Order of the Sinking Star Development Time: About 10 Years

Order of the Sinking Star took about ten years to make. Thekla’s own announcement of the design method describes a game that “transformed into an incredible 10-year puzzle adventure and programming project”, the art team has spoken about the game’s ten years in production, and press coverage at reveal described nearly a decade of development. It grew out of a small experiment Jonathan Blow intended as a proving ground for Jai, the studio’s own programming language, and it shipped on October 8, 2026.

Quick Answer

The short version: development ran roughly from after The Witness shipped in 2016 to the October 8, 2026 launch, it never had a public reveal until December 11, 2025, and the project kept growing because a tiny language test turned into four colliding puzzle worlds. The table below is what is actually on the record.

Official source: Making of Order of the Sinking Star: The Game Design Supercollider

When did development actually start?

Nobody has published an exact start date, and this page will not invent one. What is on the record is duration: the official site says the game has been in the making for “nearly a decade since the release of The Witness”, press coverage at the December 2025 reveal described it as nearly ten years in development, and Thekla’s September 2026 announcements call it a 10-year project outright. The honest answer is therefore “about ten years, starting sometime after The Witness shipped in 2016”.

[!note] The three official-ish framings all land in the same place — “nearly a decade” in 2025, “10-year project” in 2026 — so “about ten years” is the safe statement. A specific first day of production does not exist in any official material, interview or store page, and no timeline on this site claims one.

The origin story is documented, though. Press reporting from Game File describes the project as one Blow “first intended to be tiny, just a proving ground for a new programming language he was creating”, with his own retrospective line: “I should have known that it would turn big.” The language was Jai, and the reason for writing it goes back to The Witness: by the end of that project Blow had grown tired of C++, telling Edge he “felt very sad that I had to go into the office every day and program on C++”. So the ten years are, in a real sense, two projects at once — a game and the language and engine it is built with.

Why did a small experiment become a ten-year game?

Because the experiment worked and the scope followed. A language test needs a game to test against; this one needed four worlds.

The design method is the official explanation for the scale. Instead of building from a single core concept, Thekla created what it calls a “game design supercollider” — smashing together ideas from multiple puzzle game designers to produce, in its words, a combinatoric explosion of possibilities. Four foundation games by other designers were licensed, expanded into the four worlds, then wired together so their systems collide in the Gold Rooms. The scope of that arrangement is easier to grasp on the page about the four games it’s based on, and it is the main structural difference from Braid and The Witness, which each grew from one designer’s rule set — see the comparison with The Witness and Braid.

The iteration data from Blow’s May 2026 interview gives the clearest picture of where the years went: some levels were revised more than twelve times, and Thekla cut somewhere between half and two thirds of the puzzles it designed. A game that ships over a thousand puzzles after cutting two thirds of everything built is a game that designed several thousand.

[!tip] Ten years here is mostly iteration, not stalling. The public evidence points at one designer-scale studio revising levels over and over, discarding most of the work, and holding the reveal back until December 2025 — not at a project sitting idle.

Who spent the ten years, and how big was the team?

Thekla stayed small, and the staffing statements that exist all describe gradual growth rather than a large crew.

• Art: Art Lead Ethan Scheu has been on the project “for the last four of the ten years”, and the art team “has ranged from three to now eight artists in total”.
• Sound: Lead sound designer Hans Christian Kock has been on the game since 2019 — “7 years for the sound design” as of September 2026 — and music designer Martin Stig Andersen also started in 2019. Composer Peter Due joined in 2021.
• Music recording: the lute you hear on the Bard was sampled with player Fredrik Bock in 2022, and the string parts were recorded live at FAMES Studio.

Those dates are themselves a development timeline: audio work was underway by 2019, roughly the midpoint of the ten years, while the art leadership quote is the source of the “ten years in production” phrasing used everywhere else.

What was the public timeline from reveal to launch?

The game was in production for years before anyone could see it. Once it went public, the road to launch was about ten months.

The demo matters in a development-history sense too: it was the first time players touched ten years of work, and its progress carried into the launch version. The demo length and puzzle count page breaks down what that slice actually contained.

Does the ten years of development show in the finished game?

At the level of raw volume, yes. The store copy promises over 1000 interconnected, hand-crafted puzzles, Blow has put the real figure above 1400, and the four worlds were designed to be combined rather than completed one at a time. The how long to beat page puts the two headline estimates in context: up to 250 hours at a normal pace, past 500 hours for full completion — both Blow’s own estimates from reveal day, not measured playtimes.

The development approach also shows in structure. Early puzzles work on a 2D plane and later ones move into 3D, which press interviews describe as a deliberate way to keep play fresh over that length; the top-down tile-based form itself exists because a shared grid is what let four separately authored puzzle games combine at all. And the language work is a visible legacy of the decade: Thekla built its own engine in Jai, and the studio’s stated post-launch plan is to ship the compiler — binary first, source later — and release the language plus the engine’s non-game parts for free.

[!note] Ten years is a duration, not a quality score. None of this page’s claims depend on the game being good; they depend on statements Thekla and press published between December 2025 and October 2026, which is the window this page covers.

For what the finished package contains on release day — platforms, editions and pricing — the release date page carries that breakdown.

Related

• Order of the Sinking Star vs The Witness and Braid — how the ten-year project compares with Blow’s earlier games
• The four games it’s based on — the licensed designs the supercollider method collided
• How long to beat: 250 or 500 hours? — what the reveal-day estimates measure
• Demo length and puzzle count — the slice players got in June 2026
• Beginner tips and first hour — starting the finished game

This page is re-checked if Thekla publishes a firmer start date or post-launch statements about the development span.