Conversations with Large Language Models: Battle Decks
I have trouble believing I am the only person who has thought to have a large language model play battle decks
, which Wikipedia describes as an improvisational activity in which a participant must deliver a presentation based on a set of slides that they have never seen before
. Lots of other people must have thought about this already but I never hear mention of it. Maybe I just travel in the wrong circles.
I finally got around to trying it the other day just to see what came out the other end, using my talk about generative systems and the content deficit
from the Museum Computer Network conference last year as the input. I tried this first with the open Llava 1.6 model but it wasn't able to get past the fourth slide before getting stuck and repeating the same thing over and over again. Eventually, I held my breath and used Microsoft's CoPilot reasoning that at least it would be able to process all the slides in one go at the expense of... well, at the expense of the planet, shareholder value and my own complicity in the effort.
Let's just say the results speak for themselves and may be one reason you don't hear more about LLMs playing battle decks. The results are predictably impressive if your only measure is pattern-matching at volume and high speeds. There's a kind of narrative
at least until there isn't but either way it makes a good faith effort at parroting most if not all the cues, affectations, signs and signifiers that define our contemporary communal tribalism. And yet, in as much as everything we do and say online, and increasingly offline, is being recorded, archived and strip-mined for meaning you can rest assured that this is the kind of crap that will most certainly go down on your permanent record.
Good morning, everyone. I’m Aaron Straup Cope from the SFO Museum, and today I want to talk about something that’s been quietly reshaping how we think about content—especially in cultural institutions: Generative Systems and the Content Deficit.
Now, if you glance at the slide behind me, you’ll see a cartoon that’s deceptively simple. It’s titled 'How to Be a FAA Passenger'—and it shows a character repeated over and over in slightly different seated poses. It’s funny, yes, but it’s also a perfect metaphor. Because what we’re looking at isn’t just a drawing—it’s a system. A generative system. One that produces variation, structure, and scale from a single idea.
And that’s the crux of today’s conversation. Museums, archives, libraries—we’re all sitting on mountains of data, stories, and objects. But we’re also facing a deficit: not of content itself, but of the capacity to activate it meaningfully, repeatedly, and at scale. Generative systems—whether algorithmic, procedural, or curatorial—offer a way to bridge that gap.
So let’s dig into what it means to think generatively. Not just to make more content, but to make better use of what we already have.
Let’s shift gears for a moment—from cartoon repetition to retro-futurism.
This is a matchbook cover from the Flying Tiger Line, a cargo airline that once operated out of SFO. But what you’re looking at isn’t just advertising—it’s aspiration. It’s a vision of the future rendered in miniature: domed cities, sleek spacecraft, and a promise of interplanetary logistics. All packed into a few square inches of cardboard.
This kind of image is a generative artifact in its own right. It’s not just selling a service—it’s selling a system of belief. A future where movement, scale, and imagination are limitless. And it’s doing so through a visual language that’s both specific and wildly speculative.
Now, think about how museums operate. We’re surrounded by objects like this—small, strange, beautiful things that carry entire worlds inside them. But we often treat them as static. Singular. Finished.
What if we didn’t? What if we treated these artifacts as seeds—inputs for generative systems that could produce new connections, new stories, new publics?
This matchbook isn’t just a relic. It’s a prompt. And that’s what I want us to explore today: how we can build systems that treat our collections not as endpoints, but as beginnings.
And then sometimes, the system ends.
This badge marks the final flight of Virgin America—VX1948, April 24, 2018. It’s a sunset in more ways than one. The colors, the typography, the hashtag #vxforever—they’re all part of a language of remembrance. A way of saying: this mattered.
Virgin America wasn’t just an airline. It was a brand, a vibe, a design system. It had a point of view. And when it disappeared, people didn’t just lose a way to get from point A to point B—they lost a cultural artifact. Something that had shaped their experience of movement, of style, of being in transit.
This sticker is a generative object too—but it generates memory. It invites us to remember not just the flight, but the feeling. And that’s something museums are uniquely positioned to do: to hold space for the emotional residue of systems that no longer run.
