Game review
Canabalt
7.5 / 10
- Developer
- Adam Saltsman (Semi Secret Software)
- Publisher
- Semi Secret Software (self-published)
- Released
- 2 September 2009
- Platforms
- Browser (Flash), iOS, Android, PlayStation Portable, Ouya, Commodore 64, Browser (HTML5)
- Genre
- One-button endless runner
- Monetisation
- Free browser original with no ads, no MochiAds preroll and no single-portal exclusivity — Saltsman let every site mirror it at once rather than selling a sponsorship. Money came later, from a $2.99 iOS release and subsequent paid console/ports, not from the Flash version reviewed here.
- Players
- Single-player only
The short version
Canabalt reduces an entire genre to one input: a button that makes a running man jump, over a city that is falling apart behind him for reasons the game never bothers to explain. That reduction is the whole design achievement — built in five days for a jam whose theme was literally "Bare Minimum," it found a control scheme so immediately legible that it spawned an entire subgenre of imitators within a year and earned a spot in the Museum of Modern Art's permanent collection as a piece of design history, not just a game.
None of that historical weight changes what actually happens when you sit down and play it: one mechanic, tuned close to perfectly, that does not evolve across a five-minute run or a fiftieth one. Canabalt is a genuinely important artifact and a genuinely thin play session at the same time, and being honest about the second half of that sentence doesn't take anything away from the first.
Biggest strengths
- The single jump button is tuned with real precision — timing, momentum and the height of a jump all read clearly at a glance, which is the only way a game with one input can work at all
- Procedurally generated rooftops mean no two runs repeat exactly, a meaningful technical choice for a five-day jam game in 2009
- Danny Baranowsky's score and the stark black-and-silhouette art direction do a lot of work to make a technically simple game feel tense and cinematic
Biggest weaknesses
- There is exactly one mechanic and it does not change: no upgrades, no unlockable routes, no difficulty tiers, so the hundredth run tests the same reflex as the first
- A run lasts however long you can last, and the game gives you nothing to do with a good run beyond a distance number — no persistent meta-progression in the original Flash version
- The premise (why is this city collapsing, who is chasing you) is never addressed, by design, which is a deliberate minimalism some players will find atmospheric and others will find empty
Who it's for: Anyone curious where the "one thumb, one button, procedurally generated obstacles" mobile-game template actually came from, and players who enjoy short, score-chasing arcade sessions with nothing to unlock or manage.
Our recommendation: Worth five minutes for the history alone, and worth returning to for short score-chasing sessions — just go in knowing the entire design is on the table from your first run, with nothing new held back for later.
7.5 / 10
Canabalt asks one question of the player, over and over, for as long as they can keep answering it correctly: jump now, or don't. There is no second button. There is no menu of moves. A man runs, permanently, across the rooftops of a city that is visibly falling apart around him — glass raining out of windows, something enormous crashing through buildings in the background that the game never names or explains — and the only thing standing between him and a fall to the street is whichever key you've bound to "jump." It was built in five days in August 2009 for that month's Experimental Gameplay Project theme, "Bare Minimum," and it is difficult to imagine a more literal answer to that brief than a game with exactly one input.
Made in five days, given away everywhere at once
Adam Saltsman — credited then and largely still known as Adam Atomic — built Canabalt on Flixel, his own open-source ActionScript library, and posted it to his personal site before it spread, unlocked, across Kongregate, Newgrounds, Armor Games and other portals within roughly the same week (accounts of the exact first day differ by a day or two; this entry follows Kongregate's own listed date of 2 September 2009). That mattered more than it sounds: the standard Flash-portal deal of the era was a sponsorship that bought a site a period of exclusivity before a game went wide. Saltsman skipped that, and the decision to let every portal carry it simultaneously is a real part of why it reached the volume of players it did inside a month. There was no MochiAds preroll, no site-lock, and — per Saltsman's own account a year on — comparatively little money in it either; he described the game as having done far more for Semi Secret Software's reputation than its bank balance, even after a paid iPhone port arrived that October at US$2.99. The Flash original, the version this review is about, never charged anyone anything.
One button, tuned until it barely needs explaining
What actually earned Canabalt its reputation is how well that single mechanic is built. The run speed is constant and slightly relentless, obstacles arrive with just enough warning to react and just enough variety in shape and gap that memorization never fully replaces reaction, and the jump itself reads clearly — you can tell, watching the character's arc, whether you're going to clear the next gap or clip it. Rooftops, crates, filing cabinets and window-frames are laid out by procedural generation rather than a fixed sequence, so no run is identical to the last, which was a meaningfully ambitious technical choice to bake into a five-day jam prototype rather than a shipped commercial product. None of it is complicated to describe, and that's the point: a game with one input has to get that one input exactly right, with nothing else to fall back on, and Canabalt does.
