Mark Brown

He thinks by making a video about a problem. Brown says researching, writing and presenting a video organises his thoughts better than any other method, and he then asks viewers for comments or a poll.

He used this when hobbyists cloned Word Play, turning a confusing email into a video and a community vote on what to do.

Hobbyists cloned Word Play with Claude Code in days. Claude went from a finished spec to a playable core in 12 minutes, against seven months for Brown, and a list of 21 modifiers was added at once.

The pair built Words Upon Words in 11 days, added a timer mode, daily challenge, multiplayer and meta progression, and asked his permission before releasing. He asked them to hold off.

He accepts borrowing but sees this clone as different. Word Play itself blends Balatro, Bookworm Adventures, Scrabble, Slay the Spire and Vampire Survivors, but a mod needs the original bought and this clone does not.

He says video games are built from the bones of other video games. Fast cloning removes the pause where a copier might reconsider, and he jokes that 12 minutes leaves little time for introspection.

He admits the clone is better in places, and that it hurts. Brown had half-finished daily and speed-run modes in his Unity code but was too burnt out and in technical debt to finish them.

He thinks buyers do not care about a developer's hard life. He still expects many players to choose the original for its UI, sound and music, and he disagrees with the clone's extra complexity.

He doubts guardrails or simplicity advice will solve cloning. Anthropic added two sentences to Claude Code's prompt to avoid recreating specific games, but it came too late and prompts can be circumvented.

He pushes back on Valve's advice to avoid single easy-to-copy mechanics, pointing to Tetris and Wordle, and notes someone used AI to clone Photoshop and Premiere.

His second game, Word Play, shipped in seven months. It grew from a two-day Patreon-only game jam prototype in December, mixed Balatro with Scrabble and Bookworm Adventures, and went on to sell more than twice the copies of Mind Over Magnet.

He says the planning he learned on his three-year first game is what let him ship without burning out.

He treats prototypes and playtests as the only proof an idea works. His recurring line is that brains are bad video game simulators, so ideas only show their flaws once built.

He draws on Valve, which treats designs as hypotheses and playtests as experiments, and on designers who follow the fun by keeping the best part of a prototype and cutting the rest.

He sees difficulty as a toolkit, not one dial. Hard games make wins memorable, fit the story and push players to use every mechanic, but only when the challenge feels fair.

Silksong's Last Judge kept him going because he could see a path to victory, while long runbacks and random patterns pushed players to rage.

He wants players able to adjust or sidestep challenge. He uses a 100m dash example: 12 seconds is tough for him, impossible for a child and laughable for Usain Bolt.

Tunic's options let his wife enjoy its secrets despite tough bosses. He argues a Souls game's hopelessness-to-triumph arc is not reason enough to leave a game inaccessible.

His second game, Word Play, shipped in seven months. It grew from a two-day Patreon-only game jam prototype in December, mixed Balatro with Scrabble and Bookworm Adventures, and went on to sell more than twice the copies of Mind Over Magnet.

He says the planning he learned on his three-year first game is what let him ship without burning out.

He made spelling feel good before adding any roguelike systems. December went only into a juicy spelling game that worked on mouse, controller, keyboard and touch; shop, perks and special tiles came in January.

He worked in layers: core verb, then game loop, then content, then balancing. Perks grew from 11 test perks to 52, 74, 90 and finally 160.

He judged every choice by which option took the least time. He used simple icons for all 160 perks instead of unique art like Balatro's jokers, which saved weeks.

He accepted that this could make the game arguably worse, because finishing while he still had energy mattered more.

He kept going when two similar games appeared mid-development. Wordatro and Birdigo both mixed Balatro with words, and he briefly considered cancelling and regretted not showing the game earlier.

He decided that several word roguelikes made a genre, not a copy to beat, and that he could only control how good his own game was.

Outside feedback shaped many of Word Play's rules. Playtests at GDC and on iPads led to tile rearranging and bonus points for long words, and the Next Fest demo drew 20,000 players in a week.

A Discord user suggested renaming lives to plays, which fixed losing a life for a correct action and gave the game its title. A final QA group of patrons caught overpowered and pointless perks.

Brown started Game Maker's Toolkit as a way to learn design. The first episode went up on November 6th, 2014. Ten years later the channel spanned about 200 videos and 8 game jams.

He built each episode from research, interviews and analysis. His interest traces to Valve's developer commentary in Portal and Left 4 Dead, which he says opened his eyes to game design as a craft and carried him through games journalism into the channel.

His first game, Mind Over Magnet, took three years instead of one. He announced the plan at the end of 2021 and shipped on Steam on November 6th, 2024, exactly ten years after the first GMTK episode, on PC, Mac and Steam Deck.

Along the way he changed the game's genre and its entire art style. A composer, an artist, a lawyer, playtesters and Patreon backers who funded the work all helped him finish.

He ranks every bug from 1 to 5 and fixes the worst first. GMTK Discord testers found almost 100 issues, including a platform that stuck when the polarity button was mashed and a scroll-wheel binding that made Uni speed like a jet engine.

Fixes created new bugs, so he ran repeated QA rounds until testers stopped finding anything. He says nobody tells you a game is done, so you have to call it yourself.

Brown started Game Maker's Toolkit as a way to learn design. The first episode went up on November 6th, 2014. Ten years later the channel spanned about 200 videos and 8 game jams.

He built each episode from research, interviews and analysis. His interest traces to Valve's developer commentary in Portal and Left 4 Dead, which he says opened his eyes to game design as a craft and carried him through games journalism into the channel.

He judges a game's feel by its mechanics, not its plot. Far Cry 2 and Far Cry 4 share a setting and villain type, yet jamming guns and malaria make one harrowing and plentiful ammo makes the other a blockbuster.

He uses the MDA framework: mechanics live in the code, dynamics in the player's actions, aesthetics in their feelings. A designer can only change the code.

He says no mechanic is right or wrong, only right for the experience. Leaderboards, lives, roguelike saves and fixed cameras all get the same answer from him: it depends.

Alien: Isolation is his favourite case: auto-checkpoints let players die near one to escape the alien, while slow manual save points made saving itself frightening.

He checks who a game is for before criticising it. He first faulted Insomniac's simple Spider-Man swinging, then realised a PS4 tie-in targets casual players who want to feel like Spider-Man at once.

He uses this revision as proof that no design lesson, including his own, should be taken as gospel.

He treats prototypes and playtests as the only proof an idea works. His recurring line is that brains are bad video game simulators, so ideas only show their flaws once built.

He draws on Valve, which treats designs as hypotheses and playtests as experiments, and on designers who follow the fun by keeping the best part of a prototype and cutting the rest.

He values hiding information to create mystery and planning. Mystery comes from a curiosity gap the player closes through their own discovery, ideally with knowledge instead of items.

He applies the same idea to detective games, praising Her Story and Return of the Obra Dinn because they test answers without prompting or allowing guesses.

His first game, Mind Over Magnet, took three years instead of one. He announced the plan at the end of 2021 and shipped on Steam on November 6th, 2024, exactly ten years after the first GMTK episode, on PC, Mac and Steam Deck.

Along the way he changed the game's genre and its entire art style. A composer, an artist, a lawyer, playtesters and Patreon backers who funded the work all helped him finish.

He cut the plan to hit his deadline rather than slip a year. Mind Over Magnet went from 50 puzzles to about 40, five worlds to four, and three magnets to two, and became playable from intro to credits within a couple of months.

The complete build then showed what to restore: the scissor gate, a fifth short world and transition stages. He notes Nintendo cut two Wind Waker dungeons and Shadow of the Colossus shipped 16 of 48 colossi.

He locks content, then polishes and playtests almost daily. New content made the game longer but not better, while small daily fixes, like a pop-up for players stuck on the hint system, improved it with every update.

His weekly rhythm was new content Tuesday to Thursday, bug fixing Friday, a build to at most three testers so they play over the weekend, and footage review on Monday.

He ranks every bug from 1 to 5 and fixes the worst first. GMTK Discord testers found almost 100 issues, including a platform that stuck when the polarity button was mashed and a scroll-wheel binding that made Uni speed like a jet engine.

Fixes created new bugs, so he ran repeated QA rounds until testers stopped finding anything. He says nobody tells you a game is done, so you have to call it yourself.

He built art before the camera system and had to redraw everything. Big pixel sprites in a 1280 by 720 room could not scale cleanly to 1080p, the most common PC resolution, since there is no half a pixel.

He spent months bouncing between rounding, anti-aliasing and borders, then remade every sprite as 4K-ready vector art and kept pixel backgrounds behind a blur. He thinks the compromise hunting took longer than the remake.

