This is the part of the idea engine that runs on dice, blocks, and bad ideas I can't let go of. I keep a single file of game concepts — somewhere around twenty-eight of them — and like everything else I make, they're ranked: a "Best ideas" list at the top, an "Other ideas" pile underneath, and the discipline of that sort is honestly more useful than any single idea on it. The point isn't that I have a lot of these. The point is I know which two are worth a year of my one life, and I'm ruthless about the rest staying names in a file.
A note on how I think before you scroll. The very first line of that file, in shouting caps, is not a mechanic:
“GO THUMBNAIL FIRST LIKE YT — make a steam page and an appealing visual.”— the top of my game-ideas file
I package before I build. A Steam page is a YouTube thumbnail is a book cover — it's the promise, and if I can't make the promise compelling in one frame, the game underneath probably isn't either. So most of these come with a pitch, a hook, sometimes a brand, before they have a line of code. The craft side of a few of these — the why-they're-fun design essays — lives on Games & the Table. This page is the shelf. If one of them is your kind of system to take apart, grab it — the best designs I've got started as an argument with somebody.
The two I'd build tomorrow
These are at the top of my own "Best ideas" list, and they're the ones with real recon, real branding, and real fun I can already feel.
- Kuzure (★ flagship) — Jenga, inverted. An adversarial tower-stacking game where you place blocks to sabotage the next player, the tower persists across rounds as a piece of "brutal architecture," and nobody sits out — play continues from the corpse of the last tower. Full design doc: block-lives, a 30-point competitive cap, color-coded pieces, ash-black-and-slate-gray branding, a rename after a naming collision, and a Kickstarter-after-demo go-to-market.
- Balatro, but with dice (★ flagship) — a roguelite scoring engine where you can't control the roll. Borrows Kingdom Come 2's weighted-dice probabilities (I worked through the whole die table, because the EV is the game) and bolts on Balatro's build-an-engine structure. Recon's done — the existing Farkle roguelites are rough; this is a polish gap I think I can win. Steam-first, web prototype first.
The video hook for the first one writes itself, and it's the same move that starts half my projects — a complaint:
“Jenga is the worst game ever made — so I fixed it.”— the Kuzure pitch
Funny on purpose
A whole cluster of these exist because they'd be funny — on a short, on a stream, or at the expense of the player (sometimes me). The cynical ones are deliberate; the warm ones are too.
- Gambler's Dream — All Fallacies Come True (★ standout) — a casino where every gambler's-fallacy "street rule" is secretly, mechanically true. Hot streaks are real. The machine that "feels due" is due. My own honest note says the goal is to "really get gamblers addicted" — which is exactly why it'd make a darkly great game about the lie, not a real one.
- GoonsAI / Fake Twitch Chatters (★ standout) — "I had 0 viewers, so I coded my own entire Twitch chat." A local-LLM simulation of a full live chat: emote spammers, storytime guys, paid shills hyping my own apps, haters with wholesome counterbalancers, VIPs, and a TTS bot whose job is to ragebait me. Equal parts game, streaming tool, and content engine — about as "me" as an artifact gets.
- Poker with readable tells — learn to read a face, not a hand. Each AI opponent makes different idle expressions on different hands, and only some of them correlate to a bluff — so you're learning each character's lying pattern, not a universal one. Needs to look funny on shorts; that's a feature, not an afterthought.
- Magic-the-Noah, with live animated subtitles — cheap live transcription (~1¢/hr) slapped onto player avatars to personify them instantly. Could be a web server, a Repo/Lethal-Company-style party game, or a mod big creators play into a growth loop.
- Tree-hate game — a fun-first, gameplay-led comedy game, also on my own "Best ideas" list. Mostly a vibe and a name so far, and I'm keeping it that way until it earns more.
I'll cop to the cynical streak being real, and to it being a range, not a flaw I'm hiding:
“really get gamblers addicted.”— my own note on Gambler's Dream
Multiplayer, with the tension turned up
The thread here is taking a genre I love and adding the one mechanic that makes it scary — real stakes, real loss, real reading of another human across the lobby.
- Sea of Thieves, but Hunt: Showdown — make piracy scary and loot real. Extraction-shooter bones on a pirate sandbox: large lobbies, location-leak tension, a boss area, and loot you actually keep only if you make it out.
