Probably the most boring of the interfaces in the story (from an interaction design perspective, rather than an ethical one), these are monochrome monitors showing feeds from hidden, closed-circuit television cameras in places around the castle—including places you’d expect to be private.
Magenta and Riff Raff watch Janet’s bedroom from the laboratory.
Magenta and Columbia watch the laboratory feed from a castle room.
Riff Raff lets Frank-N-Furter know that Rocky has escaped via the monitor in Brad’s room.
Dr. Scott appears at the front of the castle as seen from the laboratory monitor.
The monitors are unusual for being octagonal, but this looks like a frame on a 4:3 television, standard for the mid-1970s. Note that the show that made this treatment much more famous was Battlestar Galactica, which would air 3 years after this movie’s release. (Another potential influence?)
Interfaces
Magenta and Columbia’s monitor has a single dial in the upper right. We never see it in use.
The one interface we see in use on any of the monitors is the arm to the right of the one in the laboratory. It reminds of me of a slot machine arm, but when we see Janet and Riff Raff operate it, they manually pull and return it to the upright position. So it lacks the spring-return behavior of slot machine arms. No, it seems to be a positional lever, and the raised and lowered positions do…something…but it’s unclear what.
Such arms would most befit two states, because the extents of the range are easy for a user to achieve. But there are more than two states.
How do they turn it off and on? (We see the laboratory monitor off and on.) How do they change the channel to view other feeds? (The laboratory monitor shows both Janet’s bedroom and the front door feed.) How do they control the volume?
Then if we were to think of these things for general use: How do they turn off the cameras for privacy?
Could we backworld it?
Yes, it’s possible that the arm position is not binary up-or-down but continuous, with angles defined as having meaning, and timing helping set the boundaries of input. Like, pull it 3° and hold for a second and that puts the system in the select-new-channel state. Or maybe each of the Transsexual keyboard keys is mapped to individual angles, like A = 85–90°, B = 80–84°, and so on. The user would have to move it to the right angle and hold it there for a second for that character to be sent to the processor before returning it to 90° as an “ENTER” function. This has historical precedent in the Quadrant telegraph by Louis Francois Clement Breguet in the mid-1800s. Pictured below is the sending mechanism for such a system.
Quadrant telegraph by Louis Francois Clement Breguet
But note that the usability here is provided by a printed guide below the lever that shows what regions map to what letters, so it’s easy to position and confirm that the arm is in the right spot before pressing the commit control. It would be a terrible experience to try and use this without guides or feedback. It would require too much precision and be very error prone. I can’t backworld this into any semblance of sense or usability. Especially since Janet seems to have no problem using it when she encounters the device.
Before anyone comes at me with, “Maybe Transsexuals have more precision and memory and…”, remember we design for humans, and must evaluate interfaces from that POV.
There is an odd moment when Dr. Scott first appears on the laboratory monitor. The shot is close up on the monitor, so it’s hard to tell if it was due to some action a character took, but the screen goes from staticky to the correct channel showing Dr. Scott. Many modern systems do that—think of how Zoom automatically switches to the camera showing the current speaker—but it would take a vigilant person doing technical direction back in 1975. I’m unprepared to ascribe context awareness to the design of these interfaces, but it would be way ahead of its time if it did. After the close-up shot, a pulled back shot shows Riff Raff returning the lever to vertical. This implies he was there off camera pulling the lever in the first shot, but the film does not show what signal might have had him reaching for it in the first place, or what he did to select the channel where Dr. Scott was shown.
So, in the end, the monitors are a narrative interface, conveying the voyeurism and communication plot moments. Their interfaces are handwaves that enable “character sees thing” beats, and we have to leave them as such.
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.
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.]
The “AI”.
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.
The “user”.
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 assistant, untethered from that pesky user. In the background, a multi-agent architecture. (The rats, not the people.)
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 everything collapsed.
Mounting panic.
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 your ability atrophy. (Or not developing it when you had the chance. Students take note.)
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.
I haven’t been completely honest with you.
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. (!)
In the AI analogy, this should have been a learning moment for Linguini, not a training iteration for Rémy.
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).
Last year I managed to publish very little to the blog in general because life.
This year I hope I made up for it with *…counting…* 14 posts across three months and more than 70,00 words of analysis, neologisms, and celebration. It was a rich collection of movies. I’m not done. I already have more posts and analysis scheduled for later this year.
While I did all of that, I got caught up in making some improvements to the site.
I fixed a lot of special-character typ0z and missing images. (n.b. Many, but not all.)
I got rid of a lot of “update” posts—as opposed to that aren’t really relevant now. (Even this one may get chopped in time.)
Claude helped me write some WordPress shortcode for forward-looking links that announce in text when a post is scheduled, and that automatically convert to a link once it goes live. This saves me from having to remember to go back and update links.