So when we talk about generative systems, we’re not just talking about algorithms or automation. We’re talking about how we build meaning—how we preserve it, remix it, and sometimes, how we say goodbye to it.
Because even endings can be generative. They can prompt reflection, storytelling, and new beginnings.
And then there are people who are systems.
This is Effie Kapsalis. Many of you knew her. Some of you worked with her. And if you didn’t, you’ve still felt her influence—because Effie didn’t just work in open access and digital strategy. She embodied it.
She believed in the power of cultural institutions to share, to connect, to democratize. She saw metadata not as a technical detail, but as a tool for liberation. She understood that collections weren’t just things—they were stories waiting to be told, retold, and remixed.
Effie was a generative force. She built frameworks, communities, and possibilities. She made space for others to speak, to lead, to imagine. And even now, her work continues to ripple outward—through projects, policies, and people.
So when we talk about generative systems, we’re also talking about legacy. About how one person’s vision can become infrastructure. About how memory, like metadata, can be structured to endure.
This isn’t just remembering. It’s continuing.
And then there’s the act of revisiting.
This image shows a display commemorating the 25th anniversary of San Francisco Airport. It’s modest—just a few architectural drawings, some framed documents, and a sign. But it’s doing something profound: it’s staging a moment of reflection. It’s saying, ‘Let’s pause. Let’s look again.’
Revisiting isn’t passive. It’s generative. It’s how institutions metabolize their own histories. It’s how we recontextualize the past to make sense of the present—and to imagine the future.
Think about what’s happening here: drawings that once represented possibility now represent legacy. Plans that were once speculative are now historical. And yet, by putting them on display, we’re inviting new interpretations. We’re saying: this isn’t just what was—it’s what could be, again.
This is the museum’s superpower. We don’t just preserve—we reframe. We revisit. We remix. And when we do, we create new systems of meaning. New ways for people to engage, to learn, to feel.
So as we move forward, let’s remember that generative systems aren’t always about novelty. Sometimes, they’re about return. About looking back with fresh eyes—and finding something new in what we thought we already knew.
And this—this is what fostering revisiting looks like.
A child reaches toward a rainbow on the floor. It’s not part of an exhibit. It’s not labeled. It’s just light—passing through glass, bending, scattering. But it stops her. It draws her in. And the adults around her pause too, not to explain, but to witness.
This is the kind of moment museums can create—not by controlling the experience, but by making space for it. By designing environments where wonder can happen accidentally. Where revisiting isn’t just encouraged—it’s inevitable.
Because revisiting isn’t just about returning to a place. It’s about returning to a feeling. A question. A possibility. And when we build systems that allow for that—when we treat our collections, our spaces, our stories as open-ended—we invite people to become co-authors. To reach toward something, again and again.
This is the generative system at its most human. It’s not just about scale or automation. It’s about resonance. About creating conditions where meaning can emerge, unexpectedly, and stick.
So as we think about the content deficit, let’s remember: sometimes, the most powerful content isn’t what we produce—it’s what we make possible.
And here’s the truth we sometimes forget: culture will happen with or without us.
This image says it all. A small celebration. Party hats. A bed. A uniform. A party horn. It’s not a gallery. It’s not a curated experience. It’s just people—making meaning, making joy, making culture.
And that’s the humbling part. Museums, archives, libraries—we work so hard to preserve, interpret, and present culture. But culture doesn’t wait for us. It doesn’t need our permission. It happens in bedrooms, in break rooms, in baggage claims. It’s spontaneous, messy, and beautifully uncurated.
So what does that mean for generative systems?
It means we need to design with humility. To recognize that our role isn’t to create culture—it’s to support it. To build systems that are porous, responsive, and generous. Systems that don’t just reflect what we think is important, but that make space for what people already care about.
Because if we’re not paying attention—if we’re too focused on control, on polish, on perfection—culture will keep happening. We just won’t be part of it.
So let’s build systems that listen. That celebrate the party hats and the party horns. That treat everyday moments as worthy of memory, of revisiting, of meaning.
