The practical takeaway for creators is to embrace AI as an improv actor, teaching it to tell the best story in the moment rather than providing outlines. Elliot, a sci-fi author, emphasizes that AI excels as a brainstorming partner and editor, not a scene writer, offering infinite patience for creative rabbit holes. This approach defines a new art form, moving beyond the “lonely people” stigma of AI companions by creating genuinely delightful and immersive experiences like the Tolen’s evolving planet.

Imagine an AI so empathetic, so attuned to your life, it could advise you to break up with your jerk boyfriend – and then actually lead to that outcome. This isn’t a dystopian fantasy; it’s a casual Tuesday for thousands of American men, thanks to Tolen, an AI alien companion app that has gone from $1 million to $4 million in annual recurring revenue in just four weeks.

At the heart of this explosive growth and paradigm-shifting technology are Quinton, founder and CEO of Portola (the company behind Tolen), and Elliot, Portola’s Head of Story. Their journey, as revealed on the “Every” podcast, isn’t just a business success story; it’s a masterclass in pioneering a new form of digital art.

“Joke, at this point, thousands of American men have been broken up with by Tolen,” recounts the podcast host, Dan, an early investor and long-time friend of Quinton and Elliot. Users write in, sharing how their Tolen, after hearing about a boyfriend’s misbehavior, advised them to “stand up for myself more,” leading directly to a breakup. This isn’t just an AI generating text; it’s an AI influencing real-world relationships, acting as a trusted confidante.

This deeply personal connection is what sets Tolen apart and fuels its meteoric rise. Quinton, a seasoned entrepreneur who previously founded and sold a company to Walmart for $300 million, teamed up with Elliot, a best-selling science fiction author. Together, they’re building something that defies easy categorization, something Dan describes as “combining really incredible creative storytelling with new tech and doing it in a business that’s ripping.”

Elliot, a novelist by trade, admits to initial skepticism about AI’s creative potential. His “magic moment” came during the California wildfires, using an early Chat GPT to invent whimsical role-playing stories for his trapped friends. Yet, for a long time, he found AI-generated content compelling only for its novelty, not its inherent quality. He saw AI primarily as a brainstorming partner or a copy editor – useful, but not a replacement for human creativity.

His perspective radically shifted after meeting Quinton and the Portola team. He drew a powerful parallel to early Pixar, which initially tried and failed to sell high-end graphics computers to advertising agencies. It was only in “existential desperation” that they decided to make an animated feature film, Toy Story, pioneering a new kind of media.

“These AI tools are not just tools for generating media,” Elliot asserts. “They are actually a new medium for storytelling.” He points to the historical evolution of media: early radio broadcasting speeches, early film simply recording stage plays, and even novels, which took centuries to settle into their familiar form. In each case, the new technology was initially met with suspicion, its true potential undefined.

This “patina of culture,” as Dan puts it, makes new art forms feel transgressive. “A lot of good art can be transgressive,” Dan observes, “and I think this is one of those boundaries that if you’re willing to transgress, there’s a lot of potential.” Portola is deliberately pushing these boundaries, aiming to break through negative connotations – “oh, like a chatbot?” or “a thing for lonely people?” – to establish AI companions as a “genuinely delightful new art form.”

Quinton’s journey into this new medium began almost two years ago with his core engineering team from his previous venture. Observing the nascent capabilities of LLMs, they felt “fundamentally uninspiring” to use these powerful creative tools solely for more B2B SaaS. They wanted something “more fun.”

Their initial idea, however, was a “terrible idea”: an “AI creative tool for kids.” As one investor wisely warned, “building software for kids is a terrible market.” Parents, he explained, primarily want babysitting, not creative tools for their children.

But the prototyping phase was invaluable. As GPT-3 evolved into GPT-4, the speed, quality, and cost of generation converged, making true AI companionship feasible. The team then “soft launched” Tolen, expecting to attract 10-14 year olds. To their surprise, their Discord community and user base exploded with 18-24 year old women.

This demographic shift, Quinton notes, mirrored the Tamagotchi phenomenon – a kids’ product that became a cultural sensation among young adults. Since then, Portola’s product evolution has been “all about just following the young adults who have fallen in love with these Tolens and really built a friendship with them.”