I also switched to a service to let readers know where a given show can be watched, and it is true as of the moment a page loads, so no *womp womp* links that raise false hope. (Sadly I couldn’t filter out awful billionaire sites from their lists, but you can’t win them all.) Look at any of the Report Card pages to see these in action. They’re usually right at the bottom.
I created a Glossary page of all the nerd terms related to the ongoing discussion.
I hope those improvements make using the site better.
Oh, and we now have final awards tally for the 2026 Fritzes! It and the others are linked on the main Fritzes page should you want to go back to any prior years’.
2026 is looking to have a number of interesting interfaces already, but of course if you see something that you don’t want me to miss, hit me up. You know where to find me.
With the fascist-billionaire media-capture tactic underway in the United States, entertainment that smells at all “woke” to them is being summarily lined up against a brick wall and shot. Star Trek is not immune, and with Paramount’s takeover by asshat-in-chief David Ellison, every Star Trek property has been canceled and their sets dismantled. The episodes for Season 2 of Starfleet Academy and Season 5 of Star Trek: Strange New Worlds have already been shot, so they will probably be released. But that will be it. I would not be surprised if every single Star Trek fan canceled their Paramount subscriptions and found other ways to watch those final episodes. It would be just desserts.
Given all of that, I don’t expect we’ll see a Section 31: Party on Turkana IV sequel anytime soon. This may be the one and only Section 31 movie we get until this ugly fascist phase is behind us. If you’re a US American, you’re registered to vote, right?
If this is the last Star Trek show we see for a while, it’s not a bad one to have as a bookend. I mean, it’s not perfect. (Did no one realize that even if San got back to the Mirror Universe, he’s not biometrically authorized to use the MacGuffin? Like, the last third of the movie could have just been our characters chatting over drinks in Baraam, collectively shrugging.) But the interfaces are worth celebrating, and that’s what this report card is all about.
Sci: F (3 of 4) How believable are the interfaces?
With the exception of the Godsend, they are completely believable, given the novum stack of the 24th century Star Trek world. The mission briefer’s issues are largely social, not technological. The phase pod makes sense and is a pretty good wearable measured against the criteria. Fuzz’ interfaces—the displays and the programmable-matter lever he pulls—are largely what they should be. The Federation interfaces look like they come from the same place, as do the scow’s, and the Terran ones.
And let me be clear—if you have to make one interface that doesn’t model efficacious design, a weapon of mass destruction would be it. I’m very glad that one isn’t believable.
Fi: D (4 of 4) How well do the interfaces inform the narrative of the story?
Here’s where the Godsend should shine and does, being the MacGuffin that drives the action and even amps up the tension of the climax. The phase pods help us understand what’s happening throughout a fight that happens between worlds. Fuzz’ interfaces sell us on how a being of such tiny scale can be a part of—and betray—a human-sized team. The high contrast, saturated-red-on-black interfaces distinguish the Terran environs from Section 31’s and the scows. The Droom doll (described in Best Robots) tells us some deep things about a culture we never meet, all in just a few seconds.
Interfaces: F (3 of 4) How well do the interfaces equip the characters to achieve their goals?
Georgiou wants to nab and keep the mysterious suitcase, and the phase pods help her easily meet and fight San whatever plane he’s on. We only see one of Fuzz’ interactions, but it’s perfectly designed for his need. And his whole plan depended on agents he could control at a distance. Despite not knowing the language, Quasi is able to pilot the scow, use it to delay San’s passage through the portal, and drop a bomb on a ship in hot pursuit.
Final Grade B+ (10 of 12), Must See.
Star Trek is a cultural force. In the 60 years since the original series aired, it’s become a cultural touchstone around the world. Its warp drive will fire up again, sometime. We’ve already seen it crop up under different guises across the years, and it was still on air.
It means too much to too many people to have some hoarding-class religious-political extremists ground it forever. They think they are burying it, but they are planting a Phylosian seed.
As part of the FritzesBest Interfaces award for 2026, I am reviewing the interfaces in Star Trek:Section 31. This post is about the quadrant-destroying weapon of mass destruction called the Godsend. Note this blog generally eschews analysis of weapons, but this one is more MacGuffin than blueprint, and it has the worst interface in the film.
The Godsend is a weapon that Georgiou had created when she was Emperor. It is meant to function as a scorched-earth deterrent to her enemies.
“It triggers a chain reaction, like a virus passing between planets. Everything in its path incinerates. An entire quadrant would be lost.”
To her credit, she says she ordered it destroyed, but it was secretly stored by San and later brought to Prime. It is a metal object, roughly a sphere, and slightly smaller than a human head. Around its “equator” it has a smooth belt punctuated by 10 mistily-glowing circles. The hemispheres outside this belt are faceted. There are lots of flat nurnies and greebles on the surface with no obvious purpose. They’re not even aposematic, which would be appropriate.
When Georgiou and Sahar teleport to San’s ship, combat ensues, and in the fray San accidentally knocks the Godsend off its pedestal. It hits the floor and malfunctions, exhibiting a complex set of behaviors I’ll just call the “tick”:
We hear a mechanical clockwork ticking.