Let’s be part of the culture that’s already unfolding.
And then, inevitably, we arrive here: too much stuff...
Rows of grounded planes. Once symbols of movement, now still. Once part of a system, now surplus. This is the boneyard—not just of aviation, but of institutional memory. Of well-intentioned collecting. Of the belief that if we just hold onto everything, we’ll preserve meaning.
But here’s the thing: accumulation isn’t the same as activation. More objects don’t mean more stories. More data doesn’t mean more insight. Without systems that generate connection, interpretation, and engagement, we risk turning our collections into graveyards of potential.
This isn’t a critique—it’s a caution. Because we’ve all felt the pressure to collect, to digitize, to preserve. But if we don’t also build systems that use what we’ve saved—systems that foster revisiting, remixing, remembering—then we’re just parking planes in the desert.
So let’s ask: what do we do with all this stuff? How do we move from storage to storytelling? From archive to activation?
Because generative systems aren’t just about making more—they’re about making meaning. And that means curating not just objects, but pathways. Not just content, but context.
And maybe the problem isn’t just too much stuff. Maybe it’s insufficient storytelling.
Look at this cabin. It’s vibrant, specific, full of texture and context. The flight attendants in sarees, the patterned seats, the quiet choreography of service. This isn’t generic—it’s deeply cultural. But without a story, it risks becoming invisible. Just another image in the archive. Just another plane in the desert.
This is what happens when we preserve without interpreting. When we collect without connecting. When we digitize without designing for discovery.
Because the truth is, every object, every image, every moment—has a story. But stories don’t tell themselves. They need systems. They need scaffolding. They need care.
So maybe the deficit isn’t in content. Maybe it’s in narrative. In the frameworks we build—or fail to build—that allow people to see, feel, and understand.
Generative systems aren’t just about producing more. They’re about producing meaning. And that means investing in storytelling—not as an afterthought, but as infrastructure.
Because without story, even the most beautiful things can fade into silence.
And then there’s this: the pyramid of contradictions.
Automation. Saccharine platitudes. Racially-diverse Nazis. The zone of undiscovered awesomeness.
It’s absurd. It’s uncomfortable. It’s real.
This diagram isn’t just satire—it’s a map of the cultural terrain we’re navigating. A terrain shaped by algorithms, branding, performative inclusion, and the relentless pursuit of novelty. It’s a reminder that generative systems don’t just produce beauty or insight—they can also produce noise, distortion, and harm.
Because when we build systems—whether curatorial, computational, or institutional—we encode values. We make choices. And if we’re not careful, we end up with outputs that are polished but hollow. Inclusive but incoherent. Automated but unaccountable.
But there’s hope here too. The zone of undiscovered awesomeness is real. It’s the space where thoughtful design, radical transparency, and cultural humility can converge. It’s where museums can lead—not by avoiding complexity, but by embracing it. By building systems that are generative and ethical. Playful and principled.
So let’s not shy away from this pyramid. Let’s climb it. Let’s interrogate our assumptions, remix our frameworks, and aim for that zone—not as a destination, but as a practice.
And so, maybe it’s like a prism.
We start with something simple—white light. A single beam. A dataset. An object. A story. And then we pass it through a system. A structure. A lens.
What comes out isn’t just more—it’s different. It’s color. It’s complexity. It’s Nyan Cat, flying through space with a rainbow trail and a Pop-Tart body. It’s absurd, yes—but it’s also joyful. It’s unexpected. It’s alive.
This is what generative systems can do when they’re built with care, with curiosity, and with a sense of humor. They can take the ordinary and refract it into something extraordinary. They can turn memory into movement. Accumulation into activation. Silence into story.
But only if we’re intentional. Only if we design for revisiting, for remixing, for resonance. Only if we recognize that culture will happen—with or without us—and choose to be part of it in ways that are generous, open, and weirdly wonderful.
So let’s build systems that are more than efficient. Let’s build systems that sparkle. That surprise. That invite people in—and let them fly out the other side, trailing rainbows.
