Larry Hryb on What Makes a Good Game When You Are Just Starting Out. Most conversations about making games open with graphics, engines, and budgets. Wrong end of the problem. A good game starts with an idea a player gets in seconds. Chris Zukowski counted 20,282 games released on Steam last year. Six hundred and eight of them crossed a thousand user reviews, which works out to roughly three percent. The other ninety-seven percent didn't lose due to technology. The ones I looked at were usually unclear about what they were, and a good number of them were never really finished. About that "simple engine" idea There is a comfortable line going around that beginners should be grateful for limited tools, because constraints force better decisions. The Sandbox Studio makes that line obsolete, so let's set it down. Studio hands you a character controller that already walks and jumps, cameras in first and third person, and a physics engine doing collisions, stacking, and ragdolls. Input is handled across keyboard, mouse, gamepad, and touch, so the same build runs on a phone without device-specific code. The UI kit arrives stocked with health bars, minimaps, crosshairs, and score displays. Multiplayer is a switch you flip when you are ready; nobody is asking you to architect it in week one. And there is an AI assistant living inside the live editor, moving objects around in your scene while you watch. What actually helps a first project is the length of the loop. You build in a browser tab, press Play, and you are testing. Publishing produces a link that runs in any modern browser: no install, no store submission, no gatekeeper. Send that link to a friend in Lisbon and she is playing before you finish typing the next message. The speed cuts both ways. An assistant that will generate a crafting system before lunch makes it much easier to build nine systems nobody asked for. Speed did nothing for your judgment. Say it out loud in one sentence Your first project should be something you can describe to a friend in one breath. Open-world RPGs and hundred-page design documents come later, if ever. Tetris is arranging falling blocks. Pac-Man asks you to eat the dots and avoid the ghosts. Vampire Survivors: stay alive while the screen fills with monsters. Geometry Wars started in 2003 as a minigame hidden behind an arcade cabinet in the garage in Project Gotham Racing 2, built by Stephen Cakebread at Bizarre Creations. Twin sticks, one screen, shoot the shapes. Players spent more time in that garage than on the track, so Bizarre expanded it into Retro Evolved and shipped it on Xbox Live Arcade on November 22, 2005, alongside the 360. It went on to hold the record for the most downloaded game on the service. The sentence never got any longer. The recent breakouts are built the same way. Balatro is poker hands that break their own math, made by one person going by LocalThunk. PEAK asks you to climb a mountain with friends who will get you killed. And Meccha Chameleon, which turned up out of nowhere in June at $5.99, is hide-and-seek where you paint your own body to match the wall behind you. Depth comes later. Get the sentence first. Thirty seconds of footage should be enough for a stranger to know what they would be doing. If it is not, you have a design problem, and the player is not the variable you get to adjust. Build a goal. Put an obstacle in front of it. Make the payoff land when the player gets through. Run that loop until it starts to get interesting. Fun is a feature Players forgive simple graphics all day. Confusion is what loses them. PEAK shipped in June 2025 at $7.99, discounted to $4.95 for the first week. It sold 100,000 copies in a day, a million in six, and passed ten million by August. The core of it came together in about four weeks, built by seven people from Aggro Crab and Landfall who flew to Seoul, rented an apartment, and worked out of it. Total spend was under $200,000. Nick Kaman, who runs Aggro Crab, has been happy to tell anyone who asks that the code is a mess and that players file bug reports on it daily. Nobody bought it for the rendering. They bought it because ragdoll physics and proximity voice chat and a very tall mountain add up to stories, and the stories were funny. So the real question is whether somebody wants another round. One mechanic The most common way a first project dies is system sprawl. Crafting, inventory, classes, vehicles, pets, quests, leveling, lore, achievements, and three currencies, all built before anyone confirmed the thing is fun for ninety seconds. Make jumping feel great if the game is about jumping. If it is about exploring, make discovery worth the walk. Everything else is decoration on a core that either works or does not. Studio's AI assistant earns a specific warning here, because it will never tell you no. Ask for an inventory system and you will have one, wired into your scene, in a few minutes. The Studio documentation suggests pointing it at the tedium first: placing and arranging objects, setting repetitive properties, boilerplate logic. That is the right instinct. The assistant is there to remove typing. Deciding what your game is remains your job. The rule there was that if the audience changes the station, nothing else matters. Same rule applies here. A boring core loop cannot be patched with content. Write down three