The belt rotates a few degrees clockwise (as seen from the north pole).
The upper hemisphere rotates a few degrees counter-clockwise.
One of the white circles turns pinkish-red.
(I know that north and south are arbitrary conventions here, but it helps with the description.) After the tick is complete, its computer voice says, “Detonation sequence activating”. The voice is low, raspy, and appropriately menacing.
After a beat, it ticks again. A second circle turns red, and the voice says, “Awaiting biosignature confirmation”. Amidst the ongoing fighting and ship careening, the Godsend gets kicked around a lot and, at intervals, continues to tick.
San and Fuzz are defeated, and as the ship nears the portal, Georgiou picks the Godsend up off the floor. It ticks again. She places her hand on the “north pole” for about two seconds.
The remaining white circles turn red, and the voice says, “Biosignature confirmed…Detonation in 60 seconds.” It announces again at the 30 second mark and continues to rotate at intervals. The voice warning comes again at 10, and then each second from 5 to 1. At zero there is a blinding light as it explodes just inside the portal on the Mirror Universe side as Georgiou and Sahar beam back to safety on the scow.
It arms accidentally? From being dropped on the floor? That can’t be its intended operation, so, a quadrant-destroying weapon of mass destruction was just, you know, poorly engineered? No one thought that this heavy, spheroid, metallic object might ever slip out of a hand? Or was it sabotaged like the Death Star, adding this flaw somewhere along the engineering process? Let’s hope that saboteur also immediately fled the quadrant afterward, taking along…I don’t know…every single one of their loved ones with them, along with all the innocents who might get incinerated in the blast? What size getaway ship were they working with?
Next, why is there a countdown for a detonation sequence that still requires authorization? What would happen if the detonation sequence completed without being authorized?
If nothing, then the countdown is just a goofy, extradiegetic tension-building function.
If something, shouldn’t the voice alert the user to those stakes?
Why is the countdown visualizer spread in a ring around a sphere? That makes it entirely possible that those critical signals are hidden from view for about half the time they are relevant. And they’re inset, meaning that even when looking at the facing side, at most three of them are clear. We will just see slivers of the other two. The design hides most of the visual part of the countdown from view.
Either the countdown isn’t underway here, or it’s halfway through. Who knows?
The choice of authorization (two-second hand on the pole) is easily understandable by the audience, but seems really, really prone to accidental activation. The pole is how one might, you know, carry it, or hold the damned thing while dusting the shelf underneath it.
One of the key principles for deterrents (we got “good” at this during the Cold War) is automaticity. If the one person who can trigger it can be killed before they activate the deterrent, then it’s just a tactical exercise: separate the authorizer from the device, or assassinate them quickly before they can activate it. If it’s biometric, tactics can be just making sure that body part is destroyed first. Both of these interventions are possible given the design of the Godsend. Really it should have a dead man’s switch, not an activation trigger.
If it was left as an activation trigger, the biosignature long-hold is the moment that a countdown is relevant. It would give the carrier a beat to think, “Oh, gods, no. I was just cleaning!” and reposition their hand for safety before going to change pants. The moment her hand is in place, the device should then signal a countdown in a way that it is undeniably perceptible to Georgiou—no matter in what orientation she is holding it. And it shouldn’t just be visual with intermittent audio, as we hear in the film. The audio should be constant, visuals should be on every side of the device, it should provide rising haptic feedback, and reach out to all nearby computer-controlled actuators to have them broadcast that everything’s about to be borked, send a last 🩷 SMS to your loved ones. Having it announce that it’s going to blow after a silent long-hold is very, very bad design. We can argue security through obscurity here, but the cost of accidental activation is far too catastrophic.
Maybe the thing that’s been keeping the Prime Universe safe all along from the fascists in the Mirror Universe is that they’re terrible designers and rotten engineers. It is a testament to how much I like the other interfaces that this one didn’t drag the rest of them down with it, because it’s just an immersion-breaking misery.
As part of the Fritzes Best Interfaces award for 2026, I am reviewing the interfaces in Star Trek:Section 31. This post is about the interfaces used by Fuzz.
Fuzz is a Nanokin, a species of microscopic, squidlike beings with impressive, tiny spaceships. To engage with his teammates in the human-scale world, he does so by flying into a black-market android built to look like a Vulcan, and controlling it from within. In the film they call both the android and the Nanokin “Fuzz”, but that would get confusing in writing, so I’ll call the android the Vulcanbot. I want to believe that the character concept began as a tardigrade or amoeba, but it got more octopus-like over development. From its tiny spaceship, it can get through tiny holes and cracks in machinery or body modifications, hook in, and cause plot-critical mischief.
