Lab Rat synthesizes comedy and logic into a compound worth testing

Devin Stone, 13 May 2025

It’s a tale as old as time. Drink a little too much on a night out. Wake up in a scientific research facility. Get cattle-prodded through a gauntlet of lateral thinking puzzles by a quirky AI with questionable motivations. The Turing Test. The Entropy Centre. The Talos Principle and its sequel, to some extent. Now Lab Rat. It’s a testament to the remarkable extent of Portal’s influence that games with the same AI-administered lab testing theme are still being released eighteen years later. Even more remarkable is the fact that this one, for the most part, gets it right.

Point one in Lab Rat’s favor is that it doesn’t disguise its inspiration. Both self-aware and genre savvy, the game lampshades its familiar theming early and often.

Turns out some of this game’s developers also worked on Bioshock. Who would have guessed.

Better yet is the clever way Chump Squad, headed up by Kine developer Gwen Frey, twists that theming to modernize it. Where Portal trains its satirical crosshair on corporate science, positioning its AI GLaDOS as the falsely smiling face of industrial greed and exploitation, Lab Rat aims to lampoon something more specific: video games developed by AI. And we’re not talking about some crufty vibe-coded shovelware. Lab Rat’s AI, S.A.R.A, has a fully fledged personality, complete with opinions about game dev best practices, and a healthy respect for QA. Thus, the player adopts the role of not only tester, but playtester. Meta enough for you?

Lab Rat’s fiction might present its puzzles as AI slop, but don’t worry—Chump Squad sacrificed no human jobs in developing the game. The real-life puzzle designer is Lucas Le Slo, known in the Thinky space for his fiercely cryptic, sometimes wordless offerings, often born from game jams. “CONFUSION + FUN”, his itch.io page proclaims, and the wares on display almost all look like E.T. responses to the Voyager Golden Records.

A touch of that opaque, alien sensibility carries over to Lab Rat’s design. In the first puzzle, for example, the player must infer not only how to achieve the goal, but the goal itself. It’s the kind of hands-off teaching that Thinky fans should relish.

Yup, that’s all you get..

Expect nothing as obtuse as Alephant or IFO, however. At its core, Lab Rat is a traditional Sokoban, and discovering the rules should come easily to anyone with box-pushing on their resume. Moreover, while major mechanics don’t always come with a tutorial, the levels between stick to classic stepwise puzzle game values: available actions are explicitly defined, and the only unknown is the order in which to apply them.

Part of me rues this conventional approach as a missed opportunity. I can imagine a more daring version of the game where S.A.R.A. is a misguided fan of rules-discovery puzzlers, trained exclusively on Jonathan Blow lectures and Game Maker’s Toolkit, who clumsily and infuriatingly obscures and subverts her own mechanics in order to “deliver aha moments”. Much meta-commentary could be made here about the relationship between solver and designer, student and teacher.

Instead, S.A.R.A offers mostly genuine, albeit condescending, advice. She teases the player, but never directly opposes them, at least in the first two of the game’s promised six acts. This is, to be fair, what Valve did with GLaDOS—place the AI in narrative opposition to the player, but rarely mechanical opposition. And it’s hard to blame Chump Squad for streamlining things. Judging by the fact that they released on Xbox One in addition to Steam, the team is trying to attract more than diehard Thinky fans.

I don’t mean to imply S.A.R.A’s presence doesn’t brighten the experience. In fact, her accidental(?) snark is, against all odds, genuinely hilarious. Not every joke landed, but enough made me cackle that I always exited the game with a smile. I only wish that delightful theming could have been more integrated with the gameplay.

Thanks a lot, S.A.R.A.

As for the quality of the puzzles themselves, there’s much to praise. At least in the game’s initial two acts, or “training modules”, Le Slo excels at crafting challenging, satisfying brain teasers without relying on too many moving parts. This elegance means that puzzles typically appear impossible until solved, which heightens the astonishment upon doing so.

Wait, green!? You didn’t tell me about green!

My only nitpick here is the game’s failure to tease future depth. This was true to the extent that, when the game did complicate things with new mechanics, I was often shocked. A pleasant, tingly sensation, to be sure, though I’d be more inspired to progress through the game if such innovations were foreshadowed. For example, in my two hours of playtime, I saw no sign of the existence of an endgame meta. Maybe my expectations are too high, but in a world of Void Strangers and Blue Princes, it’s hard not to feel a little unmotivated without even the hint of a culminating challenge.

While the game’s depth is uncertain, its breadth certainly isn’t. Sprinkled between the core puzzles are novel sequences centered on one-off mechanics, reminiscent of Frey's clever weaving of puzzle and narrative in Kine, each parodying a different non-puzzle genre. One such sequence presents as a fantasy dungeon crawler, complete with hammy RPG dialogue and focuses on chain-detonating robotic spiders. Some of these one-off mechanics—like said spiders—were breaths of fresh air, enough that I hope they become integrated into the core gameplay in later acts. Others felt conspicuously extra. Still, the change of pace was always welcome.

Another delight worth mentioning is the in-game analytics, which inform the player of how many of their peers completed each puzzle. I’m ashamed to admit that this triggered my competitive side. “99.9% of players solved this? Welp, I can’t quit now.” I also can’t help but wonder whether the game ever lies with this data. Given S.A.R.A.’s willingness to fib, I wouldn’t be surprised if a gag like this appears at some point.

94.5%? Should be easy, then… Uh…

Lab Rat’s startlingly funny presentation, combined with its breadth of mechanical ideas, makes it easy to recommend to any Sokoban fan. As for depth, more testing is required to determine whether the game’s back half trends toward the mean, or, like AI output so often does, descends into abstract hallucination. Given Frey and Le Slo’s pedigree, I’d like to believe it’s the latter. Here’s hoping the trip is a good one.