answers. What does the player do? Why do they care? And what brings them back tomorrow, once the novelty has worn off? If those come out clean, you are pointed somewhere useful. Where the stories come from Community belongs in the design conversation early, not in the marketing plan afterward. Games get stronger when players have something to hand each other. A score. A physics accident. A shortcut you never intended to leave in. Lethal Company gave players a radio, a monster, and no efficient way to help each other, leaving the panic to drive the gameplay. Content Warning went further and made the camera the entire game: film the horrible thing, upload the footage. In PEAK, proximity voice chat means the funniest moment of any session is somebody screaming as they fall past you. Geometry Wars had no multiplayer at all. What it had was a leaderboard with your friends' names on it, and that turned out to be enough to keep people coming back for months. This doesn't require complex engineering. It's about making early decisions on where the story lives. Studio puts this within reach on a first project. Multiplayer is a toggle instead of a rewrite, and publishing is a link you drop into a group chat. Build one twelve-second moment worth recording and you have given people a reason to bring friends. Feedback Players need to know when they succeeded. A sound, an animation, a number ticking up, a burst of particles, a bit of screen shake. Every action should connect to a result. Press the button, something happens. Plenty of games feel dead because feedback got scheduled for the polish pass, and then the polish pass got cut. Imagine a basketball game where the hoop does nothing when the ball goes through. The score increments in silence. Works fine. Feels like nothing. Studio's UI Kit gives you health bars, ammo counters, and score displays that already share a visual language, and the September update added sprite-sheet animation to the particle editor. Reach for them when a moment needs marking, and leave them alone the rest of the time. Sit on your hands You know how your game works because you built it. Nobody else has that advantage. Put someone in front of it and say nothing. Harder than it sounds. You will want to explain the controls. Do not. When three players make the same mistake, the game is teaching the wrong lesson. That one observation beats a week of art tweaks. Because Studio publishes to a browser link, a playtest is a message. Send it to five people tonight. Study the reason Inspiration is healthy. Cloning puts a ceiling on you. A platformer works because the movement is precise. Survival games work because the choices cost you something you actually wanted. With a puzzle game it is usually that each challenge quietly teaches the idea the next one is about to require. Balatro is a useful test case. Copy the pixel art and the poker theme and you have nothing worth playing. What the game is built around is the moment a player notices their joker combination has made the numbers ridiculous, and every system in it exists to deliver that moment again. Get out of the player's way The Xbox titles that lasted got people into the game quickly and paid off effort early, without a lecture on the way in. New developers keep building friction where none belongs: long intros, tutorials explaining what the player figured out in the first ten seconds, and a full menu system for a game with one button. Studio's template catalog exists partly to prevent that. Third Person Action, Side Scroller, FPS, Vehicle, Visual Novel, and as of this month, 3rd Person Platformer and First Person Exploration. There are finished examples in there too. Grindwave, Drift Chaser, The Still Sun. Play them, remix them, or take them apart for the systems underneath. Beginning from something already playable is a head start, and I have never met a player who asked how you got there. Prefabs, sooner than you think Build, test, improve, go again. That beats waiting for the perfect idea, and plenty of people spend a year planning a game they never start. Make something small this week instead. One piece of workflow advice from the Studio docs is worth taking literally: the second you place the same object a third time, turn it into a prefab. They are first-class now. Create one, save it, edit it in place, override individual instances without breaking the link back to the original. Retrofitting prefabs later is painful. Doing it early costs you sixty seconds. Every finished project teaches you something. The failures teach more, and they teach faster. Finishing A small finished game teaches more than a large unfinished one. Shipping forces decisions you would otherwise defer forever, and it swaps imagined solutions for real ones. This industry has never been short on ambitious concepts. Completed games are the rare part. Deadlines help, which is why jams work. Can I tell what I am supposed to do without anyone telling me? Does something satisfying happen when I get it right? And did you actually finish it? That last question may be the most telling. Do not measure your first game by what is under the hood. Measure whether a player understands the goal and wants another turn. If they come back tomorrow, you have something. If they come back the day after that, you may have a winner. The rest is version two.