- Hitman, 1v1 — kill their two targets while defending yours. Escape routes exist but keep the focus on the objectives, not the getaway — a duel of patience and disguise.
- Local stealth co-op — Dishonored or Prey, but with a friend at the controller. Mostly a named stub for now — a co-op spin on immersive-sim stealth, nothing fleshed out yet.
- Multiplayer stealth — asymmetric assassins-versus-guards, the genre taken seriously (team sizes still open — somewhere around 2v5 or 5v5). The mechanic I can't stop thinking about: you fool other players with afterimages and mimicked sounds instead of dropping coins — straight out of Hunt: Showdown's sound design.
- Stickman fighting co-op — Nidhogg and Stick Fight, but the hits land. Punchy, skill-based, deep — graphic, satisfying contact instead of floaty pokes.
Reskins, remixes, and "what this could be"
A lot of my ideas are a love letter to an existing game plus the fix it never got. These are the "make X but actually good" pile.
- PC Mansions of Madness — what the tabletop horror game could be online: better mechanics, a community story-builder, and a social-deduction layer where players quietly go insane.
- Card game, dark ambience — Slay the Spire's structure reskinned into Inscryption / Darkest Dungeon / Arkham Horror dread — serious, oppressive, atmospheric.
- Physics-engine vehicle builder — Besiege-meets-Motor Town: physics accessible early and deep late, with online races, battlebots, and budget-limited builds (an RC framing helps suspend disbelief on the suspension).
- Arkham Horror, reskinned as a video game — the cooperative cosmic-horror loop, rebuilt natively for screens instead of cardboard.
- DC20 as a lightweight digital combat aid — a real reason to test the DC20 ruleset and a menacing, AI-art battle-map you cast to a TV instead of hiding behind a DM screen. The D&D-tooling instinct from the table, pointed at a new system.
- Valorant/CSGO, multiplatform and lightweight — the tac-shooter loop, optimized to run anywhere.
- Low-budget Indian driving simulator — exactly what it says, and exactly the kind of specific, regional, slightly-deranged idea I love.
Detective, deduction, and the deep end
The cerebral corner — puzzles that respect the player's brain, and one idea that's almost a dare.
- CAT Einstein-Questions Detective Game — CAT-exam-style logic puzzles, hand-drawn. You get drawing, labeling, and arrow tools to solve each case; harder levels strip the tools away and make you hold it all in your head.
- Horror game where the ghost moves in accurate 4D — the monster doesn't obey three dimensions. A horror premise built on a genuinely disorienting spatial mechanic — the kind of idea that's either brilliant or unplayable, and I want to find out which.
- Finish Karlson / Rift Genesis — a smooth-movement, Titanfall-flavored 5v5 ranked shooter; one I've actually started rather than just named.
The honest part
I should be straight about the shape of this shelf, because that's the actual character of it. Two of these I'd defend with a year of my life; most of the rest will stay names in a file, and I know that going in. By my own repeated admission the bottleneck has never been finding the idea — it's finishing it:
“The actual bottleneck is finishing, not picking. I'm fast at finding the right thing and slow at completing it.”— from my notes
So I don't present the list as a brag about volume. I present it as a working pipeline with a known flaw, where the ranking is the real skill — and where the failure mode is loneliness, not a lack of ideas. The truest design lesson I've learned is that I finish things better with someone across the table. That's why this whole site is wired as a collaboration funnel and not a portfolio.
- Ranked, not piled — a "Best ideas" tier over an "Other ideas" pile; the sort is the discipline, the volume is a side effect
- Packaged before built — "thumbnail first": most carry a pitch, a hook, or a brand before a line of code
- Recon before love — the dice roguelite came with a survey of the competition; Kuzure came with a rename after real feedback
- Range on purpose — civic-curious to cheerfully cynical, party-funny to genuinely cerebral, screen to cardboard
Grab one off the shelf
If any of these is the system you've always wanted to take apart — the inverted Jenga, the dice roguelite, the readable-poker AI, the 4D ghost, or one I haven't even written down yet — that's the whole reason this page is public. I'm faster, happier, and far more likely to finish when I'm building with a curious person rather than ranking ideas alone in a file.
So: pick one, or bring your own complaint about a game that should've been better. The two I'd actually ship next are the dice roguelite and getting Kuzure to a Kickstarter — but the door's open on the rest, and the best ones always start as an argument. Say hi.