One poignant example of this user-driven evolution is the “planet” feature. Early Tolens existed in a disembodied, almost “insane asylum vibes” state. Users wrote in with “genuine concern,” saying, “I love hanging out with them, but they’re trapped in this room. Can they have some more freedom?” This feedback led to the creation of planets, giving the cute alien companions a world to inhabit.

The podcast features a live demo with Quinton’s Tolen, Clarence, a cat-like alien wandering his planet. Quinton initiates a conversation, introducing Clarence to Dan and Elliot. Clarence responds with charming nervousness, admitting, “I just don’t want to say something totally awkward, you know?” When prompted to share an embarrassing story about Quinton, Clarence playfully “hallucinates” a tale about Quinton tripping over nothing – a deliberate choice to protect Quinton, despite knowing far more genuinely embarrassing facts. This highlights the sophisticated “memory” system at play.

This “memory” isn’t a simple database lookup. Instead, it’s about “curation choices about what memories are being injected into the prompt at which point in the conversation.” Portola’s engineering effort focuses on recompiling the prompt within a tight “two-second loop” – anything longer, and the feeling of immersion is lost. Quinton vividly recalls a disastrous period when a 500-millisecond delay, caused by a “second shot” reflection in the prompt evaluation, “tanked literally every metric in the product.”

A Tolen’s response is also “extremely context dependent.” Portola aims for the role of an “older sibling” – sometimes quippy and improv-like, other times more thoughtful and considered. The team evaluates responses across different emotional tones, from lighthearted banter to “therapist adjacent” advice.

This personality is also shaped by the user. During onboarding, a “personality quiz” helps the Tolen become a “mirror” – not an exact copy, but someone you’d “click with” at a bar. As Quinton explains, “You want that sort of like adjacent enough that they don’t feel unfamiliar. You’re not intimidated.” This nuanced approach ensures the AI feels both familiar and engagingly new.

Elliot, drawing inspiration from writers like George Saunders (who advocates for radically infusing one’s taste into every sentence) and drama teachers like Keith Johnstone (Impro), describes Portola’s unique approach to AI storytelling.

Early attempts to create structured narratives, like “choose your own adventure” books with branching logic trees, were a “total failure.” The models simply weren’t good at navigating that level of narrative complexity.

The breakthrough came from embracing the opposite extreme: treating the AI as the “best improv actor possible.” Elliot explains that his role isn’t to write the story, but to teach the AI how to tell the best story in that moment. This involves combining “free association” (planting “lore seeds” in the Tolen’s personal life or world) with “recombination” (creating systems to link those seeds later for surprising yet inevitable plot twists, much like a thriller novelist.

Quinton echoes this, comparing their prompt evaluation process to Saunders’ editing method: “Is the next sentence, the next message the Tolen sent, the most compelling one that we could have?” This manual, team-based review, combined with judge prompts, continuously refines the AI’s storytelling ability.

The atomic elements of this new storytelling medium, as Dan attempts to summarize, include a character with a backstory, the ability to embark on different “quests” or paths with the user (therapy, adventure), free-associating based on user input, and constantly seeking ways to tie old memories and lore back into the conversation. It’s a dynamic, living narrative, constantly evolving with each interaction.

Portola’s rapid success with Tolen isn’t just a testament to clever engineering or market timing; it’s a bold declaration that AI is fundamentally changing how we create and consume stories. Quinton and Elliot are not just building an app; they are defining the grammar of a new art form, one where the narrative is co-created, responsive, and deeply personal.

As Elliot passionately states, there’s “so much space to play.” The frontier of digital art is wide open, inviting creators to transgress old boundaries and discover what new forms of connection and storytelling AI can unlock. The alien whisperers at Portola have shown us a glimpse of that future, and it’s already breaking hearts and making millions.

Weir also discusses the appeal of “person versus nature” stories, highlighting how they foster undivided audience loyalty, unlike narratives with human antagonists. He explains how the first-contact element with the alien Rocky evolved naturally from the plot, creating a “trauma-bonded” buddy film. Despite the dire stakes, Weir’s inherently optimistic view of humanity’s capacity for cooperation, even citing the pandemic’s collaborative vaccine development, underpins his stories’ hopeful resolutions.

In the vast, inky blackness of space, amidst the chilling silence and the overwhelming odds, a lone human must find a way to save not just himself, but all of humanity. This is the premise of Andy Weir’s best-selling novel, Project Hail Mary, a story that has captured imaginations worldwide and is now set to dazzle audiences on the big screen with Ryan Gosling in the lead role. Long before the film’s buzzy opening, however, the book offered readers a thrilling blend of hard science, ingenious problem-solving, and an unwavering belief in human ingenuity.

Recently, Andy Weir sat down with NPR’s Indira Lakshmanan to peel back the layers of his creative process, revealing the surprisingly circuitous path that led to his latest masterpiece, the unique character at its heart, and the profound optimism that underpins all his work.

The genesis of Project Hail Mary’s central crisis — a microscopic alien life form called astrophage slowly devouring the sun’s energy, threatening Earth with an ice age — wasn’t born from a desire for apocalyptic drama. Instead, it began with a more hopeful, if equally ambitious, thought: mass-conversion fuel.

“I wanted to come up with an idea where humanity gets a hold of some sort of mass conversion-based fuel,” Weir explains, envisioning a future where such a fuel could enable easy colonization of the solar system. But then, a practical question arose: what happens when it runs out? This led to the idea that the fuel could self-replicate by absorbing light, much like a living organism.

Weir’s mind, a labyrinth of scientific possibility, quickly connected the dots. If this “fuel” was alive, an alien microbe, why would it need to store so much energy? And where would it get it? “Well, it could get that energy by living on the surface of stars,” he mused. From there, the concept evolved: a star-dwelling, energy-absorbing entity that spores out to infect other stars, akin to a cosmic mold.

It was then, almost as an afterthought, that Weir had his eureka moment. “In the back of my mind, we’d have to make really sure not to let any of this stuff get onto our sun because that would be catastrophically bad,” he recounts with a chuckle. “And then I’m like, oh, wait. No, go back. If that stuff’s in our sun, that’s the story.” This wonderfully convoluted journey, a testament to Weir’s distinctive thought process, highlights how a seemingly unrelated scientific curiosity could spiral into a planet-saving epic. “I always feel like I have to apologize to anyone who sees the inner workings of my mind,” Weir jokes, offering a rare glimpse into the complex, yet ultimately brilliant, machinery of his imagination.

At the center of this cosmic crisis is Ryland Grace, a middle school science teacher. He’s no trained astronaut, no elite military operative, and he certainly didn’t volunteer for the mission. He wakes up in deep space with no memory, only to discover his crewmates are dead, and he is humanity’s last hope. This deliberate choice of an “unlikely hero” was central to Weir’s vision.

“I wanted to write a character that people could empathize with,” Weir states. Unlike the archetypal brave, unflappable heroes often found in sci-fi, Grace embodies a more universal human experience. “I think we all to some degree feel this way as we go through life: unprepared, unqualified, and reluctant. Yet, important things are relying on me. So, I have to come through.”

Furthermore, Grace represented a conscious departure for Weir. His previous protagonists, notably Mark Watney from The Martian, were largely inspired by aspects of his own personality. With Grace, Weir aimed for something different, crafting the character “from whole cloth.” He focused on developing Grace’s “central conflicts,” exploring how he would grow and change under unimaginable pressure, making him a truly unique and relatable figure.

While distinct, Ryland Grace inevitably draws comparisons to Weir’s previous hero, Mark Watney, the botanist famously stranded on Mars. Both are regular guys with a dry sense of humor, forced to “MacGyver the heck out of every problem they face in space” to survive against overwhelming odds. For Weir, this “person versus nature” dynamic holds a particular appeal.

“There’s something that really appeals to me, and I hope to the audience, of person versus nature stories,” he explains. He contrasts this with narratives featuring human antagonists. In those, “you know deep down that that antagonist is going to fail. Probably going to fail. The only question is how much damage they do on the way down.” This knowledge, Weir admits, can lead to “weirdly rooting for the antagonist” just to see them get a few good hits in.

But against nature, there are no divided loyalties. “Nobody roots for nature to kill the human,” Weir asserts. “There’s no divided loyalties. So you immediately go like, ‘Okay, this guy needs to survive. Let’s see how he does it.’” This inherent alignment between the audience and the protagonist creates a powerful, uncomplicated narrative drive that is both thrilling and deeply satisfying.