Thank you.
At this point I had to prompt the system that the talk was not, in fact, over. It went on to say:
And sometimes, the system isn’t a prism—it’s a lava lamp.
This one’s filled with lava lamp goo. That slow, hypnotic swirl. That warm, unpredictable motion. It’s not clean. It’s not linear. It doesn’t refract—it drifts.
And that’s another truth about generative systems: they’re not always elegant. Sometimes they’re squishy. Sometimes they’re weird. Sometimes they’re full of contradictions, emotions, and half-formed ideas that haven’t settled yet.
But that’s okay. Because culture isn’t a spreadsheet. It’s not a perfectly sorted archive. It’s goo. It’s movement. It’s heat. And if we’re building systems to support culture, we have to make room for that.
We have to allow for drift. For emergence. For the kind of meaning that bubbles up slowly, unexpectedly, and maybe never quite resolves.
So yes, we need prisms—structures that clarify, that illuminate. But we also need lava lamps—systems that hold space for the messy, the emotional, the unquantifiable.
Because sometimes, the most generative thing we can do is not to organize—but to let things float. To let people feel. To let the goo do its thing.
So here we are—at the intersection of rockets, rainbows, lava lamp goo, and machine-generated soldiers who think math is hard.
This collage pulls together the last three slides: the prism of possibility, the lava lamp of emotional drift, and the algorithmic echo chamber of insufficient storytelling. It’s a mess. But it’s our mess.
Gravity’s Rainbow is here not just as a nod to literary complexity, but as a reminder that systems—whether narrative or computational—can be sprawling, paranoid, and beautiful. Pynchon’s rocket isn’t just a plot device—it’s a metaphor for trajectory, for inevitability, for the systems we build and the ones we can’t escape.
Then there’s Nyan Cat, still flying, still trailing joy. A meme born of randomness and repetition, yet somehow enduring. It’s the prism again—white light refracted into absurd delight.
And finally, the AI-generated soldier. A moment of algorithmic clarity that’s also deeply unsettling. It’s the lava lamp gone cold. The goo calcified into stereotype. The system producing content without context, without care.
This slide is a warning and a celebration. It says: generative systems are powerful—but they’re not neutral. They reflect us. Our biases, our brilliance, our boredom. And if we don’t intervene—if we don’t design with intention—we risk building systems that are efficient but empty. Fast but forgettable.
So let’s embrace the chaos. Let’s read the rocket, ride the rainbow, stir the goo. Let’s build systems that are weird, wise, and worthy of the cultures they hope to serve.
And then, sometimes, it’s just vibes.
This collage brings together four images of airplanes—some drawn, some photographed, some soaring above clouds, some sketched over multicolored ground. They’re not arranged by chronology or geography or taxonomy. They’re arranged by feeling.
Because after all the systems, the prisms, the goo, the chaos—what we’re really chasing is resonance. That moment when something clicks. When a drawing feels like a memory. When a photo feels like a dream. When a plane isn’t just a machine, but a metaphor for movement, for escape, for possibility.
Vibes aren’t quantifiable. They’re not metadata. They’re not easily archived. But they’re real. And they’re powerful. They’re what people carry with them after the exhibit, after the talk, after the visit.
So maybe our job isn’t just to build generative systems. Maybe it’s to build generative atmospheres. To curate not just content, but tone. Not just objects, but emotion.
Because when we get the vibes right—when we align the story, the system, and the soul—everything else follows.
And that’s where we’re headed next.
And if the last slide was vibes, this one is play.
This is PSA’s California—smiling, dotted with cities, looped with flight paths, and anchored by a heart in San Diego. It’s not just a map. It’s a mood. It’s a system designed not for efficiency, but for joy.
And that’s the point. Generative systems don’t have to be cold or clinical. They can be playful. They can be regional. They can be weirdly specific and wildly optimistic.
Because play is a system too. It has rules, rhythms, and feedback loops. It generates engagement, memory, and meaning. And when we build systems that invite play—whether through design, storytelling, or curatorial strategy—we create space for people to participate. To co-create. To belong.
