contained HTML file — no build step, no backend, no dependencies. The season's clue ladders live in a reviewed editorial document alongside the game, so every puzzle's provenance is inspectable.

Play it

Condary — Season 0 and Quizaconda are both playable now. A new puzzle every day at 00:00 UTC. How deep can you climb before the mercy clue?

Sources & notes

  1. Condary Season 0 clue ladders — reviewed editorial document with per-clue difficulty notes and sources (Incubation, 2026).
  2. Distro-DB project records: summaries, upstream PURLs, install names.
  3. Per-package primary sources: pandas project docs and maintainer pages, plus the official sources cited on each ladder.