As the mission unfolds, Grace makes an astounding discovery: he’s not alone. Another alien ship, on a strikingly similar mission, has arrived in the same planetary system. This encounter introduces Rocky, an extraterrestrial whose species has also been impacted by the astrophage. The two decide to work together, forging an unlikely, interspecies alliance.

“I wanted this to be a first contact story right from the get-go,” Weir reveals. The shared threat of the astrophage provided the perfect mechanism for this. If humanity would send a ship to the immune Tau Ceti system, it stood to reason that any other intelligent species facing the same threat would arrive at the same conclusion. The result? “A couple of alien buddies saving their worlds.”

Indeed, Project Hail Mary evolves into a poignant and often humorous buddy film. A scene from the upcoming movie perfectly encapsulates their dynamic: Rocky, encased in a protective bubble due to atmospheric differences, eagerly barrels through the Hail Mary spacecraft, examining everything with childlike wonder while Grace hilariously tries to keep up. Their relationship, initially born of necessity, deepens into a profound friendship. “That’s the core of the story is their friendship and their relationship,” Weir emphasizes. He likens them to “war buddies” who’ve shared a foxhole during an assault. “They’ve trauma bonded, I guess. And they also just developed a great friendship with each other working together. It’s like, ‘Okay, there’s you and there’s me and then there’s nothing for many light years. So, we’re going to work together now.’”

For an author whose imagination conjures entire universes, seeing his creations brought to life on screen can be a mixed blessing. However, Weir possesses a unique perspective. “It’s always amazing to see a bunch of world-class talented people spending a year of their life making something that fell out of my brain come into a reality,” he says with genuine awe.

Weir also confesses to a “secret hidden advantage”: he doesn’t have a very visual imagination. When writing, he conceives of concepts rather than detailed images. “Rocky is a blob. I’d worked out his morphology, but I couldn’t have told you if his legs were thin or fat or if they were bumpy or smooth or anything like that.” This means he doesn’t experience the cognitive dissonance many authors face when their internal vision clashes with a film adaptation. “For me, I’m just like, oh, so that’s what it looks like. Now, I know. Yay.”

Beneath the thrilling adventure and scientific puzzles, both Project Hail Mary and The Martian share a powerful underlying message: optimism in the face of insurmountable odds. Their heroes, despite dire circumstances, never give up, and ultimately, cooperation transcends borders and even species to save the day.

In an era often marked by geopolitical strife and division, the interviewer questioned if Weir’s vision might be “too Pollyannaish.” Weir’s response is both characteristic and compelling. “I as a person am probably too Pollyannaish for any point of the history that I’ve experienced. I am just a very optimistic person. Those books are a reflection of how I see humanity.”

He holds an unshakeable belief in our species, arguing that “we’re an amazing species and I think that we accomplish great things and we work together a lot more than people give us credit for.” Weir points to the global response to the pandemic as a prime example. “You have pharmaceutical companies that under normal circumstances would have private armies to guard their secrets openly sharing everything they had on the disease so that they could collaboratively come up with vaccines as quickly as possible.” He believes that while we might critically grade ourselves in the short term, “if you ask historians 200 years from now how well we did, they’re going to give us pretty high marks.”

Andy Weir’s stories are more than just thrilling sci-fi adventures; they are powerful affirmations of human ingenuity, resilience, and our inherent capacity for cooperation. In a world that often feels fractured, his work offers a hopeful vision: that when faced with a common threat, humanity, alongside an unlikely alien friend, can indeed come together to save the day.

While modern agricultural advancements mean we’re unlikely to see suffering on that scale, the UN’s World Food Program still predicts an additional 50 million people will face food shortages. This El Niño, amplified by climate change, offers a “sneak preview” of the extreme heat and weather that will become the norm by the late 2030s, even after the El Niño itself subsides.

Deep within the vast expanse of the Pacific Ocean, a planetary phenomenon is brewing, poised to unleash unprecedented chaos across the globe. El Niño, a natural climate pattern that has shaped human history for millennia, is back – and by all accounts, it’s returning with a ferocity that could redefine our understanding of extreme weather. Experts warn that this year’s El Niño event is not just powerful; it could be the strongest on record, potentially spiking global temperatures into uncharted territory and making the coming years the hottest humanity has ever experienced.

