Inside Derek Yu's Philosophy — Finishing as a Skill, Not a Talent

Introduction

Derek Yu (b. 1982) is an American game maker and pixel artist. He made the cave-diving roguelike Spelunky (2008; remade 2012) and Spelunky 2 (2020), and UFO 50 (2024), a collection of fifty games presented as the library of a fictional 8-bit console. He is also long known as a custodian of TIGSource, the indie games community site. He is not a puzzle designer. Even so, I think he belongs on the shelf next to them.

The reason is simple. The texts he has left behind are not about what goes inside a game but about getting a game finished. The essays on his own site, MAKE GAMES WITH DEREK (derekyu.com), dissect not how to have ideas but the machinery by which ideas fail to reach completion. For anyone making puzzles alone or in a tiny team, that is a far more urgent subject than aesthetics.

This is not a review. It is a study of the person — his philosophy, his obsessions, how he handles failure, his dilemmas and his acknowledged influences — built only from what he has written or said on the record.

Background

He met games through an Atari 2600 as a child and was designing them on paper by second grade, moving through text adventures and Klik & Play before taking a computer science degree and working as a freelance illustrator in San Francisco (Wikipedia). He is an author who holds both the art and the code himself.

In 2007, working as Bit Blot with Alec Holowka, he released Aquaria and won that year's Seumas McNally Grand Prize at the IGF. In 2008 he put out the free original Spelunky, made in GameMaker; after the 2012 HD remake, its shape — a hand-authored cave of traps reshuffled every run — entered the general vocabulary of design.

In 2016 Boss Fight Books published his own book, Spelunky, covering randomisation, challenge, player feedback, team dynamics and what it actually takes to finish a game (Boss Fight Books, 2016). That same year he and Jon Perry began UFO 50, which took eight years to finish.

Philosophy — finishing is a skill that accumulates

Across everything he has said, the loudest repetition is this: the ability to finish grows in proportion to the number of things you have finished. In "Death Loops" he writes, "Finishing games gets easier with each game you finish, and the more games you finish, the better you will understand what it takes to make a successful game" (derekyu.com). In "Indie Archetypes" he says it again — "You get better at finishing games the more of them you finish" — and gives the order of operations: summit a boulder, then a bigger boulder, then a hill (derekyu.com).

The same line surfaces in 2024: "Finish a game by any means possible. Release one sloppy, fun little game quickly and then make a game that's slightly bigger and slightly less sloppy" (GameMaker, 2024). Any means, sloppy is fine, but ship it. Not a word has moved in over a decade (gamemaker.io, 28 Nov 2024).

Underneath it is an observation about time itself deforming the work: "The longer a game is in development, the more you feel like it has to succeed to justify its development time" (Death Loops). Development length raises not only quality but the bar the author sets for their own success. Finishing quickly, then, reads less like quality control than like designing the author's own state of mind.

And he refuses to universalise method: "the 'best' tools, algorithms, etc. are the ones that get the job done, and what gets the job done depends very much on the person using them" (Indie Archetypes). Tools and methods are subordinate to the temperament of the person holding them.

Obsessions — naming the shapes of failure

The recurring gesture in his writing is giving proper names to vague troubles. The state of not finishing becomes the "Death Loop", split into endless restarting and endless polishing. Of the first he offers: "Imagine running a marathon but you're running in circles instead of toward the finish line" (Death Loops).

"Indie Archetypes" sorts non-finishing developers into five types — Daydreamer, Inventor, Burrower, Wanderer, Finisher. Not to condemn anyone, but as a self-diagnostic that selects a remedy. He goes as far as to suggest that if you find you would rather not finish, you might finish alongside someone with a high aptitude for finishing.

"Assessing Risk" proposes a two-axis frame — resources in, breadth of appeal — and warns: "The more you put in, the more you'll probably need to get back to call it a 'success'" and "It's easy to fall into that Morass in the Middle where you're spending lots of time on aesthetics that will not make an appreciable difference" (derekyu.com). Again: name the swamp first, then walk out of it.

His feedback essay has the same shape. Sympathetic feedback (friends, fellow makers) is separated from unsympathetic feedback (strangers), and only the latter measures real appeal. Then: "Players will know HOW they feel about your game, but they may not necessarily know WHY they feel that way" (derekyu.com). Symptom is not diagnosis.

Failure — filed as structure, not as war story

This section needs care. As far as I have read, he almost never tells a war story about a cancelled project or a personal disaster. What he does instead is generalise failure into something others can use. Death Loops, Archetypes and Risk are all written as prescriptions, not memoirs. That, I think, is his particular way of settling accounts with failure.

The one place he steps into feeling is criticism: "The lack of understanding and empathy can definitely feel bad and it's natural to focus on the negative feedback and mean, dismissive comments, even if they're outweighed by the positive 100-to-1" (Getting Feedback). Having granted that, he moves the remedy from emotion to volume analysis: "A LOW VOLUME of positive feedback is worse (from a financial perspective) than a HIGH VOLUME of mixed feedback." Whether it hurts and whether it sells are measured on different rulers.