When the camera is at the nano-scale, the film uses tilt-shift and floating-particle techniques to emphasize the smallness of Fuzz. That means that only a small strip of things are in focus in any given shot, giving us less visual information to work with than usual. So though I’ll cover it, know I’m working with a lot less than I might ordinarily have.
Nanoship
The ship he flies around in is roughly spherical, and about ten times his own diameter. It kind of looks like him, which is both a funny and philosophical design choice. Its surface ripples in waves similar to the surface of the unnamed Section 31 ship that Sahar pilots above the safehouse planet. I think the implication is that it is made from programmable matter.
It has retractable, tentacle-like appendages coming out from the hull. They can be extended to surfaces to hold the ship in place and interface with electronics. I counted 20 tentacles in one screen shot, but if they’re programmable matter, they can be made ad hoc.
The interaction design question is how these are controlled, but, with programmable matter, general artificial intelligence, and agents all part of the novum stack for the movie, it might be as simple as a prompt: “When you are near safe access points, create connectors to them.” Since it’s never shown in the film, though, we have to leave it as a guess. I leave it as an exercise for the reader to imagine how it might work with a modern technology stack.
There is a curved viewport at the front of the ship, subtending around 120° from the pilot’s view. Additional displays to the left and right of the viewport extend the display surface to around 180° degrees. The viewport features an augmented, highly dynamic display, able to show live video, star charts, big red labels, waveforms of audio—whatever is needed in the moment. Language in the display is both English and Nanokinese (for lack of an official known name of that script in the lower left). Stylistically it has a cyan border with white contents, with dusty lavender highlights. Semi-randomly-wandering line segments appear throughout. Sadly, we do not see Fuzz futzing about with this interface at all, so we cannot evaluate that part of it. But it is the context of both the nanomap and nanolever, discussed below.
Nanomap
A curious element in the center of the volumetric projection console is that of an edge-lit, standing human figure with a transverse ring around the waist. It is always there and does not appear to change throughout the film, regardless of the position of the body he’s in or controlling. It might serve simply as a map of the current body-in-question for alert and display purposes. Stuff like wayfinding or a damage control diagram.
We don’t see it when Fuzz is in Zeph or Dada Noe, but it would be cool if we saw it change to match the current host. Even cooler if we saw some vague indication of the surroundings around the host. Even coolest if we’d seen one virtual body for Vulcanbot and a second one for Zeph on the dashboard when Fuzz had the ability to remote control both.
Nanolever
When Fuzz’ deception is figured out by Georgiou and his Vulcanbot is face-to-face with a phaser, Fuzz grabs a lever and pulls it toward himself. In response Zeph’s corpse—controlled by his mechsuit—begins to rise, again under the control of Fuzz.
The lever is interesting for two reasons.
First, it’s the only physical control visible we see in the ship. (Fuzz has his tentacles raised above the viewport in a number of scenes, but the shot is from the outside of the ship, so we don’t know if he’s operating controls or just bracing himself.) A physical control is persistent and can’t get lost in occluding windows of a digital display. This tells me that Fuzz knew he might get exposed, and might need to pull the lever at any moment to initiate his ace-in-the-hole plan. The physical lever facilitated that much better than a digital one would.
Second, look at the physical design. It is textured and curved. These are both features which make it easier for octopus arms to grasp and manipulate. (I’m not a cephalopod expert, but this study says so.) We don’t know if Fuzz’ tentacles function similarly to octopus arms, but it’s a reasonable place to start.
I have less confidence in the two rings at the top of it. A shopping search for “lever controls” shows that none of them feature rings or holes. I’m not an industrial designer, but having those rings seems error prone. Not to grip, but to release. If your fingers or tentacles are in those rings, and some emergency situation requires you to quickly grab something else, you might be critically delayed by the fine motor control required to withdraw from the rings. If the lever is just a stick, releasing is practically a non-issue. So I’m less fond of the rings. If you can think of a good reason for these, let me know in the comments.
An Agent!
Since I started thinking in-depth about agentive technology, I’ve been noting when I see them in sci-fi. It’s rare. Up until Fuzz, Dr. Strange’s Cloak of Levitation has been my go-to example. Literacy in agents is becoming more important over time, and popular media is one way that people learn about it. (Especially its risks.) I was delighted to see a plot-centric use of them in this film.
Look close and you’ll see “CONVEYANCE AUTOPILOT ENGAGED” across the screen.
Vulcanbot is an agent while Fuzz is in Zeph, and then Zeph-corpse is an agent as Fuzz is fighting Georgiou to escape. Vulcanbot even handles the b-plot battle with Sahar before being caught in the climactic explosion.
This literacy of what an agent is and what it’s capable of is critical to the protagonists’ fates. If Georgiou hadn’t sussed it out, the team might have split up from unresolved suspicion. Fuzz would have snuck away and San would have returned with the Godsend to the Terran Empire and used it to return and conquer Prime. So her agent-literacy saved the day.