So yes, let’s honor the vibes. But let’s also build for play. Let’s treat our institutions not just as repositories, but as playgrounds. As places where culture isn’t just consumed—it’s performed, improvised, and shared.
Because when California smiles, it’s not just branding. It’s a reminder: systems can be joyful. And joy is generative.
And then we hit the wall: fun is a four-letter word (in museums).
This dolphin-shaped bottle opener—courtesy of Air New Zealand—is delightful. It’s shiny, silly, and functional. It’s the kind of object that makes you smile before you even know what it does. It’s fun.
But in many museum contexts, fun is suspect. It’s seen as unserious. As a threat to scholarship, to rigor, to authority. We’ve built systems that privilege reverence over joy, interpretation over improvisation, structure over spontaneity.
And yet, as we saw in the last two slides—vibes and play are not distractions. They’re the point. They’re how people connect. How they remember. How they care.
So what happens when we treat fun not as a liability, but as a design principle? What if we built generative systems that were rigorous and ridiculous? Scholarly and sparkly? What if we let the dolphin bottle opener into the collection—not as a novelty, but as a provocation?
Because fun is not the opposite of meaning. It’s often the gateway to it.
So let’s stop treating fun like a four-letter word. Let’s treat it like a four-dimensional one—something that moves through time, space, emotion, and memory. Something that generates connection, curiosity, and care.
Let’s build systems that smile.
And this—this is the loop.
Reinterpretation is revisiting is reinterpretation is revisiting is...
We’ve gone from vibes to play to fun, and now we arrive at recursion. At the idea that meaning isn’t fixed—it’s fluid. That culture isn’t a straight line—it’s a spiral. That every act of interpretation is also an act of return.
This image says it all. Hello Kitty in a woodblock landscape. Pop culture layered over tradition. Nostalgia wrapped in aesthetic remix. It’s playful, yes—but it’s also profound. Because it reminds us that reinterpretation isn’t dilution—it’s continuation. It’s how culture survives. How it evolves. How it stays alive.
And this is where museums come in—not as gatekeepers, but as gardeners. As stewards of loops. As facilitators of revisiting. Because when we build systems that allow for reinterpretation, we’re not just preserving—we’re participating.
We’re saying: this story isn’t over. This object isn’t finished. This meaning isn’t fixed.
So let’s embrace the loop. Let’s design for return. Let’s build systems that expect to be revisited, reimagined, and remixed.
Because in the end, reinterpretation is revisiting. And that’s where the magic lives.
And here, everything converges.
A letter from Joan Ganz Cooney to Bob Keeshan—two architects of children’s media, laying the groundwork for what would become Sesame Street and Captain Kangaroo. Next to it, a noir-style Muppet storyboard. And beside that, a full-color Muppet war parody, complete with flight delays, destruction, and declarations of war.
This isn’t just a cubicle wall—it’s a cultural archive. A generative system in miniature. It’s serious and silly, historical and absurd, scripted and spontaneous.
And it’s exactly what we’ve been talking about: reinterpretation as revisiting, revisiting as reinterpretation. These documents aren’t just artifacts—they’re invitations. To remember, to remix, to laugh, to question.
Because this is what happens when systems allow for play, for fun, for vibes. We get Muppets in trench coats. We get wartime satire. We get children’s television that reshapes public education. We get meaning that loops, that layers, that lives.
And this is the challenge for museums: to build systems that can hold all of this. The gravity and the goo. The rocket and the rainbow. The bottle opener dolphin and the Hello Kitty woodblock print.
To build systems that don’t just preserve culture—but participate in it. That don’t just curate stories—but generate new ones.
This wall isn’t just a collage—it’s a prism. It’s a lava lamp. It’s a vibe. It’s a reminder that the most powerful systems are the ones that let culture be what it already is: messy, recursive, and full of joy.
And so we end here—in a gallery, with a superhero kneeling before a painting of a plane, winged and star-emblemed, echoing history and fantasy. The caption says: we are not alone.