He locks content, then polishes and playtests almost daily. New content made the game longer but not better, while small daily fixes, like a pop-up for players stuck on the hint system, improved it with every update.

His weekly rhythm was new content Tuesday to Thursday, bug fixing Friday, a build to at most three testers so they play over the weekend, and footage review on Monday.

He faked his trailer from unfinished content and it worked. Using unplaytested levels and half-finished backgrounds, cut to a stock track's drum stem, the trailer helped push Mind Over Magnet past 30,000 wishlists before Steam Next Fest.

Viewers congratulated him on finishing the game. He recorded in 4K for a 1080p trailer so he could crop, and built from simple mechanics to a montage to a call to wishlist.

He built accessibility into his own game at the right moment. Mind Over Magnet has remapping, toggle aim, options to turn off particles and shake, and tintable magnets with plus and minus symbols.

He says accessibility is best added once code is near final: too early and it gets thrown out, too late and it is a nightmare to retrofit.

Working without a plan left only nine of 24 months as real progress. After GDC feedback he stopped for six months, and an earlier break lasted five; even when working he felt like he was aimlessly noodling around.

Juggling the YouTube schedule that funds the game broke his momentum, so he put the channel on the back burner. With a concrete plan he made about 40 final levels in one month.

He plans scope only once a prototype has shown what is fun. Mind Over Magnet became 50 levels across five worlds, each a single-screen puzzle of 2 or 3 minutes, and each mechanic got three levels: teach, test, twist.

He warns that planning too early stifles creativity and planning too late wastes months. A Trello board of high-level tasks gets a detailed sub-plan as each task starts.

He values hiding information to create mystery and planning. Mystery comes from a curiosity gap the player closes through their own discovery, ideally with knowledge instead of items.

He applies the same idea to detective games, praising Her Story and Return of the Obra Dinn because they test answers without prompting or allowing guesses.

He treats prototypes and playtests as the only proof an idea works. His recurring line is that brains are bad video game simulators, so ideas only show their flaws once built.

He draws on Valve, which treats designs as hypotheses and playtests as experiments, and on designers who follow the fun by keeping the best part of a prototype and cutting the rest.

He thinks open worlds need guidance the player barely notices. He praises Breath of the Wild's landmarks and triangle-shaped terrain and Metroid Dread's points of no return and locks placed near keys.

He warns that waypoints and dotted lines pull players out of the world, and played Fallout 4 with markers off to see what that cost.

He assumed players knew what he knew, and padded puzzles into tedium. Fearing a clean puzzle would be solved instantly, he added traps and red herrings; his dad found the levels tedious and handed back the controller.

Patrick Traynor of Patrick's Parabox had told him to simplify, and he ignored it. After cutting a level from two magnets and four linked steps to one magnet, GDC players got stuck for a minute or two and then hit an aha moment.

He judges a game's feel by its mechanics, not its plot. Far Cry 2 and Far Cry 4 share a setting and villain type, yet jamming guns and malaria make one harrowing and plentiful ammo makes the other a blockbuster.

He uses the MDA framework: mechanics live in the code, dynamics in the player's actions, aesthetics in their feelings. A designer can only change the code.

He says no mechanic is right or wrong, only right for the experience. Leaderboards, lives, roguelike saves and fixed cameras all get the same answer from him: it depends.

Alien: Isolation is his favourite case: auto-checkpoints let players die near one to escape the alien, while slow manual save points made saving itself frightening.

He generates puzzles by crossing every mechanic with every other. A matrix of mechanic pairs turned the vague goal of making an interesting puzzle into concrete experiments, and he made about 26 levels in January 2023.

From the electromagnet and drill pair he found that a magnet stuck near the drill tip escapes fast and far from it escapes slowly, then built the level backwards from that trick. A puzzle bible of mechanics per level showed where progression jumped too fast.

He threw away his first rough demo instead of patching it. Feedback on the 30-day minimum viable product said the robot felt floaty, rules changed between levels, puzzles lacked aha moments and the magnet did not feel central.

He made the magnet a design pillar, decided to focus on puzzles instead of mixing in platforming, and rebuilt the character controller, which alone took almost two months.

He built accessibility into his own game at the right moment. Mind Over Magnet has remapping, toggle aim, options to turn off particles and shake, and tintable magnets with plus and minus symbols.

He says accessibility is best added once code is near final: too early and it gets thrown out, too late and it is a nightmare to retrofit.