The central role this agent played in the film is one reason I really loved it. Of course even more interesting would have been to see how Fuzz expressed his commands for the agents and monitored their performance against those goals, but because this needed to be hidden for the Big Reveal, we don’t get to.
A missing signal
One important feature that is only weakly implemented in the Vulcanbot and should be stronger when we implement similar technologies in the real world: Agent-mode signals. These signals would convey to observers whether the technology is being operated by a human sentience or when it is being driven by agentive software.
Of course Fuzz is deeply vested in deception. Vulcanbot acts a little strangely when in agent mode, but it’s because the AI is not rich enough to mimic Fuzz on autopilot. It’s easy to imagine that if it could have been a perfect mimic, Fuzz would rather that.
But for us in the real world we want to know what we’re dealing with. It changes how we interact and what our expectations are. I argued for these deliberate design interventions in the context of Google Duplex way back in 2018, just not on this blog. So let me assert them here. A more ethical Vulcanbot would shift to a modulated voice as a hot signal when it was operating agentively, and interject a cold signal when circumstances called for it.
Delicious woke
Star Trek has addressed queerness before. I’m glad to see it again, considering how the weird MAGA Trump-suckup regime is trying to villainize and scapegoat trans people like the Nazis did with Jewish people here in my home country. And, to be clear, fuck that nonsense.
Though there’s a diegetic “excuse” as to why it is, the perceptual truth is there’s something invisible inside a character that has us accepting a masculine version for most of the movie, and then accepting a feminine version at the end. Same body, different behaviors, sci-fi reason.
There’s just something inside that informs who this character is and how they behave, even if it doesn’t match your expectations from the outside. Best not to think too much about it.
The rationale is there, so the queer-o-phobes don’t have a good excuse to reject it outright. Diegetically, the invisible part is binarily gendered. Diegetically, that’s what informs the Vulcanbot’s outward behavior, not *gasp* actual genderqueer-ness. It’s fantastically designed for the right kinds of cognitive dissonance.
Perfect for Pride Month. Maybe we can have Nanokin as a teeny tiny marshal for the next sci-fi Pride Parade.
When Georgiou escorts Noe into his hotel room, she activates and tosses a device onto the container he’s carrying. It’s a palm-sized, metallic saucer-dome shape with intricate detailing, and sports three rounded signal lights that glow white. It magnetically grips to the surface of Noe’s container. He cautiously asks what it is and she pulls out another one to show him, explaining it is a “phase pod”. She attaches hers to her belt, and a beat later its white lights turn green. Her appearance becomes blurry and shifting and there is an audible low purr. A beat later the lights on the pod attached to the container turn green, and it gains that phasing appearance as it slips from his hand to land with a thud on the floor. He tries to shoot her with a phaser, and the blast passes through her to hit the wall behind. He tries to hit her physically and passes through. When San enters through the wall a little later, we see a similar pod attached to his belt and the phase-fight begins.
There’s a moment where Georgiou pulls her phase pod off to render herself immune to the knife he’s slicing at her while phased. San pulls his off to reengage her in unphased-space. A beat later we see them crash through the glass window separating the room from the nightclub floor.
When the fight takes them both onto a raised dias, Georgiou taps her pod to turn it on again. We see its lights and the lights on the case pod instantly turn green. She’s phased. San turns his on again, too, and the fight continues. About midway through the fight, San stabs her pod. He doesn’t quite disable it, but it starts to malfunction, its status lights flickering between green and white.
San drops the case and kicks it through a wall. Georgiou tries to run through the wall to retrieve it, but the malfunctioning pod fails to phase her left shoe. It “catches” and won’t pull through the wall. The status lights of the pod flicker green and red. In frustration, Georgiou smashes the device a few times to no avail.
Caught, she watches as San appears, opens the case, removes the Godsend, and teleports away. Too late, the pod sorts its shit out, and Georgiou is able to pull her foot through and stand up to be confronted by Sahar.
Caveat: Some things are unexplained, here
It’s not clear how the floor holds a phased thing, while the walls do not. Maybe it’s some aspect of the artificial gravity? It’s also not clear why Baraam doesn’t have the equivalent of a red alert when the fight starts, which would prevent San’s beaming away. Most casinos have outrageous levels of security and Georgiou is paranoid, so one would expect it. These are script questions, admittedly, not related to the interfaces, but things that the skeptic in me must voice.
One production gotcha
Before they reach the dias (below) we can see an unphased Georgiou holding the case. Its signal lights are green, but since she’s holding it, it must be unphased, too, which breaks one of the two diegetric rules already established.
Green means Thing is Phased.
Phased and non-phased things do not interact.
Rather than trying to backworld this (which would get complicated fast) I’m going to presume this is just a production mistake.
Evaluation as a wearable
On this blog I’ve established some guidelines for what makes a good wearable. I’ve used those to evaluate this interface.
Sartorial? Yes.
Yep. Palm-sized, one flat side, other side rounded. Lovely textures and shape. All fit being worn.
