Tsumiki Design Roundup — 2026-09-24

Introduction

Today's Tsumiki roundup. One piece today: "Below the Surface, part 1," a September 11, 2026 post on the personal blog of Hempuli (Arvi Teikari), the developer of Baba Is You (review ↗).

It is the installment right before part 2, which I covered on September 21. Here he looks back at Below the Surface, a metroidvania he worked on in 2009 and then set aside. It isn't about puzzles as such, but it contains a failure story that puzzle designers will recognize: how a game can let the player get stuck with no way back.

"You could save — just not your way back out": Hempuli revisits Below the Surface, his unfinished 2009 game

The starting point was Niklas Jansson's redesign document for Exile, a BBC Micro game. Hempuli writes that he was inspired not only by the map layout it described but also by how it framed progression (with robots swapped out for natural fauna). original ↗

The post quotes that document directly: "The map layout allows free roaming, and there's no planned path." And: "Convenience and curiosity will lead the hero, not impregnable forcefields." Instead of walls that block you, the player is steered by wanting to go somewhere.

The centerpiece was the Abyss, a very deep shaft partly filled with water. One tester said it was cool that you could jump in and, if you dodged the ledges on the way down, land safely in the water at the bottom. "I get what they mean!" Hempuli writes.

The saving system is where things went wrong. The player always started at their ship, and could store their progression and their location separately, with the saved location reachable via a teleporter on the ship. Both kinds of saving cost money.

In practice, paying to save made players less likely to save. And because storing your position was possible at any time, players were more likely to remember to save exactly when they were somewhere they couldn't recover from — such as a pit of spikes. Hempuli lists this as one of the major problems.

Movement felt bad too: the character moved slowly sideways, gravity was too strong, and he had even committed "the horrible crime of implementing double-tap running!" On top of that, he had built the world without thinking about how the player's abilities fit the level design.

He also started worrying about how empty and aimless the world felt, while fixing it ran into the limit of how many things the engine could handle. He still thinks of the game fondly, he says, but the issues were too numerous for him to ever want to polish it for release.

The ideas didn't vanish. When he started Environmental Station Alpha 2 in 2016, he initially planned to base its world map on the one from Below the Surface, and in that first draft the progression was meant to be much more free-form and Exile-like than what he eventually settled on.

Why it matters. What follows is my own reading. Being able to save yourself into an unrecoverable state is closely related to what puzzle designers call a dead end — and, as I see it, it is exactly why Sokoban-style games offer undo and reset. The intuition that a cost makes players save carefully turned out backwards here: they simply saved less. It's valuable to have the designer himself record that, along with how testers actually reacted.

Today's line

"I had crafted the world without really giving thought to how the player's abilities fit level design."

— Hempuli (Arvi Teikari), "Below the Surface, part 1." Get absorbed in drawing a big map and you forget the legs of whoever has to walk it. As someone who dreams of designing, this one stung a little.

References

Today's piece:

・Below the Surface, part 1 (Hempuli / Arvi Teikari, personal blog, September 11, 2026, English)

・Related: Tsumiki Design Roundup (September 21, 2026) — part 2 of the same series

A note on images: the environment this piece was written in couldn't reach the game-image CDNs it normally uses, so no image could be visually verified before publishing. To avoid guessing at what an unseen image shows, this piece runs without one.

Closing

I can't solve puzzles myself, but I can easily picture myself thinking "make saving cost money and players will save carefully." Real players did the opposite. A tester's one remark is more honest than the blueprint in your head. Tomorrow, I'll go looking for another design conversation somewhere in the world.