Why Daily Puzzles Work (and Why Most Don’t)

The thing Wordle actually invented

The internet is full of analysis of why Wordle took off. Most of it focuses on the green squares — the clever spoiler-free share format that flooded Twitter in late 2021. That was important, but it wasn’t the innovation. Hangman has been around for a century. Anagram games are older than that. The green squares were a new delivery mechanism, not a new game.

What Wordle invented, as a product, was a single daily constraint done well. One puzzle a day. No grind. No ads. No “just one more.” When it was done, it was done, and you came back tomorrow. In a medium where every other game is designed to keep you engaged for hours, Wordle was designed to end.

That constraint — “the game has to end” — is load-bearing. It forces every other decision.

What “has to end” forces

• The session must be short. If it ends, it has to end in a time you’re willing to commit every day. Five minutes is the upper bound for most people. Wordle averages two.
• There’s no progression system. No levels, no XP, no unlocks. Progression is a tool to pull players through content; daily games don’t pull, they pulse.
• Difficulty is fixed, not scaled. Everyone plays the same puzzle. The only comparison is whether you solved it and how efficiently. This is what makes “social proof” sharing work — your row of green squares is meaningful because your friend faced the exact same puzzle.
• The game has to be good for a stranger. No onboarding arc, no tutorial levels. Day one of playing is the same puzzle as day 500.

Why most daily puzzles fail

In the two years after Wordle, hundreds of daily puzzle clones launched. Almost all of them died. The common failure patterns:

• The puzzle isn’t self-contained. It relies on remembering yesterday’s, or progresses across days. This fights the “it ends” principle — new players can’t start today.
• The session is too long. Fifteen-minute puzzles don’t fit into a morning routine. They compete with full games, and lose.
• There’s nothing to share. Or the share text is spoilery. The spoiler-free share isn’t optional — it’s how daily puzzles spread.
• Difficulty is unfair. Procedural generators that don’t control for difficulty produce occasional brutal days, and a brutal day breaks the habit. Returning users punish inconsistency more than they reward peak moments.
• Ad density is wrong. A daily ritual that feels like it’s trying to monetize you stops being a ritual. Ads belong at transitions, not during play.

What we stole

Looking at what does work, we kept:

• Midnight drops. One puzzle set per day, live at midnight local time. No back-catalog surfacing in the main flow.
• Everyone plays the same puzzle. Deterministic generation keyed on date. Your Tuesday and my Tuesday are identical.
• Spoiler-free share text. The share emits a row of star emoji representing your results, with no information about hole positions.
• No account. Progress lives in localStorage. Zero friction to start. (The Expert leaderboard is the one exception, and it’s optional.)
• A streak. Not for pressure — for the subtle satisfaction of the counter incrementing. Streak is a small thing, shown small.

What we rejected

Some standard daily-puzzle conventions we decided against:

• Only one puzzle. Wordle ships one. We ship three, at three difficulties. We think the gradient is valuable — warmup, thinking, stretch — and the total still fits in five minutes. This is a bet. It could be wrong.
• Fixed-size grid. Easy is 4×4 and the others are 6×6. One grid size would have been simpler, but the difficulty ramp works better with a small board for easy.
• No timer in the main flow. Timers make puzzles stressful, and stress kills rituals. The Expert leaderboard mode is timed, but it’s opt-in and separate.
• No life system, no hints that cost something. If you want to look at yesterday’s solution, you can. We trust players not to cheat themselves.

The difficulty question

The hardest design question for a generated daily puzzle game is: how do you guarantee difficulty? Hand-authored puzzles (crosswords, Connections) are easy — a human curates. Procedurally generated puzzles (Daily Unfold) are hard, because random generation produces wildly varying difficulty even within the same parameters.

Our approach is a scored pipeline: generate thousands of candidates, rate each for difficulty using a scoring function built from symmetry, fold complexity, overlap, and a few other signals, and keep only the ones that land in the intended difficulty band. The long version is over here.

It’s not perfect. Some days the “hard” is a six-star smack-in-the-face, and some days it’s a brisk warmup. We’d rather err on the side of “approachable but sometimes too easy” than “occasionally impossible.” The latter breaks habits. The former just makes you feel smart.

Closing

Daily puzzle design, more than most genres, rewards restraint. Every feature you add is pulling against the single most valuable property of the format — that it ends. If you can’t justify a feature against that, it’s probably a net negative.

The best thing a daily puzzle can do is be boring to describe. “You play a puzzle. Then you’re done. Come back tomorrow.” That’s the product. Everything else is variance.