Blow uncovers a puzzle, then builds it

Jonathan Blow has said that building Braid, Number None, 2008, felt less like inventing content and more like uncovering something that was already there. It is a flattering account of puzzle design, and it is easy to misuse. A rule that feels inevitable in the designer head is not yet a puzzle. It is a rule.

The Witness, 2016, makes the other half of the method obvious. The line-panel rule is simple, and the world keeps finding new consequences for it until the player feels they discovered the consequence themselves. That feeling is constructed. Someone placed the panel where the old assumption would fail, and placed the earlier panels where the assumption was still true. Uncovering is the player experience. Placement is the job.

Jonathan Blow has said that building Braid felt less like inventing arbitrary content and more like uncovering things that were already there. By his account, more of the finished game came out of development than he had consciously put in at the start.

Behind that remark sits a method that runs through both Braid and The Witness. Rules and puzzles were located through programming and playtesting instead of being handed down by a completed plan.

The sequence is always the same. Take a mechanic, work out what it quietly implies, then turn the surprising implications into puzzles. What results is not a corridor of arbitrary obstacles. It is a run of discoveries about one small universe and how it works.

A rule tends to arrive once the mechanic is running and someone starts poking at it. Braid opened with one strong idea: rewind time by a practically unlimited amount. With that coded, further ideas surfaced on their own. If everything in the world can be rewound, then some things can be exempt from rewinding. That exemption is a rule.

The work that follows is play. You look for consequences that were not visible when the rule was written down. Suppose a moving platform ignores time manipulation. The player can now rewind their own position back to a moment when the platform is somewhere else entirely. Time resumes, and the character drops.

Calling that a gimmick undersells it. It is a true and interesting fact about how Braid's world behaves. A puzzle can then exist to illustrate that fact, so the player runs into the same odd consequence the designer hit while building the system.

The Witness was built the same way, starting from the mechanic of drawing a line across a grid. Playtesting surfaced something Blow noticed: a drawn line frequently partitions cells. An observation like that is the raw material for a rule.

A puzzle type takes shape from there. One panel might fall easily to a path looped around part of the grid. Move the exit on a near-identical panel and the same loop now seals the player away from finishing. Tracing by habit stops working, and understanding partitioning becomes the point.

The chain runs mechanic to rule to consequence to puzzle. None of it began with a decision that the game needed a certain kind of challenge. It began with small questions asked very early, while the systems were still being written.

None of this leaves the designer idle. The first job is to chase the ramifications of every change as far as they will go.

In The Witness, that means interrogating each piece in turn and asking how it can be twisted: the grid, the cells, the line, the environment, the panel itself. In Braid, it means asking what a given rule change does to every object in the game.

So when Braid gives certain objects immunity to rewind, the idea does not get used once and dropped. Enemies, keys, doors, clouds, platforms, and the player character can each carry that property. Every version puts the rule under a different light and asks what it really means.

Job two is presentation: arranging the puzzle so the player gets the cleanest possible shot at the interesting fact sitting at its center.

Misdirection earns its place here. Braid's puzzle Hunt asks the player to kill all the monsters and arranges them so the most obvious order fails. That failure is doing real work. It demonstrates why the underlying rule matters and blocks anyone from brute-forcing their way past without understanding it.

Sequences, pairings, and reprisals pull in the same direction. A small puzzle about opening two doors with a single key sets up a heavier version further on. A layout the player already knows can reappear in a later world under different rules, so the shift in consequence is visible.

Expected behavior can also be subverted outright. One level is solved by realizing that rewind must not be used at all. Another bends time so that only one of two gates will ever open. Puzzles like these ask for noticing, not just for steps.

Job three is curation. Mechanics, rules, and puzzles get cut when they hold no surprise, overlap too heavily with something else, or say nothing worth saying.

Braid and The Witness both grew out of earlier projects that were shelved, because the main mechanics in those projects did not open up a rich enough space. Rules went the same way. Braid once had a turn-based world, and Blow removed it when its consequences failed to surprise and the rule started to feel contrived.

What makes curation tricky is that the survivors are not always the conventionally fun ones. An awkward piece can stay because it is interesting, because it surprises, or because the universe feels incomplete without it. A puzzle in which a key wanders off by itself is odd, and it can still express something real about the rules.

Under this approach the puzzle carries a message. It is an idea traveling from designer to player, and solving it is how the player answers, "I understand."

Compare that with "I finally figured it out." Plenty of puzzle games run on arbitrary steps, elaborate sequences, red herrings, and obscure mechanisms. Victory arrives without anything coherent being revealed.

The fairness of Braid and The Witness comes from somewhere else. Their hardest puzzles are usually about seeing what known mechanics imply in a new layout, combination, or setup. Every element is introduced up front, early puzzles teach it quickly, and later puzzles ask for a deeper reading of the same rules.

Compactness helps too. A puzzle often explores a single idea on a stage small enough to hold in your head, without red herrings or demanding execution. Seeing the solution is the hard part; carrying it out is usually simple.

The moment of solving does not resemble grinding a Rubik's cube or guessing at a riddle. It feels like spotting something that had been sitting in plain view. The answer was in front of the player from the beginning, waiting on the right way of looking at the world.

That is where the force of the aha moment comes from. It is not relief at stacking up enough steps, or at finally guessing what the designer had in mind. The world simply got clearer.

The more a puzzle is about something specific and real inside the rules, and the less it leans on arbitrary difficulty, the more that epiphany is worth. Design may begin as exploration, but the finished puzzle hands the discovery to the player in its cleanest form. You only learn whether it lands by getting the thing running and playing it.

In the Flockbay app, state one rule in a sentence, then ask for two rooms. The first room should make the rule obvious. The second should break the assumption a player would bring from the first. Play them in order. If the second room needs a caption, it is not a discovery yet.

An AI puzzle maker and a logic puzzle template will try to explain the rule in text. Delete the text and see if the rooms still speak.

More on this: Puzzle game maker: the rule is the game, and the rule gets written