Easy to access and use: Tap to activate
I wonder about accidental activation. The fight has a lot of bumping around. If it’s a dumb momentary button, capacitance sensor, or accelerometer, some accident of the fight might accidentally turn it on or off, so let’s presume it’s not that. If it’s a biometric signature requiring the authorized user to tap it, it still seems riskily “out there”, ripe for an accidental touch that could have the wearer slamming into a wall they thought was passible, or dephasing in the middle of something. A long-touch, as seen with the mission briefer, might make more sense. Long-touch trades discoverability (how is anyone supposed to know to long-touch a thing) for increased certainty (accidental long-touches are less likely than accidental brushes.)
A long-touch introduces some challenges mid-fight, but I suspect this is primarily meant as an infiltration tool, not a combat one.
Social and Apposite I/O: The glow is almost fine
You might think that having the glow there gives too much away, diegetically. It would draw the attention of onlookers and raise suspicions if not alarms. Generally speaking, a covert wearable announcing itself brightly runs counter to its purpose. But the visual and audio effects of the underlying tech are far more conspicuous, and I presume, unavoidable. So subtlety is not an option using this tech. And the signal lights help convey to the user the states we see.
Status
Display
Off
Dark
Ready
White
Active
Green
Damaged
Green and white
Error
Green and red
That’s quite useful, even if they require a glance down. There’s an additional status which is “these two pods are paired” which might be accomplished with a synchronized blink, but that might also give too much information away to assailants.
As we see in the film, one problem is when phase-ees want to kick other phase-ees out of phase-space. Then the light on the surface provides a helpful signal of exactly where to target. Let’s trust that the jerky visual effects we see from our non-phased vantage point are still in effect when phasing, so that little target’s going to be jumping around anyway. Also, as long as all pods have the same lights, it’s not granting an advantage to anyone in particular. Georgiou could just as easily targeted San’s pod.
Haptics, probably
The use case of most concern for me is when the pod is malfunctioning. It looks like the tech prioritizes its effects for living matter, so bodies don’t unphase inside of solid matter. That would make for graceful degrading that could inform the user that the tech is trending in an unsafe direction. “Hey, why is my scarf stuck. Oh crap!”
We can’t rely on the phasing side-effects, because those seemed to continue as usual even during the half-functioning. Whatever that signal is shouldn’t rely on her looking at the device, either. A haptic feedback seems most fit, and specifically where it vibrates when the device is working, pulses when it is half-working, and buzzes or goes still when it is not working at all. Since film generally does not convey haptic feedback well, it might actually be part of the device we see. That would explain why Georgiou was willing to risk passing through the wall. The stakes were high, and she could tell, without looking, there was a good chance she would make it.
Between the tap that should be a long-tap, the lights that clearly signal mode, and the haptics that we’re going to presume are there anyway, the phase pod makes for a plausible wearable that meets basic usability and story needs, too.
As part of the FritzesBest Interfaces award for 2026, I am reviewing the interfaces in Star Trek:Section 31. This post is about the mission briefer.
When HQ needs a team to get moving, they send a mission briefer. (n.b. Ths is my term. They don’t mention it by name in the movie.) This little faceted matte-black pod is the size of an orange with one flat side. Rest it on a surface, and when an authorized person long-touches the top, it spins open like a lotus flower. A lens rises up and emits a holographic video projection above it with mission information. The projection has a highly pixelated translucent appearance. The movie begins with the decontextualized briefing for the pre-Georgiou team, and ends with the final team standing around a table in Baraam, receiving a briefing for a mission that will take them to Turkana IV. (!)
One excellent design aspect is that there’s no indication from the outside what it is or how to use it. Ordinarily of course we designers work hard to make sure use is clear to the novice user, but in this case obscurity is security. No rando off the street should be able to figure out how to open the top secret clearance container. This aspect might be even better if it looked and functioned like some other mundane object, so that said rando wouldn’t even suspect it was worth investigating. But that introduces other risks and complications, and for an object that is not quite plot-critical, would require too much screen time to explain.
Otherwise I have some minor questions about the device. Each of these can be dismissed as “well, it’s really high tech, you see”. Sophisticated tech is a plausible explanation, but that’s the unsatisfying “a wizard did it” answer that doesn’t help us with design lessons.
Shouldn’t it have strong multifactor authentication?
I suspect all briefings contain highly-sensitive information. And sure, we can give it the benefit of the doubt that mere contact provides a biometric signature “something she is.” We should see some indication that she provides one of the other two: Something she knows like a password or something she has like a combadge. (I’m not a security expert, but I believe holding the briefer itself might count as “something she has” but it’s a prohibitively weak authentication factor since it’s coupled to the content.)
Isn’t orientation a problem?