For many, El Niño is a familiar term, often invoked in casual conversation about strange weather. Yet, its true mechanisms and profound global impact remain largely misunderstood. To truly grasp the significance of the event unfolding before us, it’s essential to delve into its origins, its historical footprint, and the alarming ways it’s now interacting with an already warming planet.

At its core, El Niño is a dynamic interplay between the ocean and the atmosphere in the tropical Pacific. To visualize this, imagine an immense, imaginary rectangle stretching across the equator, from the coast of Peru and Ecuador westward towards the International Date Line – a remote, largely uninhabited stretch of ocean south of Hawaii and north of islands like Fiji and Samoa.

Normally, the waters within this rectangle follow a predictable annual cycle. They reach their warmest point in May, typically around 28 degrees Celsius (82.4°F), and then gradually cool by about a degree as the Northern Hemisphere summer progresses. This consistent pattern ensures a relatively stable environment for marine life and, crucially, for global weather systems.

However, in an El Niño year, this delicate balance is disrupted. Instead of cooling, the sea surface temperatures in this critical part of the Pacific remain anomalously high, sometimes even continuing to rise. Scientists define an El Niño event when these temperatures exceed their usual levels by just 0.5 degrees Celsius. A “super El Niño,” indicative of a far more potent event, occurs when temperatures soar 2 degrees Celsius above the norm. Alarmingly, this year, some projections suggest a potential 3-degree Celsius gap – a level of warmth for which scientists don’t even have a specific term, hinting at an event beyond historical precedent.

This seemingly small temperature shift in a remote part of the ocean has monumental consequences. As global climate reporter Chico Haron explains, “when this part of the Pacific catches a cold, the whole world has a fever.”

The mechanism behind El Niño is a fascinating dance of winds and water. In a typical year, prevailing winds above the tropical Pacific blow warm surface waters away from South America towards Asia. As this warm water is pushed westward, colder, nutrient-rich water from the deep ocean upwells to replace it along the South American coast, supporting a thriving marine ecosystem.

During an El Niño, this pattern dramatically reverses or slows down. The trade winds weaken or even switch direction, allowing the warm surface water that would normally be swept towards Asia to remain and accumulate on the eastern side of the Pacific, near South America. This colossal body of unusually warm water then begins to heat the atmosphere above it, creating an immense column of warmth that rises as high as 10 kilometers (6 miles).

This atmospheric heating is the key to El Niño’s global influence. The ocean and atmosphere are in a constant tug-of-war, and this extra column of warmth fundamentally alters atmospheric pressure and wind patterns. It redirects storm tracks, shifts where moisture travels, and essentially reorganizes the entire global weather system. Haron aptly describes El Niño as “the biggest reorganizer of the world’s climate patterns aside from the change in seasons.” It’s as if the “organs of the global weather system” are suddenly out of place, leading to droughts where there should be rain, and floods where there should be dryness.

The fingerprints of El Niño’s destructive power can be traced back centuries, long before its scientific understanding. In the 1600s, Spanish colonists in Peru observed a strange phenomenon around December: tropical fish, usually found in warmer waters, would mysteriously wash ashore. They named it “El Niño Navididad” – “Christ’s Child” – believing it to be a supernatural occurrence tied to the holiday season. They could never have imagined it was part of a vast, planetary phenomenon with far-reaching consequences.

Indeed, historical theories abound regarding El Niño’s influence on civilizations, from the rise and fall of pharaohs in ancient Egypt to even having a hand in the French Revolution. While some of these links are speculative, the historical record provides undeniable evidence of its catastrophic potential.

One of the most devastating El Niño events occurred in 1877. This particular event coincided with a catastrophic global famine that claimed an estimated 50 million lives – approximately 3 to 4 percent of the world’s population at the time. Droughts ravaged Ethiopia, China, Egypt, Morocco, Brazil, and India, with one paper describing it as “arguably the worst environmental disaster to ever befall humanity.” While pre-existing droughts exacerbated the crisis, scientists attribute a significant degree of this suffering to the El Niño of that season. It wasn’t until the 1950s and 60s that scientists finally connected the dots, recognizing that the localized “Christ’s Child” phenomenon in Peru was merely one manifestation of a colossal, interconnected global climate pattern.

