Overview
Most programming games hand you a toy assembly language and call it minimalism. A=B, from the aptly named Artless Games, calls that bloat. Released on Steam in March 2022 for Windows and Mac at $4.99, it hands you exactly one instruction — A=B, meaning “replace substring A with B” — and then asks you to convert case, count characters, sort letters, and eventually perform binary arithmetic and multiplication. With find-and-replace. That’s the whole pitch, and it’s an audacious one.
If you’ve studied computation theory, you’ll recognize the trick: string-rewriting systems of this kind are Turing-complete, which means this ludicrously constrained language can, in principle, compute anything a laptop can. A=B is essentially a playable proof of that theorem. Each chapter escalates the demands while grudgingly introducing a handful of extra keywords that expand what the language can express — and every addition feels like the developer conceding ground an inch at a time, keeping the core deliberately spartan.
The market positioning is unambiguous. This sits on the same shelf as Human Resource Machine, TIS-100, and Shenzhen I/O: games that treat constraint as the puzzle. Artless Games made a bet that a sufficiently elegant idea doesn’t need dressing up — no art budget, no narrative, no mercy. For the right player, the bet pays off.
Key features
- A one-instruction language: Every program is an ordered list of substitution rules. The engine scans your rules top to bottom, applies the first match, and starts over. That’s the entire execution model — and internalizing its consequences is the game.
- A curriculum in disguise: Chapters progress from trivial case conversion to sorting, counting, and full binary arithmetic. Later chapters introduce a few additional keywords and constructs that widen the language without ever diluting it. You’re effectively taking a course in rewriting systems and computability, except nobody tells you that until you’ve already learned it.
- Escalating algorithmic challenges: The puzzles don’t test whether you know an answer; they test whether you can invent an encoding. Discovering that you can use marker characters as cursors, counters, and state machines is the game’s central, repeatable thrill.
- Zachtronics-style open-endedness: There’s no single intended solution. Your kludge and someone else’s elegant five-rule construction both pass, which makes post-solve refinement its own optional metagame.
Analysis
The Gameplay Loop
The loop is simple: read the spec, write rules, run, watch your string mutate, swear, revise. What makes it hard is how quickly the single instruction stops feeling simple. Rule ordering is everything — the engine always applies the first matching rule, so your program’s control flow lives entirely in the sequence of your list. Get it wrong and you’ve built an infinite loop out of two lines. Get it right and you’ve built a state machine out of typography.
The early chapters are onboarding done properly. Converting a to A teaches the mechanic; the game then immediately complicates it with tasks that can’t be solved by naive replacement, forcing you to discover auxiliary symbols — scratch characters you inject into the string as temporary markers, cursors, or unary counters, then erase before the checker sees your output. Nobody teaches you this technique. You derive it under pressure, and the moment it clicks is the best “aha” the puzzle genre can offer: not spotting a hidden object, but inventing a tool.
By the back half, A=B stops being cute. Sorting a string with substitution rules is a legitimate algorithms exercise (it’s bubble sort, if you squint — adjacent out-of-order pairs are exactly what a substitution rule can see). Binary multiplication is the kind of problem that will eat an evening and leave you sketching encodings on paper. The difficulty curve is the game’s most divisive property: the final stretch assumes a comfort with algorithmic thinking that casual puzzle fans simply won’t have. That’s not a flaw so much as an honest declaration of audience, but buyers should know the water gets deep fast.
Where It Beats Its Peers — and Where It Doesn’t
Against the genre’s touchstones, A=B carves out a real niche. Human Resource Machine gates you gently and wraps everything in charm; TIS-100 buries you in an obtuse manual and parallel architecture. A=B is more conceptually radical than either. Its language is smaller than TIS-100’s and yet the ceiling is just as high, which makes every solution feel more like a discovery than an implementation. When a tiny rule set turns out to be this powerful, the power feels earned.
What it lacks is the Zachtronics connective tissue. There are no histograms comparing your cycle count against friends, no narrative framing, no in-fiction manual to pore over. The optimization impulse — can this be done in fewer rules? fewer steps? — is left entirely to the player’s own motivation. For self-directed players that’s fine. For everyone else, the absence of external scoring pressure means solved puzzles stay solved, and some of the genre’s longest-tail replayability evaporates. It’s a real omission, and the one place where “artless” reads as a limitation rather than a philosophy.
Friction, Honest and Otherwise
Most of A=B’s difficulty is the good kind — conceptual, fair, yours to overcome. But debugging can turn tedious. When a 30-rule program misbehaves on one edge case, you’re stepping through substitutions one at a time, and the tooling gives you the minimum viable window into execution. A richer trace view or breakpoint-style inspection would have cost the developer little and saved players real frustration on the marathon puzzles. The game trusts you to be your own debugger. Sometimes that trust feels like flattery; on the hard chapters, it feels like an unpaid internship.
Visuals & performance
There is almost nothing to evaluate here, which is itself the design statement. A=B looks like a text editor cohabiting with a test harness: monospaced type, an input pane, an output checker, and effectively no ornamentation. That’s not beautiful — but it is legible, and in a game about reading strings character by character, legibility is the entire job. The presentation never gets between the player and the problem, which is more than can be said for flashier puzzlers.
Performance is a non-issue by design. This is a program that rewrites short strings — a 2D, text-based game of this scope is light on hardware, and a decade-old laptop should handle it without breaking a sweat. The one aesthetic critique that sticks: the sparseness offers zero celebration. Solve a puzzle that took an evening of hard thought and the game acknowledges it with the enthusiasm of a passing unit test. A little more ceremony wouldn’t have violated the minimalist creed.
Platform nuances
Steam is the natural — and only — home, and the fit is comfortable. This is a keyboard-first experience through and through: you’re typing rules, not dragging blocks, so mouse involvement is incidental. Anyone hoping for Steam Deck couch play should temper expectations; nothing about text entry on a handheld improves this game, and it isn’t the intended context. Desktop, physical keyboard, ideally a notepad within reach — that’s the configuration A=B was built for.
The Steam ecosystem also does quiet, load-bearing work here. Community discussions have become the game’s de facto lecture hall, where players trade encoding strategies the way Zachtronics fans trade cycle counts. Given the absence of in-game hints, that community layer is closer to essential infrastructure than a nice-to-have — factor it in if you’re the type who refuses to look anything up.
Final verdict
A=B is the rare game that gets more impressive the more you know about its subject. Artless Games took a footnote from a computability textbook and turned it into forty-odd puzzles that teach, humble, and occasionally infuriate — with a confidence in the player that big-budget puzzle design abandoned years ago. The presentation is a shrug, the difficulty curve ends in a cliff face, and the lack of optimization scoring leaves easy replayability on the table. Those flaws matter, and they cap what the game can be. But the core achievement stands: if you’ve ever felt the specific satisfaction of a constraint snapping into an elegant solution, A=B delivers it at a purity level few games on Steam match — for less than the price of a coffee. We recommend it, with reservations, to the audience it was clearly built for.
Reviewed by the Rankeno Puzzle & Casual Desk. Scores are out of 10; the overall score is the desk’s editorial judgement, informed by the four sub-scores above. How we review