Editorial Policy
RiddleCrypt publishes two kinds of content: puzzles (the interactive game) and articles (the long-form writing at /articles/). They are produced through different processes, with different roles for AI assistance and human editorial judgment. This page explains both, plus how we handle corrections.
How our articles are written
The long-form articles on this site are drafted with the assistance of large language models — specifically, Claude by Anthropic — and then substantially edited by a human editor before publication. The editorial pass is not cosmetic; it involves checking facts against authoritative sources, rewriting sentences for voice, cutting filler, adding specific examples, and verifying that any cited dates, names, and quotations are accurate.
We make this disclosure for two reasons. First, because misleading readers about authorship is dishonest. Second, because the alternative — pretending every article was hand-written from scratch — is a posture no thoughtful reader believes at scale. The honest position is that LLMs are a useful drafting tool when paired with rigorous human editing, and we have chosen to use them that way.
How our puzzles are made
The puzzles in the game itself are generated in batches by Claude Haiku 4.5, then deduplicated and indexed by difficulty before joining the active rotation. Each puzzle's nominal difficulty is recalibrated empirically based on how players actually perform on it — so a puzzle the model rated "hard" but every player solves becomes an easy-tier puzzle, and vice versa.
Players can flag puzzles that are unclear, have multiple valid answers, or contain factual errors. Flagged puzzles are pulled from rotation pending review.
Fact-checking standards
For article content, we hold ourselves to the following standards:
- Dates and names are verified against at least one reliable secondary source (Encyclopedia Britannica, Stanford Encyclopedia of Philosophy, established academic publications, or the official source where applicable).
- Quotations are checked against the original text or a primary archival source. We do not include quotations we cannot trace.
- Scientific claims (for example, the neuroscience cited in our article on pattern recognition) are sourced from peer-reviewed publications. Where research is contested or preliminary, we say so explicitly rather than presenting it as settled.
- Statistics include their date and source where material to the claim.
Despite these safeguards, errors will sometimes slip through. When they do, we want to fix them quickly.
Corrections
If you spot a factual error, a misattributed quote, a wrong date, a broken link, or any other accuracy problem in an article, please email playsupport123@pm.me with:
- The article URL
- The specific passage in question
- What you believe is correct, and your source if you have one
We will read every correction email, verify the claim, and update the article (with a visible correction note) if the error is confirmed. Response time is typically within two business days.
Independence and conflicts of interest
RiddleCrypt is an independent project, not part of any larger publication or media group. We do not accept sponsored content, paid placements, affiliate revenue from product links, or "guest posts" from third parties. Our only revenue sources are display advertising (Google AdSense) and optional Premium subscriptions, neither of which influence what we choose to write about or how we write about it.
If we ever recommend a third-party book, app, or service in an article — for example, the further-reading suggestions at the bottom of some articles — that recommendation is independent and unpaid. We have no commercial relationship with the publishers of those works.
Use of generative AI: what we will and will not do
We will:
- Use LLMs to draft article structure, generate first-pass prose, brainstorm topic angles, and propose puzzle ideas.
- Edit those drafts substantially before publication.
- Disclose this use openly, here and on our About page.
We will not:
- Publish AI output without a human editorial pass.
- Claim any article was "hand-written by a human writer" if it was AI-drafted.
- Generate fake author identities or fictional editorial staff to imply scale we do not have.
- Generate fabricated quotations attributed to real people.
- Use AI-generated images and present them as photographs without disclosure.
This policy will evolve as the technology and norms around AI-assisted writing continue to develop. The current version was last updated on 3 June 2026.
Contact
Editorial questions, corrections, feedback, partnership inquiries: