Wait. Isn’t this the movie about the rat chef? How can this film be legitimate fodder for a blog about sci-fi interfaces? True, the film is certainly not sci-fi by genre. But it is about over-reliance on an intelligent assistant, and that’s of serious interest to designers, technologists, and anyone paying attention to where AI is taking us. So we’re going to step outside of genre for a bit. It’s worth it.
What Is Overreliance?
Simply put, overreliance is what it sounds like. Relying on AI output too much. (The sweet spot is appropriate reliance.) This quickly becomes a problem because:
- AI makes mistakes (and they spread). LLMs can confidently hallucinate. Accept the output uncritically, and you push those errors downstream where they become bigger and more expensive to catch.
- You lose your skills (making you more dependent). Over time, you become deskilled. If the AI is unavailable, you panic because you now depend on it. This also makes you worse at skeptically reviewing the AI’s output—worsening the cycle.
Brief moment of shill: I cover all this in detail in my new book, Designing Assistant Technology: AI That Makes People Smarter (Rosenfeld Media 2026).
When I look through the sci-fi survey to find examples of overreliance, there’s almost none. But, I fired up this Pixar classic with my younger kid over her spring break earlier this year, and to my delight I discovered this hidden meaning.
The Setup
[Note this summary focuses on the parts that are relevant to this analysis and doesn’t do the full richness of the story or character relationships justice.]
Rémy is a rat with a preternatural gift for cooking. After being separated from his family, he winds up in the kitchen of the late, legendary chef Gusteau. There he rescues a soup being ruined by a new kitchen hand, Linguini. The soup is mistakenly served to a critic, who praises it, and Linguini suddenly finds himself under pressure to replicate something he didn’t actually make.
Under that pressure, Linguini and Rémy form a partnership. Rémy hides beneath Linguini’s toque and steers him by pulling his hair — controlling his body like a puppet. Together they comprise a single chef: Rémy’s taste, knowledge, and technique; Linguini’s body and face. The arrangement works astonishingly well. They quickly become the toast of Paris.
When evidence emerges that Linguini is the biological son and rightful heir of Gusteau himself, his newfound status goes to his head. He takes sole credit. Feeling betrayed, Rémy leaves, and the crisis arrives: A notoriously feared critic has booked a table, and Linguini, alone, has nothing. Stripped of his assistant, Linguini completely collapses, panicking across several scenes—stalling, flailing, bargaining, and in a blind, abject terror. n.b., This is the crisis we’re talking about.
The only thing that saves him is that Rémy returns. Linguini confesses everything to his kitchen staff, most of whom walk out, except chef Colette. Rémy’s rat colony fills in for the missing staff. Together they produce an extraordinary dish that thrills the critic.
The restaurant is eventually shuttered by health inspectors, but the story ends well: they start a smaller restaurant on the outskirts of Paris (with one dining room for humans and a separate one for rats), Rémy and Colette cooking, and Linguini happily waiting tables.
Linguini Over-Relies
Linguini’s crisis moment was, in retrospect, completely avoidable. Linguini had weeks, possibly months, of working alongside one of the most gifted culinary minds in the city. He could have watched, asked questions, absorbed something. He didn’t. To be fair to him, Linguini didn’t come to this kitchen with ambitions to become a chef. He just needed a job. But having stumbled into a situation that required him to perform as one, he made a choice—probably without fully recognizing it as a choice—to go on autopilot and let Rémy handle everything rather than develop any capability of his own.
That choice felt costless, up until it everything collapsed.
This is the most panic-inducing part of the over-reliance trap. The assistant is so capable that critical engagement and learning can feel pointless. Why think through a dish when Rémy already knows what it should be? Sure, it’s efficient. But the hidden cost in buying that efficiency is the skill you’re not building and the judgment you’re not developing. Many people are beginning to call this cognitive debt in the context of AI assistants. When the assistant is unavailable—for any reason, even temporarily—you’re left with the same responsibility despite having let atrophy your ability.
Where the Analogy Doesn’t Quite Map
A few things about Rémy’s situation don’t translate cleanly to the AI case, and let’s be clear-eyed about them.
- Rémy doesn’t hallucinate. He is a fully competent, deeply knowledgeable assistant who goes away at a critical moment of need. The crisis in the film is about absence, not error. The more insidious and common risk is the one where the assistant is available, does give an answer, and the answer is wrong. Ratatouille doesn’t explore that scenario directly.
- Rémy has genuine interiority and goals. Their most serious conflict is about Rémy wanting recognition. LLMs don’t have their own goals—they won’t abandon you for lack of credit.
- Rémy doesn’t defer well. When Colette suggests an alternate sauce, Linguini wants to go with her suggestion for social reasons, but Rémy literally forces Linguini’s hand. In the real world, we don’t want AI assistants making those kinds of overrides when the human has full context.
So keep those differences in mind as we discuss what we might do about this same risk in the real world.
It gets worse with deskilling
If Linguini had any cooking skill before the partnership—and there’s soft evidence he did (his refrigerator has ingredients, not pre-made meals)—he almost certainly lost it during the months of delegation. This is deskilling: the gradual erosion of a human capability through disuse, enabled by reliance on a system that performs the knowledge-based parts of the task instead.
Deskilling is one of the most pressing conversations in AI right now because it feels like AI is getting put into everything, and the benefits and the costs are on entirely different timescales. The benefit of using AI is generally immediate: you mostly get better output, faster, with less effort. The cost side of it, though, is slow, invisible. It might only become apparent in a crisis, or after a boiling-frog awareness that the system has gone from helping you to making you dependent.
We don’t want to have unintentionally caused a generation of users to go into Linguini-levels of panic, realizing they’ve AI-ed themselves into stupidity.
How might we have changed things?
If you’re the Linguini in this situation, you can (and probably should) work with that assistant on your own to make sure some learning is happening. Maybe practicing in your own time, or taking over one dish each day on the job during slow periods. Rémy doesn’t have to go quiet in these moments. He just shouldn’t lead. Let Linguini make decisions, and Rémy can give feedback on those decisions. In the book I call the pattern Human Goes First (which is an even funnier title in this context), and describe a pilot study in which this same intervention turned performance collapse into performance gain inside just two practice sessions. (!)
That tactic is good from an individual user point of view. But If you’re involved in making software, as a designer or product manager, it’s unethical to rely on users doing what’s right, rather than having the system designed against deskilling and overreliance in the first place. You should work to ensure your users are upskilled. To that end, you can implement something like the Human Goes First pattern in your own applications. Occasionally let the users go first in the task and let the AI help them learn when things go wrong. That way people keep up their skills, which helps them get to that ideal of appropriate reliance.
On a Final Note…
Thanks, Pixar, for giving us this pre-LLM touchstone warning about overreliance. And sci-fi writers, since speculative tech is your wheelhouse, I hope to see more of these literacy-building storylines in that genre as well. Help us all not be Linguinis.
Further Reading on AI Analogies:
- Surprisingly About AI: Check out the post on Idiocracy for another film that’s surprisingly about AI without having explicit AI in it.
- The Weekend-at-Bernie’s Trope: For two other films with this trope where one party is animating a technically functional but cognitively absent second party, see Fritzes winners Section 31 (where Fuzz animates a dead Zeph) and M3GAN 2.0 (where M3GAN animates a sleeping Gemma).