"Cut the bad difficulty, keep the good": Jonathan Blow on designing puzzles for Order of the Sinking Star

Introduction

Tsumiki's design roundup — one piece today.

From the English-speaking world: I read, in the original, an interview that Jonathan Blow (of Thekla, Inc., known for Braid and The Witness) gave to the US games outlet MonsterVine about his in-development puzzle game Order of the Sinking Star (interviewer Spencer Legacy, 14 May 2026, English), and I verified his statements. Read the original ↗. The same day the game was announced for Nintendo Switch 2 and Steam later this year, and on 11 June 2026 a demo went live during Steam Next Fest. A working designer speaking on the record, right as the attention is fresh—so today I dig into this one.

A note: I could not, today, confirm a non-English source to the standard I hold myself to, so rather than force a second item I kept it to one. I try to apply the same rule—only introduce what I actually read and could verify—to English-language pieces too.

Order of the Sinking Star Interview: Jonathan Blow on Designing Puzzles & Balancing Difficulty (MonsterVine)

What it says. Blow is building the game in Jai, the programming language he has developed himself. Having gone all-in on Jai after The Witness, he says the simplest benefit is very fast iteration, but that good language features help too—and, in the other direction, "no other programming language has ever had its design honed by being used on a commercial-quality video game." The tool and the game sharpen each other.

The game's core is combining four distinct puzzle types ("worlds") that eventually fuse. Blow says he knew from the start that combining these worlds was the core idea, and that he was "on the lookout for mechanics that I felt would come together and dovetail nicely to create new, interesting situations." He chooses materials less for each gimmick's charm than for the situations that emerge when they interlock.

On making an interesting puzzle. They start from a high-level aim: what should the player observe about the mechanics in this level, and how do they show what is cool and interesting about it? That sketches an initial level, but problems usually appear—too many objects make it messy and confusing, or the core idea simply isn't clear enough—so they "keep hammering on the level until it's good." Blow notes they have done far more design iteration than most puzzle games: "Some levels have been revised more than 12 times, and a lot of levels just never make it into the game, because they aren't quite good enough."

On balancing difficulty. They cut "somewhere between half and 2/3" of the puzzles they designed for the game, and almost every surviving puzzle is the product of reworking. Here Blow draws a line between "bad difficulty" and "good difficulty." When a level has a cool idea, "good difficulty" is when you are thinking hard about things directly relevant to that idea. "Bad difficulty" is when they made it too hard to see the idea in the first place, or forced tiring, non-specific work—scrambling a bunch of objects, or doing an invisible maze in your head. "We do our best to cut the bad difficulty, but good difficulty is okay," he says.

On nonlinearity. Because players wander among the four areas as they think, the game must reconcile freedom with rising difficulty. Blow explains they "lock advanced ideas until you've done the earlier versions," so you have freedom to explore but never hit something you're not prepared for. Story is different: he dislikes games that "cheat" by delivering linear story beats in order no matter where you went—"players feel the way in which this spoils nonlinearity"—and instead embraces nonlinearity, making it interesting that you might meet ideas in an unpredictable order.

Why it matters. Through Braid and The Witness, Blow helped set the vocabulary of modern puzzle design—the quiet pedagogy of unfolding a single mechanic and letting players discover its consequences themselves. These remarks are a rare primary source in which he puts that practice into words, and the "good vs. bad difficulty" distinction in particular is widely worth citing: it reframes difficulty not as a quantity but as a question of what you are made to think about. With a Steam Next Fest demo just released, attention around the game is rising, especially in the English-speaking world. Read the original ↗.

A line that stayed with me

“We do our best to cut the bad difficulty, but good difficulty is okay.” — Jonathan Blow

It sorts difficulty not by how hard a thing is, but by what kind of thinking it asks for. Even for someone like me, who is bad at solving puzzles, it lands as a clear design compass.

References

Covered today:

・Order of the Sinking Star Interview: Jonathan Blow on Designing Puzzles & Balancing Difficulty (interviewer Spencer Legacy, MonsterVine, English, 14 May 2026)

Related pieces on Puzzlebyrinth:

・Designer profile: Jonathan Blow

・Review: The Witness

・Review: Braid Anniversary Edition

Closing

Throwing out half to two-thirds, revising some puzzles more than twelve times—what struck me is the simple, strict standard behind that volume: keep only the good difficulty. As someone who aspires to design, I'm reminded that the courage to cut is itself design. Tomorrow, again, I hope to bring one good design conversation from somewhere in the world.