On sequel anxiety his answer inverts the expectation: "Surprisingly, I felt a lot less pressure working on Spelunky 2" (The Escapist, 2020) — he credits resources and an experienced team, and adds, "Spelunky 1 would always exist no matter what we did to it, so I felt free to follow my vision for the sequel." The first game cannot be un-made; therefore the second is free. The fear of failure is neutralised by the permanence of the past work.

In the same interview he admits plain uncertainty about reception: "I was surprised, only because it's hard to know what people want out of a sequel." That refusal to overclaim runs through everything he writes.

Dilemmas — the man who said finish small shipped an eight-year project

I have a habit of setting interviews against each other. Two tensions here are hard to ignore.

First: the man who wrote "build small, ship fast" and "the longer development runs, the more it has to succeed" spent roughly seven years on Spelunky 2 and eight on UFO 50. I could not find a text where he squares this directly, but his framing of UFO 50 looks like an answer: "The development of UFO 50 often felt like a very long game jam between friends inspiring each other back and forth" (Game Developer, 2025). Not one eight-year game, but fifty small completions (gamedeveloper.com, 14 Mar 2025).

Second: "Assessing Risk" is a commercial frame balancing appeal against resources, yet the same author writes in "Getting Feedback" that "A truly great game tries to accomplish something more than 'have as many people like my game as possible.'" One says lower your risk; the other says take it. He never reconciles them. To my reading that is not a contradiction but a declaration of separate uses: the tool for surviving is kept apart from the reason for making.

Third, he treats the changing environment itself as a bind: "a lot of things have gotten easier for indie game developers, but ... I feel like the difficulty is sort of scaled with the power of indie game developers" (GDC Podcast, 2021). Better tools, higher expectations, more noise. He explains his own industry in the vocabulary of difficulty tuning (Game Developer, 29 Jul 2021).

Influences

Only what he names himself. In 2024: "Developers like Maddy Thorson and cactus inspired me a lot early on with their cool, energetic games. They showed me the power of GameMaker and of indie games in general" (GameMaker, 2024). Thorson would go on to make Celeste; cactus is Jonatan Söderström, later of Hotline Miami. We have a separate study of Thorson on this site.

His attitude to tools belongs here too: "I hope it continues to improve on what it's great at and not try to be the 'everything tool', because the latter is what causes tools to collapse under their own weight eventually" (ibid.). A tool that does everything collapses under its own weight — again the concern is not what can be built but what can be finished.

For UFO 50 the influence is less a person than the 8-bit hardware era itself. He describes imposing a 32-colour palette and simple colour rules on the team while deliberately not being strict about on-screen sprite counts (Game Developer, 2025). Not faithful emulation, but borrowing only the usefulness of the old constraints.

A reading

From here it is my own reading. I read Derek Yu as a designer of the author's endurance rather than of games. What his essays address sits consistently outside the board: loops, morasses, archetypes, the volume of feedback — none are properties of a work; all are names for routes by which an author breaks. His real arena may have been the author's experience rather than the player's.

Read that way, UFO 50 looks like an answer to his own essays. The man who wrote that long development corners you found exactly one way to spend eight years: cut them into fifty completions and share them with friends. "A very long game jam between friends" should be taken not as a metaphor but as a design spec. The exit from the death loop is not only "choose short projects" but "restructure a long project into a chain of short finishes."

And that translates directly into a prescription for puzzle makers: rather than sinking years into one large puzzle game, finish small boards and give structure to the accumulation. The daily-puzzle format is one instance of exactly that design. This is why a man who never talks about puzzles still belongs on a puzzle maker's shelf.

Closing

If you want to understand him, read the four essays at MAKE GAMES WITH DEREK before playing anything — "Death Loops" and "Indie Archetypes" land hardest on anyone holding a folder of unfinished projects. If you would rather start with a game, take the free original Spelunky (2008), or any single short title inside UFO 50: whichever of the fifty you pick, you will land on something that was finished.

For neighbouring figures, this site has studies of Maddy Thorson (whom he credits), Bennett Foddy (who designs failure from another angle) and Terry Cavanagh (who has talked for years about finishing as a solo developer). Read together, they show what the makers of this era feared and what held them up.

Sources

Primary sources used in this article:

・Derek Yu, "MAKE GAMES WITH DEREK" (the author's own essay index)

・Derek Yu, "Indie Game Dev: Death Loops" (author's essay)

・Derek Yu, "Indie Game Dev: Indie Archetypes" (author's essay)

・Derek Yu, "Indie Game Dev: Assessing Risk" (author's essay)

・Derek Yu, "Indie Game Dev: Getting Feedback" (author's essay)

・GameMaker, "Finish a Game by Any Means Possible: An Interview with Derek Yu", 28 Nov 2024

・Game Developer, "Inside UFO 50's incredible 8-bit journey", 14 Mar 2025

・Game Developer, "Spelunky, indie life, and finding success with Derek Yu — GDC Podcast Ep. 22", 29 Jul 2021

・The Escapist, "For Derek Yu, Game Development Is a Life-Long Passion, and Spelunky 2 Is Just the Start" (interview), 9 Dec 2020

・GDC 2021, "One More Run: The Making of 'Spelunky 2'" (the author's talk)

・Boss Fight Books, Spelunky by Derek Yu (the author's own book, 2016)

・Wikipedia, "Derek Yu" (used to check dates and biography)