About RiddleCrypt
RiddleCrypt is an independent, ad-supported logic puzzle platform. The format is a dungeon adventure: each room contains a single puzzle, solving it opens the door to the next room, and three wrong answers end the run. Behind that game framing is a serious project — a thoughtfully designed daily-cognitive-exercise tool with roughly two thousand puzzles spanning four categories and one hundred difficulty tiers.
Why a dungeon?
The dungeon framing solves a problem that pure puzzle apps don't. A typical Sudoku or crossword app presents each puzzle as a standalone unit — solve, get the score, move on. There is no narrative thread to pull the player from one puzzle to the next, which is why puzzle-app retention curves are notoriously flat. The dungeon mechanic creates a tiny stake: each puzzle is a doorway, and failing means the dungeon seals. That small stake produces meaningful engagement without resorting to manipulative streak-shame or notification spam.
The visual and verbal language of the site — torches, scrolls, the gold-on-parchment palette, the IM Fell English typeface borrowed from seventeenth-century engravings — is part of the same design choice. We want the experience of solving puzzles to feel like the experience of exploring a place, not the experience of filling in a spreadsheet.
Our mission
RiddleCrypt exists to make daily mental exercise both engaging and accessible. The accumulating evidence in cognitive psychology is that adults who engage in regular cognitively demanding hobbies — puzzles, board games, music, language learning — show measurably slower rates of cognitive decline in later life. The effect is modest but real. A few minutes of focused puzzle-solving each day is among the cheapest insurance policies you can take out on your own future thinking.
The format — short rooms, immediate feedback, escalating difficulty — is designed to fit any schedule. The early rooms are intentionally gentle so that beginners can build confidence. The deep rooms test the experienced solver. The Daily Riddle is a single fresh puzzle for players who want a low-commitment ritual rather than a long session.
How the puzzles are made
The RiddleCrypt library currently contains nearly two thousand original puzzles, spanning four categories — riddle, math, pattern, and word — across one hundred difficulty tiers. New puzzles are added in monthly batches.
The puzzles are generated using current large-language-model technology (specifically Claude by Anthropic), then curated, deduplicated, and indexed by difficulty before joining the rotation. The unique daily challenge is generated fresh at midnight UTC each day, so every visitor sees the same daily puzzle regardless of when they arrive that day.
We are transparent about this generation pipeline because we believe transparency matters. The alternative — pretending every puzzle was hand-crafted by a human writer — would be a lie that no thoughtful player would believe at scale. The honest position is that LLMs are an excellent tool for generating large quantities of varied short-form puzzles, that this tool requires human curation to be useful, and that the player's experience is what determines whether the puzzles are any good.
Editorial principles for AI-generated content
Generating puzzles via LLM raises real editorial questions. Our working principles:
- Quality screening, not raw output. Every puzzle is checked for clarity, solvability, and uniqueness before joining the rotation. Puzzles that fail screening are discarded or rewritten.
- Difficulty calibration is empirical, not declared. Each puzzle's nominal difficulty is adjusted based on how players actually perform on it. A puzzle the LLM rated "hard" but every player solves becomes an easy-tier puzzle. Conversely.
- Player flagging removes broken puzzles fast. Players who encounter a puzzle that is ambiguous, has multiple valid answers, or contains a factual error can flag it. Flagged puzzles are pulled from rotation pending review.
- No claims of human authorship. We will not pretend our puzzles were "hand-crafted by puzzle masters" because they were not. They are AI-generated, human-curated. That is a defensible position; the alternative would be marketing dishonesty.
- Editorial articles are human-written. The long-form pieces on our Articles page are written or substantially edited by a human, not generated wholesale. The puzzles are a different category — the article-as-cognitive-product requires a level of editorial judgment that LLMs do not yet reliably provide.
Our commitment to privacy
RiddleCrypt is designed to be played anonymously. No account, email address, or personal information is required to play. We do not use Google Analytics, Facebook Pixel, Mixpanel, Heap, or any other third-party behavioural-tracking system. We do not fingerprint your device. We do not sell or share any data with anyone — there is, frankly, almost no data to sell.
The only browser storage we use is the player's anonymous identifier (a random UUID, generated in your browser) and your current game state. Both are stored locally for continuity between sessions. Clearing your browser data ends your run and assigns you a new anonymous ID; you start fresh.
If you purchase Premium, our backend stores a Stripe customer ID and a transaction record. Card details never touch our servers — Stripe handles all card processing. Cryptocurrency purchases go through NOWPayments. Full data practices are documented in the Privacy Policy.
The advertising on this site is served by Google AdSense, which does set its own cookies. Those cookies are described in the Privacy Policy, and visitors in the EU and UK see Google's standard consent banner before any advertising cookies are set.
How RiddleCrypt is sustained
The platform is supported by display advertising (Google AdSense) and optional Premium subscriptions. The free tier — anonymous play, full puzzle access, Daily Riddle, public leaderboard — exists as a permanent first-class experience, not as a downgraded version of Premium. Premium players get an ad-free experience and unlimited hints, but the puzzles themselves are identical.
Premium pricing is intentionally low: $2.99 per month or $19.99 per year. Cryptocurrency is accepted for the annual plan. We chose this pricing because we want Premium to be a small thank-you from players who enjoy the site, not a paywall that locks anyone out. See the Pricing page for plan details.
What we don't do
It is sometimes easier to explain a project by what it refuses to be:
- We don't run engagement-baiting daily-streak notifications.
- We don't gate content behind email signup walls.
- We don't sell your data because we don't collect data worth selling.
- We don't artificially make puzzles harder to drive purchases of "premium hints."
- We don't run pop-up ads, autoplay video, or interstitials that block content.
- We don't operate on a "freemium with frustration" model. The free experience is the real experience.
Editorial and quality standards
All puzzles undergo automated quality screening before joining the active rotation. Difficulty is calibrated empirically by tracking player success rates, then adjusted to keep the progression curve smooth. Puzzles flagged by players as broken, unclear, or duplicative are pulled pending review.
The editorial articles on the Articles page are written for thoughtful adult readers. We assume our readers know history, have read books, and can handle a sentence longer than ten words. The puzzles, by contrast, are designed for the broadest possible audience — anyone who can read English at a tenth-grade level.
Acknowledgements
The fonts on this site are Cinzel by Natanael Gama (for headings) and IM Fell English by Igino Marini (for body text), both released under the SIL Open Font License. The infrastructure runs on FastAPI, PostgreSQL, Redis, and Cloudflare. The puzzles were generated using Claude Haiku 4.5. The dungeon-crawl conceit is borrowed, with affection, from a long tradition of text adventures going back to Will Crowther's Colossal Cave Adventure (1976).
Contact
For feedback, bug reports, broken-puzzle flags, partnership inquiries, or press requests:
Response time is typically within two business days. We read everything that arrives.