Last year, Martha Wells published All Systems Red, the first installment of a Tor.com novella series called The Murderbot Diaries, which follows the adventures of the titular android as it is pulled into various adventures while traversing inhabited space. Wells earned considerable acclaim for All Systems Red: it became a New York Times bestseller, and earned her the Hugo and Nebula Awards earlier this year. Since then, she’s published three new installments concerning the adventures of the titular android — the last one came out earlier this month. Individually, they’re an intriguing set of adventures. But collectively, they form a larger story about a machine coming to understand what it means to be human.

All Systems Red introduces readers to Murderbot, a SecUnit (short for Security Unit) android that’s tasked with protecting a group of scientists on a distant world. Things go south when they discover that another group of scientists were murdered, and MurderBot is pretty much the only thing that’s going to keep them alive. Wells elegantly portrays the non-human character, complete with a cranky persona and an addiction to TV dramas.

Spoilers for The Murderbot Diaries ahead.

That first book ends with Murderbot’s wards surviving, and their leader Dr. Mensah buys out the android’s contract, letting it choose to stay with them or go its own way. Murderbot promptly boards a ship headed somewhere else. Over the remaining installments — Artificial Condition, Rogue Protocol, and Exit Strategy — Murderbot tries to steer clear of exasperating humans; thwart an interstellar megacorp that’s breaking some serious laws; and figure out its past, including why it’s named “Murderbot” in the first place.

In Artificial Condition, Murderbot returns to a site where it went rogue and killed a bunch of people, teams up with a research transport named ART, and falls in with a trio of researchers who are trying to negotiate a deal with their terrible employer. SecUnits aren’t supposed to be on their own, and Murderbot’s subterfuge prompts it to think deeply about how to pass as human. In Rogue Protocol, it ends up on a terraforming station on a planet controlled by GrayCris, the unscrupulous megacorp that caused problems in the first book. Murderbot decides to gather some intelligence on what the company is up to, all while protecting another group of humans that get caught in the inevitable crossfire. Finally, in Exit Strategy, Murderbot reunites with Mensah and her crew, armed with evidence of GrayCris’ misdeeds, and ends up in a huge, firefight-filled showdown. From beginning to end, it’s a fun ride. Fortunately, there will be a novel about the character somewhere down the road.

Tor.com’s novella strategy here turns what might otherwise be a longer novel into something slightly different. Each book serves as a self-contained story, and builds on the larger story in incremental amounts. As Irene Gallo, the imprint’s creative director, told The Verge earlier this year, it’s a useful way to take a smaller idea that they might not want to “pad out into something bigger.” That’s apparent in each of Wells’ Murderbot novellas: they’re intimate character profiles as Murderbot bounces from problem to problem.

But what each installment does, beyond just introducing a new adventure for the robot, is get into Murderbot’s head and watch as it changes into more of a person. The androids in Wells’ world are part mechanical, part organic, and Murderbot is particularly different. It hacked its own coding, so it’s a completely free agent, able to reason and form its own priorities. And, while it would rather be left alone to watch the equivalent of Netflix in Wells’ universe, it keeps finding itself doing the right thing: protecting people that it comes to care about.

Murderbot doesn’t simply come off as a rebellious or dour artificial human, though — like Marvin from Douglas Adams’ Hitchhiker’s Guide books, or Bender from Futurama. Wells eschews human-like gendered pronouns and demonstrates just how differently Murderbot sees the world. It can intercept and interpret electromagnetic signals, hack its way into anything, and operate at a frighteningly fast speed. Humans by comparison are slow dullards, unable to keep up, and the android is both frustrated and exasperated by its wards. It’s always clear that while Murderbot might look vaguely like a human, it isn’t one.

But Murderbot also learns to empathize and relate to people and their own priorities, essentially learning what it means to be human in the process. It forms meaningful connections and risks its “life” for the people that it comes to protect and care about, rather than just follow its programming and leaving them to die. Murderbot’s journey is far more compelling as a result. It’s a non-human seeking to operate in a human environment, working to deal with the variety of signals and cues that it gets from the people — and TV programs — around it. And like humans, it’s just trying to get by in the world, making it up as it goes along.