HexLemma

Sudoku without guessing is a strict religion. Picross too. There's a small, stubborn community of puzzle players who treat any move that isn't provably correct as a kind of personal failure, and the games that serve them tend to come from one or two stubborn developers grinding away in obscurity. Geriatric Games is making one of those. It's called HexLemma, it lands in June 2026, and it's on Steam with a wishlist page already up.

The pitch in one line. HexLemma is a constraint-satisfaction deduction puzzle on a hex grid where every cell hides a value from 1 to 7, and you deduce them all using six families of clues that compose into proofs. No guessing. Wrong moves get blocked at the input layer before they hit the board.

That last bit matters more than it sounds.

The "no guessing" promise

Anyone who's played enough logic puzzles knows the dark secret of mass-market sudoku apps. A non-trivial chunk of them ship puzzles that require a guess somewhere in the middle. The generator can't prove a unique solve path, so it punts. Hardcore solvers smell this in about ninety seconds and quit forever.

HexLemma is staking the whole design on the opposite contract. Every commit is provably correct, every elimination provably necessary, and the game refuses to let you place a wrong value, which is a stronger claim than "we'll tell you if you mess up." It means the solver engine sits behind the input system in real time, vetting moves against the proof state. That's not a trivial thing to build, especially with six clue types interacting.

The six families do a lot of work:

• Pair-Same and Pair-Differ (=, ≠) on distant cells
• Order constraints between cell pairs
• Count (N·V), exactly N of a hex's six neighbors hold value V. The Minesweeper DNA.
• Parity, even or odd
• Vertex Sum, three cells meeting at a corner sum to a target

The interesting part isn't any single one of those. It's the composition. The marketing copy calls it an "emergent chain," which is overcooked, but the underlying idea is real. A Count clue narrows where a 3 can sit, a Pair-Differ kills one of those candidates, a Vertex Sum collapses what's left. You're building proof chains rather than running through a checklist. That's the thing genuine deduction-puzzle nerds will pay for.

The hex grid isn't cosmetic

Switching from a square to a hex grid sounds like a visual decision. It isn't. Each cell has six neighbors instead of four or eight, which fundamentally changes how Count and Vertex Sum behave. Vertex Sum only really makes sense on a hex board, where three cells genuinely meet at a single point. On a square grid you'd have four cells touching a corner, which scales the combinatorics differently and arguably less elegantly.

Eight board shapes ship at launch. Hex Disk, Rectangle, Triangle, Parallelogram, Hex Star, L-Shape, Donut, and organic blobs. The blobs are the one that raises an eyebrow, because irregular boundaries mean edge cells have fewer neighbors, which changes the difficulty of every Count clue near the perimeter. Whether the difficulty curve holds up across all eight shapes at all four difficulty tiers is the kind of thing that won't be obvious until people are grinding the Insane tier for a few weeks post-launch.

Campaign as vocabulary lesson

Twenty-five hand-tuned puzzles, five stages, each stage introduces one new clue family. Stage 5 is "Capstone" with full grammar on Insane. That's a deliberate teaching arc, the same shape Stephen Lavelle and Jonathan Blow lean on for their puzzle work. Don't dump the rules. Layer the vocabulary one piece at a time and let the player build intuition before composition gets ugly.

The Daily Challenge is one puzzle per UTC day, same for every player, Steam leaderboard attached. Deterministic seeds are standard play here. Random Puzzle mode lets you pick shape, difficulty, and seed, which is the part the solver-community types actually care about, because it means you can post an interesting seed in a forum and another player can reproduce the exact board.

No timer. No mistake cap. Cloud saves. Move replays for finished puzzles, which is small but genuinely useful for anyone who wants to study their own solve paths or share a particularly clean proof chain.

Skepticism, earned

Geriatric Games is a small studio. The trailer is a one-developer affair, posted under "Geriatric Millenial," which tells you about the scale of the operation. The localization list is ambitious for a solo or near-solo team. Twelve languages, including both Portuguese variants, plus Finnish and Norwegian, which most indie puzzle games skip. Colorblind themes for deuteranopia and tritanopia are listed under "Coming Soon" rather than at launch, which is more honest than pretending they're done.

Steam Deck support is listed as "partial" with controller scheme work ongoing. Hex-grid navigation on a stick is a real interface problem. Six-way cursor movement doesn't map cleanly to a d-pad, and the team is explicitly still tuning it. That's the right level of disclosure for a 2026 release.

The risk isn't the puzzle design. The risk is reach. A constraint-satisfaction hex deduction game with seven-value cells and six clue families is genuinely niche, and the Steam puzzle category is a slaughterhouse. Most thoughtful logic games drown inside their first week.

Who actually finds a game like this

A broad Steam-front-page push for HexLemma would be money set on fire. The audience isn't "puzzle fans." It's a specific overlap. Nikoli-style solvers, Simon Tatham's Portable Puzzle Collection diehards, the people who follow Cracking the Cryptic for sudoku variants, the Picross completionists who own every Jupiter release, and the slice of the Zachtronics crowd that wanted more proof and less optimization. Reaching them through generic ads is a losing proposition, because the signal sits inside small YouTube channels, niche subreddits, and the kind of streamer who'll spend ninety minutes silently working a board.

That's the surgical problem CreatorFetch exists to solve. Finding the deduction-puzzle YouTubers, the hex-grid math hobbyist streamers, the Steam Deck handheld reviewers willing to feature a contemplative title, the abstract-strategy creators whose audiences actually buy these games, instead of burning budget on viewers who'll bounce in thirty seconds. For a studio Geriatric Games' size, that's the difference between finding the two thousand people who'll love this thing and disappearing on launch day.

The genre has a long memory. Hexcells is still being played a decade after release. Simon Tatham's collection has been free and beloved since the early 2000s. If HexLemma's solver actually delivers on the no-guessing promise across all four difficulties and eight shapes, it'll find its people. The real question is whether enough of them hear about it before the algorithm moves on.

June 2026 is a long way off. There's time for the colorblind themes to land, for the Deck controls to settle, for the daily-challenge leaderboard to build the kind of small, obsessive community these games live or die by. Worth keeping an eye on if you're the sort of person who reads the words "Vertex Sum" and immediately wants to know how it interacts with parity.