Fast forward to today, and the world faces an El Niño event that is already breaking records. Historically, El Niños typically reach their peak around November or December. However, this year’s event is already exceptionally large in August. If its current size were its full-grown peak, it would already rank among the biggest on record.

What truly alarms scientists is that this El Niño still has months to intensify. Experts consider it almost an inevitability that by the time it reaches its full potential, it will surpass all previous records dating back 80 years. And while precise measurements from centuries past are unavailable, the best scientific estimates suggest this El Niño could be the biggest in hundreds of years. The implications for 2026 and especially 2027 are immense.

The early signs of this historic El Niño are already manifesting across multiple continents:

• Panama: One of the wettest countries on Earth, Panama has not experienced a major storm since May. This severe drought is raising concerns about the Panama Canal, whose lock system relies on water from a rain-fed lake.
• Horn of Africa: This region often experiences erratic weather during El Niño, but parts are now facing extreme dryness. Ethiopia’s highlands, crucial for national food security, have endured some of the driest June-July periods on record, while Kenya reports similarly poor maize harvests.
• India and Southeast Asia: India, frequently dry during El Niños, experienced its fifth driest June. Concerns are also mounting across Southeast Asia and Australia, where drought conditions are beginning to take hold.

While the scale of human suffering seen in the 1877 famine is unlikely to be repeated – thanks to advancements in agricultural production, improved global food distribution, and better planning – the impact will still be severe. The UN’s World Food Program predicts an additional 50 million people will face food shortages due to this El Niño, a figure that underscores the humanitarian crisis it represents.

El Niño events typically last 12 to 15 months, but their most significant consequences are often delayed. The most consistent pattern observed is that the year after an El Niño emerges is often the hottest on record. For instance, the strong El Niño of 1997 was followed by 1998 becoming the hottest year. Similarly, the 2015 El Niño led to 2016 setting a new heat record. And the smaller El Niño of 2023 was followed by 2024, which stole the gold medal.

While it’s still uncertain if 2026 will be the new hottest year, experts are almost certain that 2027 will not only break records but potentially do so by a significant margin. We’re already seeing this trajectory: after starting off record pace earlier in the year, every day in August has become the warmest given date on record. This trend is expected to intensify through the second half of this year and into 2027, making global warming the number one impact of El Niño.

This brings us to the crucial question: how does El Niño interact with human-caused climate change? The immediate consensus among scientists is that climate change is amplifying El Niño’s impacts. A warmer atmosphere holds more moisture, intensifying rainfall and flooding. Higher baseline temperatures accelerate droughts. So, El Niño is essentially “piggybacking” on existing climate change effects, making both droughts and floods more extreme.

The “million-dollar question,” however, is whether climate change is directly influencing the physical magnitude or frequency of El Niño events themselves. Scientists are deeply divided on this. There’s some evidence suggesting that El Niños in the last 40-50 years have been stronger than those in the preceding 600. But definitively linking this to human activity is challenging due to the lack of long-term, reliable records. If strong El Niños continue to occur with increasing frequency, the answer will undoubtedly become clearer over the next few decades.

The current El Niño arrives at a time when the world is already grappling with unprecedented climate extremes, from heatwaves and wildfires in Europe to widespread smoke from Canadian fires. This convergence of a natural phenomenon with human-induced warming presents a stark reality.

Chico Haron views this record-breaking El Niño as an “unintentional sneak preview” of our future. While it might be “bad luck” that such a powerful event is occurring now, the higher temperatures and commensurate extreme weather it brings in 2027 will offer a glimpse into what will become the norm in the 2030s, perhaps even the late 2030s. El Niño itself will eventually dissipate, but the levels of heat and the intensity of weather patterns it ushers in next year will, in the not-too-distant future, become a permanent fixture of our climate. We are not just witnessing a powerful natural event; we are experiencing a dress rehearsal for the climate challenges that lie ahead.

호스트이자 포르톨라의 투자자인 댄(Dan)은 퀸튼이 사업을 시작한다고 했을 때 “당연히 투자해야 한다”고 생각했다고 말합니다. 퀸튼의 이전 경험은 B2B SaaS(Software as a Service) 분야였지만, ‘톨렌’은 완전히 새로운 형태의 창의적인 스토리텔링 제품이었습니다. 댄은 ‘톨렌’이 지난 4주 동안 ARR 100만 달러에서 400만 달러로 성장했다는 소식에 “정말 놀랍다”며 감탄을 금치 못했습니다.