This one’s tiny, but how does the projection get oriented (yaw in this case, since pitch and roll are handled by the surface)? Sahar seems to fuss about its placement on the table, but the device looks the same from all angles, so I’m not sure that what he was doing was orientation. In the end scene, the projection is just of a person talking, so the orientation is not critical. It might be awkward for a projected person to be facing directly away from the listener, but not significantly hinder the information. But in the opening sequence, there is text and maps and lots of 2D information, which would be made significantly difficult to interpret if it was backwards or off-facing.
It seems silly to comp up pinpoint lights, but here we are.
Of course, it could have built-in tech that finds where the team is around it, and calculates the optimal display yaw. If we had half a second after the long-touch of tiny glowing bits around the base that demonstrated it finding them and thereby the optimal orientation, it would telegraph this feature. (See above.)
Is everyone supposed to be able to see it?
The team is watching the mission briefing in the lower right. As is anyone else at the bar, I guess.
As you can see in the wide-angle shot, the team is just watching the brief with the briefing agent in the nightclub of Baraam. It draws attention. Can’t anyone just glance that way, record it, and sell the information to the highest bidder in the underworld network? That can’t be secure. If it was just projecting into the team’s eyes, ears, or brains, that might be secure, but the film would need to change that wide-angle shot to indicate that. Projection beams or something. Somehow it should signal how this isn’t just broadcast for any eavesdropper to pick up.
What if the team has questions?
I’ve never seen this in a mission briefing in present-day spy thrillers, but there’s an opportunity here since we’re dealing with very advanced technology. If the briefer has a knowledge base, then the team should be able to ask questions of it. Clarifications or additional detail. If it was driven by something like a large language model, rather than a recording, then it could be interactive, and there could be a question and answer session at the end, and serve as a just-in-time reference during the mission, too. (c.f. related concepts in the real-time interplanetary chat post.)
Again, these are nit picks, as it hits the narrative beat just fine. It’s a prerecorded message that plays and tells them what they need to know. (And Jamie Lee Curtis!) Anything else would be gravy.
The Fritzes award honors the best interfaces in a full-length motion picture in the past year. Interfaces play a special role in our movie-going experience, and are a craft all their own that does not otherwise receive focused recognition.
Today we’ll be covering Best Interfaces. The movies nominated for Best Interfaces manage the extraordinary challenge of being believable and helping to paint a picture of the world of the story. They advance the state of the art in telling stories with speculative technology.
The winner of the Best Interfaces award for 2025 is Section 31.
As you’ll read below, my posts on the winner will be a series rather than a single post, so let me do one Also Check Out here.
Bust first, also check out: Superman
Though I have some issues with the amount of fuigetry in most of the screens, and how Lex has to call out countermoves rather than have an assistant offer next most likely countermoves; the robots in the Fortress of Solitude and the crazy-cool gestural control of his spheres by Mr. Terrific make me think that interfaces and tech will not be an afterthought in DC’s new Gunn era.
(James: reach out and I’ll send you a free copy of my book about assistants, it would have helped with that Luthor interface.)
The 2026 Best Interfaces Award goes to Star Trek: Section 31
Maybe I was out of the loop, but I don’t recall hearing a lot of buzz about this movie at the time it came out. But when I finally caught it, I was impressed with the breadth, the art direction, and some interfaces of a sort I don’t think I’ve documented before. This year I’m going to honor the winner with an old-school breakdown, interface-by-interface. In this post we’ll start with a general overview, and then move to the Mission Briefer.
Note I try to only describe just enough so the interfaces can be understood, but since this is a cloak-and-dagger spy thriller, it’s still pretty intricate.
Plot overview
In the Mirror Universe of Star Trek, the mostly-good United Federation of Planets doesn’t exist. Instead it has a morally-inverse counterpart called the Terran Empire. Philippa Georgiou became ruler of this evil empire in part by defeating and enslaving the ambitious contender San. Once ascended, she exercised cruelty and ruthlessness until sci-fi shenanigans landed her in Prime Universe (the home universe of the shows), in 2257.
[Here I bypass a lot of stuff that happened in Discovery for the sake of brevity.]
Eventually she takes up an alias as “Madame Veronique du Franc”, proprietor of the pleasure space station Baraam, outside Federation territory. Section 31—essentially the Federation’s black ops—sends a team to blackmail Georgiou to help them intercept a superweapon, which happens to be en route to Baraam in the hands of a shady middleman named Dada Noe.
The team consists of their superstrong “augment” leader Alok Sahar, a mech-suit guy named Zeph, a seductress named Melle, a shape-shifting genius called Quasi, a buttoned-up Federation overseer named Garrett, and Fuzz, a microscopic Nanokin who pilots a teeny tiny spaceship and most often inhabits a black market Vulcan-looking android body.
Using some technologies called Phase Pods, Georgiou successfully separates the superweapon from Noe only to have it intercepted by a masked person also wearing a phase pod. Masked person kills Melle and escapes with the superweapon, but on the way Georgiou learns it is the Godsend, a quadrant-vaporizing weapon she had commissioned when she was Terran Emperor.