The art direction and score carry more of the experience than the mechanic alone would suggest. The city is rendered in near-monochrome silhouette, all sharp black shapes against washed-out skies, with the destruction happening in the background rather than the foreground — you never see what's actually chasing you, only its consequences. Danny Baranowsky's soundtrack pushes the whole thing toward tension it hasn't technically earned from the mechanics alone; strip the music out and you have a fairly plain obstacle-dodging exercise, but with it, a five-second stretch of empty rooftop feels like the calm before something.
Ten minutes of content, played on repeat
Here is the honest complaint sitting underneath seventeen years of acclaim: Canabalt does not change. There is no second area, no unlockable ability, no difficulty setting, no meta-progression in the original Flash version that carries anything from one run into the next beyond your own memory of what killed you last time. Your first thirty seconds and your five-hundredth are mechanically identical. That is a deliberate design choice, not an oversight — it's what "Bare Minimum" as a jam theme demanded — but it means the ceiling on how much game is actually here is low by design, and a reader deciding whether to spend time with it today should know that going in rather than discover it after the novelty wears off. Games that came after it, from Robot Unicorn Attack onward, mostly kept the one-button structure and added the progression systems Canabalt pointedly left out; whether that improves on the original's purity or dilutes it is a reasonable argument, and Canabalt itself never tries to have it both ways.
What it actually started, and whether that's the same thing as being good
It's worth separating two claims that get merged in most retrospectives: that Canabalt is historically important, and that Canabalt is, today, a great ten minutes. The first is not seriously disputable. The New Yorker credited it with more or less inventing the endless-runner template mobile gaming ran with for a decade; it was named among 2009's best games by outlets including Rock, Paper, Shotgun and Eurogamer; and in November 2012 the Museum of Modern Art acquired it for its permanent design collection, alongside a small handful of other games chosen specifically as design objects rather than as entertainment products. None of that is a score, and none of it should stand in for one — a museum can be right that something mattered without that telling you whether it's fun to actually sit with in 2026.
Judged as a play session on its own terms, Canabalt earns most of that reputation honestly. The control feel is genuinely excellent, tuned specifically enough that it still holds up against runners built a decade later with far larger teams and budgets. What it doesn't have is content depth, and pretending otherwise because of the MoMA plaque would be exactly the kind of nostalgia this format is supposed to resist.
Playing it today
The Flash original still runs. Archive.org hosts a playable copy running through Ruffle's in-browser Flash emulation, confirmed working when checked on 5 September 2026, and the game is also present in Flashpoint's catalogue for offline preservation play. Ruffle's ActionScript 3 support — the language Canabalt is written in — is not complete for every AS3 title on the web, but this one has been specifically verified running through it rather than assumed compatible. Saltsman also released the Flixel engine source under the MIT license in December 2010, and in 2024 a fan (with Saltsman's blessing, using his original source) rebuilt the game as an open-source HTML5 port on HaxeFlixel — so between emulation of the original and a maintained native rebuild, this is one of the better-preserved artifacts of the 2009 Flash scene, not one hanging on by a single copy on a dead server.
Final verdict
Canabalt is a five-day jam prototype that turned into one of the most cited design objects in game history, and both halves of that sentence are true without needing to inflate the other. The one-button design is tuned close to perfectly, the presentation does real work punching above the game's five-day budget, and the free, unlocked, everywhere-at-once release is a large part of why it reached the audience that made it matter. What it isn't is a deep game — it's the same ten minutes, repeated, by design, and the difference between "historically important" and "worth a lot of your time today" is exactly that gap. A 7.5 reflects both halves: genuinely worth knowing, and genuinely thin once the history is set aside.
Pros and cons
Pros
- The single jump button is tuned with real precision — timing, momentum and the height of a jump all read clearly at a glance, which is the only way a game with one input can work at all
- Procedurally generated rooftops mean no two runs repeat exactly, a meaningful technical choice for a five-day jam game in 2009
- Danny Baranowsky's score and the stark black-and-silhouette art direction do a lot of work to make a technically simple game feel tense and cinematic
- Released free with no portal exclusivity and no ad preroll, which let it spread across every major Flash portal in the same week and reach the audience that made it influential
- Directly credited, including by outlets like The New Yorker, with popularising the endless runner as a genre that mobile gaming ran with for the following decade
Cons
- There is exactly one mechanic and it does not change: no upgrades, no unlockable routes, no difficulty tiers, so the hundredth run tests the same reflex as the first
- A run lasts however long you can last, and the game gives you nothing to do with a good run beyond a distance number — no persistent meta-progression in the original Flash version
- The premise (why is this city collapsing, who is chasing you) is never addressed, by design, which is a deliberate minimalism some players will find atmospheric and others will find empty
- Saltsman himself said the game earned Semi Secret Software far more press than sales in the year after release, which is worth knowing before treating "hugely influential" as the same thing as "hugely successful" for the people who made it