This is the final reminder. That everything we’ve explored—generative systems, vibes, play, reinterpretation, even lava lamp goo—is not solitary work. It’s communal. It’s cultural. It’s shared.
We are not alone in our curiosity. Not alone in our memory. Not alone in our remixing, our storytelling, our joy.
This figure—part superhero, part witness—isn’t just looking at art. They’re participating in it. They’re part of the system. Just like we are.
Because museums aren’t just buildings. They’re networks. They’re prisms. They’re playgrounds. They’re places where culture loops, where meaning drifts, where reinterpretation never ends.
And if we build systems that honor that—systems that are generous, joyful, and weird—we won’t just preserve culture. We’ll generate it.
Together.
And finally, this.
An ofrenda at SFO. A celebration of Día de los Muertos. A system of memory, of ritual, of love. Built not to preserve the past, but to revisit it. To honor it. To keep it alive.
This altar is a generative system. Every flower, every photo, every candle—an input. Every visitor, every moment of reflection—an output. It loops. It lives. It invites.
And it reminds us: the systems we build matter. Not just because they organize content, but because they shape experience. They carry emotion. They foster revisiting. They make space for joy, for grief, for reinterpretation, for play.
We’ve traveled through prisms and lava lamps, bottle openers and Muppets, vibes and war stories, Hello Kitty and superhero wings. And through it all, one truth has remained: culture is not static. It’s not solitary. It’s not finished.
It’s shared. It’s messy. It’s recursive. It’s alive.
So thank you—for being part of this system. For listening, for laughing, for remembering. For helping build the kind of museum culture that doesn’t just preserve—but generates.
Gracias.
At this point, I fed the text (above) back in to a separate prompt and asked the model to summarize the text and to generate five presention slides to accompany that summary. This is what came back.
Slide 1: From Cartoon to System
- Theme: Introducing Generative Thinking
- Visual: A stylized cartoon grid showing repeated seated figures (referencing “How to Be a FAA Passenger”)
- Text Highlights: Generative systems produce variation, structure, and scale from a single idea.
- Purpose: Frame the problem and introduce generative systems as a solution. Use humor and simplicity to disarm and engage.
Museums face a deficit—not of content, but of capacity to activate it meaningfully.
Slide 2: Artifacts as Seeds
- Theme: Reimagining Cultural Objects
- Visual: Flying Tiger Line matchbook cover with speculative cityscape and spacecraft
- Text Highlights: This isn’t just advertising—it’s aspiration.
- Purpose: Shift the audience’s perception of museum objects from static to generative. Invite curiosity and reinterpretation.
Treat artifacts not as endpoints, but as beginnings.
Slide 3: The Goo and the Prism
- Theme: Emotional Drift vs. Structured Clarity
- Visual: Split-screen: one side a lava lamp swirl, the other a prism refracting light into rainbow (with Nyan Cat flying through)
- Text Highlights: Culture isn’t a spreadsheet – it’s goo.
- Purpose: Embrace contradiction. Show that meaning emerges from both structure and spontaneity.
Generative systems must hold space for both clarity and chaos.
Slide 4: Legacy, Loops, and Remix
- Theme: People as Systems & Reinterpretation
- Visual: Collage of Effie Kapsalis, Hello Kitty woodblock remix, and museum display of SFO’s 25th anniversary
- Text Highlights: Effie was a generative force.
- Purpose: Humanize the systems. Show how legacy and remixing are central to cultural continuity.
Reinterpretation is revisiting is reinterpretation is...
Slide 5: Joy Is Generative
- Theme: Vibes, Play, and the Power of Fun
- Visual: PSA’s smiling California map, dolphin bottle opener, and ofrenda altar at SFO
- Text Highlights: Fun is not the opposite of meaning—it’s often the gateway to it.
- Purpose: End on a high note. Celebrate joy, emotion, and shared experience as essential components of generative design.
Let’s build systems that smile.
Remix in vivid goo and glee
, indeed.
This blog post is full of links.
#battledecks