He threw away his first rough demo instead of patching it. Feedback on the 30-day minimum viable product said the robot felt floaty, rules changed between levels, puzzles lacked aha moments and the magnet did not feel central.

He made the magnet a design pillar, decided to focus on puzzles instead of mixing in platforming, and rebuilt the character controller, which alone took almost two months.

He threw away his first rough demo instead of patching it. Feedback on the 30-day minimum viable product said the robot felt floaty, rules changed between levels, puzzles lacked aha moments and the magnet did not feel central.

He made the magnet a design pillar, decided to focus on puzzles instead of mixing in platforming, and rebuilt the character controller, which alone took almost two months.

He thinks open worlds need guidance the player barely notices. He praises Breath of the Wild's landmarks and triangle-shaped terrain and Metroid Dread's points of no return and locks placed near keys.

He warns that waypoints and dotted lines pull players out of the world, and played Fallout 4 with markers off to see what that cost.

He failed by starting with art and story on earlier games. His Picross game Carter's Curse had sprites, cutscenes and menus, but its puzzles ignored the enemies on screen, so it was shallow and dull.

A noir game stalled on research and a Snake remake stalled on movement code. He now prototypes the core idea first with borrowed art and buggy code.

He wants players able to adjust or sidestep challenge. He uses a 100m dash example: 12 seconds is tough for him, impossible for a child and laughable for Usain Bolt.

Tunic's options let his wife enjoy its secrets despite tough bosses. He argues a Souls game's hopelessness-to-triumph arc is not reason enough to leave a game inaccessible.

He says no mechanic is right or wrong, only right for the experience. Leaderboards, lives, roguelike saves and fixed cameras all get the same answer from him: it depends.

Alien: Isolation is his favourite case: auto-checkpoints let players die near one to escape the alien, while slow manual save points made saving itself frightening.

He treats prototypes and playtests as the only proof an idea works. His recurring line is that brains are bad video game simulators, so ideas only show their flaws once built.

He draws on Valve, which treats designs as hypotheses and playtests as experiments, and on designers who follow the fun by keeping the best part of a prototype and cutting the rest.

He checks who a game is for before criticising it. He first faulted Insomniac's simple Spider-Man swinging, then realised a PS4 tie-in targets casual players who want to feel like Spider-Man at once.

He uses this revision as proof that no design lesson, including his own, should be taken as gospel.

He thinks a good puzzle has a catch, an assumption and a revelation. A catch is two needs that seem to conflict, a planted wrong assumption walks the player into it, and the solution reveals a new truth about the rules.

He contrasts Portal 2 and The Turing Test, which share almost the same puzzle, to show how much presentation decides difficulty.

He praises easy modes that explain their purpose first. Celeste explains who Assist Mode is for and recommends a first run without it, while Cuphead's Simple Mode sits beside normal mode and cuts content.

He argues language, placement and timing tell players how much weight an option carries, so developers can offer help without asking players to design the game.

He favours systemic games that let players invent solutions. In a systemic game entities sense each other through shared rules, so a Far Cry tiger can maul a guard without a script.

He credits the immersive sim lineage from Ultima Underworld and Looking Glass through Deus Ex for treating players as collaborators, and warns that a reliable safe option kills creative play.

He values hiding information to create mystery and planning. Mystery comes from a curiosity gap the player closes through their own discovery, ideally with knowledge instead of items.

He applies the same idea to detective games, praising Her Story and Return of the Obra Dinn because they test answers without prompting or allowing guesses.

He splits Zelda dungeons into find-the-path and follow-the-path types. Four questions sort them: branching paths, choice of order, backtracking, and whether the player does that backtracking.

He prefers find-the-path dungeons because they test spatial awareness, and argues Zelda designers treat the level itself as the challenge.

He favours systemic games that let players invent solutions. In a systemic game entities sense each other through shared rules, so a Far Cry tiger can maul a guard without a script.

He credits the immersive sim lineage from Ultima Underworld and Looking Glass through Deus Ex for treating players as collaborators, and warns that a reliable safe option kills creative play.

He thinks open worlds need guidance the player barely notices. He praises Breath of the Wild's landmarks and triangle-shaped terrain and Metroid Dread's points of no return and locks placed near keys.

He warns that waypoints and dotted lines pull players out of the world, and played Fallout 4 with markers off to see what that cost.