Georgiou convinces Sahar to form a partnership to recapture the weapon. They beam to his spaceship above a desolate planet where they interrogate Noe. They learn Noe is from the Mirror Universe, where he administered a facility that housed the Godsend. He hatched a plan to sell it and with the money escape to Prime Universe to retire in peace. His portal is an unknown but routinely opening rift between worlds. He tells them he is scheduled to meet his anonymous buyer when the rift next appears in four hours. He expects that if he does not deliver the weapon to his buyer—and the Terran Empire learns that the Godsend is gone—they will trace it to the rift, surge through, and conquer Prime. At that moment a massive explosion rips through the ship. The computer automatically beams the crew to the surface, but Noe dies in fiery debris. (Narratively convenient, but consider that the ship’s computer knew enough to beam our protagonists to safe, non-fiery-debris places, raising the possibility that it chose to murder Noe.)
Sahar says the explosion was sabotage by someone working with the still-unknown Godsend thief. One of them is a mole! Accusations fly, but Garrett focuses them on finding a derelict garbage scow she knows about, as a means to continue their mission. The team splits. Georgiou, Quasi, and Fuzz search for the scow. Sahar, Garette, and Zeph work to repair an antenna in an old Section 31 safehouse so they can warn the Federation of the impending danger.
Team Scow repairs the ship. We see Fuzz behaving a little strangely.
Meanwhile Zeph skips out on Team Antenna. While Sahar and Garrett search for him, the antenna gets activated, some message sent, and then the antenna is destroyed. The whole team rejoins and begins a search for Zeph. They find him dead. When they recover the video from his mech suit, they see something was controlling his suit and made him kill himself. Georgiou reasons that the mole must be Fuzz, who left his Vulcanbot on autopilot while he flew to Zeph to hook in and control him to commit the crimes and fly back to his bot. Thusly busted, Fuzz takes remote control of Zeph’s suit (grossly with Zeph’s corpse still in it) and the two try to escape on a float. The rest of the crew pursues in a second float, and there’s a vehicle combat sequence. Fuzz tells Georgiou that he’s been working with San. Then San beams Fuzz up to his ship. San speeds toward the rift to tell the Empire everything and begin the invasion. The remaining team gets the scow running and gives chase.
They catch up near the rift and the scow tries to delay its entry into the rift via tractor beam. Sahar and Georgiou beam to San’s ship to learn that San has initiated the Godsend. San fights Georgiou. Sahar fights Vulcanbot while Fuzz escapes to watch from a safe distance. On board the scow, Garrett forges a makeshift weapon in the ship’s hold and they release it at San’s ship. It lands and explodes, giving the heroes the upper hand in their respective fights. Georgiou grabs and activates the Godsend via biometric signature. Quasi manages to beam her and Sahar back to the scow just in time, leaving San, Fuzz, Vulcanbot, and the Godsend to be destroyed in the explosion as it passes back to the Mirror Universe and seal the rift forever. (And, presumably, something about the confluence of energies neuters the Godsend so it doesn’t go on to kill quadrillions in the quadrant where the rift happened to be, because that would be multiple, multiple genocides and sully whatever victory this is.)
The movie ends with the team back on the Baraam. They meet Wisp, Fuzz’ widow, piloting a second bootleg Vulcanbot. They receive a mission briefing that has them warping the Baraam (surprise, it’s also a spaceship) towards Turkana IV.
Whew.
Star Trek: Section 31 is primarily set in the 24th-century “Lost Era” between 2324–2326. This places it roughly 66 years after Discovery (2258) and about 40 years before The Next Generation. For continuity, the designers have to find some middle ground between the glowing, 3D, multiplanar translucency of Disco and the flat, 2D, highly-graphic, vibrant oranges-and-blues palette of LCARS. I think they did a really nice job. We see circular, glowing interfaces. We see hints of the fully realized LCARS to come.
There’s also a clear delineation between Federation/Section 31 interfaces, the mining colony interfaces, the few Terran Empire interfaces we see, and those of the foreign-language garbage scow.
San’s ship interior, by the way, is high-contrast red-on-black, and very pointy, making me wonder if the production designers have read my post on the Design of Evil, because it is practically an archetype of those patterns.
Anyway, now that we have a grasp on the plot, next let’s look more closely at those interfaces.
The Fritzes award honors the best interfaces in a full-length motion picture in the past year. Interfaces play a special role in our movie-going experience, and are a craft all their own that does not otherwise receive focused recognition. In this post we round-up all the big labels across the survey.
I wrote about the Big Label in 2012 in the Visual Interfaces chapter of Make It So…
…and here we are 12 years later cataloging more. It’s understandable why: It’s familiar, it conveys critical plot information in a fraction of a second. It’s inexpensive because it’s fast to design with not a lot of moving parts. It’s not quite an interface, since it’s just output, but there were enough this year to catch my eye. So, uh, here they are.