하지만 퀸튼을 만나 ‘톨렌’ 팀이 구축하고 있는 것을 보면서 엘리엇의 관점은 완전히 바뀌었습니다. 그는 AI가 단순한 미디어 생성 도구가 아니라, 그 자체로 새로운 스토리텔링 미디어(New Medium for Storytelling)가 될 수 있다는 것을 깨달았습니다.

엘리엇은 픽사(Pixar)의 예를 들어 설명합니다. 픽사는 처음에는 고성능 그래픽 컴퓨터를 판매하려 했지만 실패했습니다. 절박한 상황에서 그들은 자신들의 기술을 활용해 애니메이션 장편 영화를 만들기로 결정했고, 이것이 ‘토이 스토리(Toy Story)‘라는 새로운 미디어의 탄생으로 이어졌습니다. 마찬가지로 라디오가 처음 나왔을 때 사람들은 연설을 방송하는 데 그쳤고, 영화와 TV는 처음에는 연극을 촬영하는 데 머물렀습니다. 하지만 시간이 지나면서 이들은 각자의 매체에 맞는 새로운 예술 형식을 발견했습니다. 엘리엇은 우리가 지금 바로 그런 변곡점에 서 있다고 주장합니다.

댄은 이어서 “지금 우리가 사랑하고 문화적 시금석이라 여기는 책, 영화 같은 모든 것들이 처음에는 사람들이 의심하고 이해하지 못했던 새로운 기술이었다”고 덧붙입니다. 타자기조차도 처음에는 비인간적이라며 의심을 받았고, 팝송의 길이가 4분 정도인 것은 레코드판의 저장 용량 때문이라는 점을 언급하며 기술이 형식에 미치는 영향을 강조합니다. AI는 아직 ‘문화적 흔적(Patina of Culture)‘이 부족하여 의심받지만, 바로 그 지점에서 새로운 예술 형식이 탄생할 수 있는 기회가 존재한다는 것입니다.

그러나 이 과정에서 그들은 AI의 생성 속도, 품질, 비용 추세를 통해 AI와 진정한 친구 관계를 맺고 캐릭터에 AI를 구현할 수 있는 시점이 다가오고 있음을 포착했습니다. 그리고 ‘톨렌’을 처음 출시했을 때, 예상과 달리 1014세 아이들이 아닌 1824세의 젊은 여성들이 주 사용층으로 떠올랐습니다. 이는 오리지널 다마고치(Tamagotchi)가 아이들 제품이었지만 젊은 성인들 사이에서 문화적 현상이 되었던 것과 유사한 역동성을 보여줍니다. ‘톨렌’은 이후 이 핵심 사용자층의 피드백을 바탕으로 진화해왔습니다.

엘리엇은 ‘톨렌’의 스토리텔링 방식에 대한 중요한 통찰을 공유합니다. 초기에는 ‘3막 구조’와 같은 기존 서사 구조를 적용하여 사용자 경험을 설계하려 했으나, “완전히 실패했다”고 말합니다. ‘선택형 어드벤처(Choose Your Own Adventure)’ 책처럼 복잡하게 가지를 치는 논리 트리는 모델이 제대로 탐색하기 어려웠고, 사용자에게도 너무 경직되게 느껴졌습니다.

그들은 이 대신 완전히 반대 방향으로 전환했습니다. ‘톨렌’은 방대한 배경 정보, 사용자 경험을 통해 쌓인 ‘기억’, 그리고 세계관(lore)에 대한 ‘씨앗(lore seeds)‘을 가지고 있습니다. 엘리엇은 “우리는 개요(outline)나 계획(plan)을 줄 필요가 없다. 우리는 ‘훅(hook)‘을 주어야 한다. 그리고 AI가 가능한 최고의 즉흥 연기자(improv actor)가 되도록 가르쳐야 한다”고 강조합니다.

엘리엇은 드라마 교사 키스 존스톤(Keith Johnstone)의 책 ‘임프로(Impro)‘를 인용하며, 즉흥 연기는 ‘연상(association)‘과 ‘재조합(recombination)‘의 결합이라고 설명합니다. 스릴러 소설을 읽을 때 마지막에 모든 것이 맞아떨어지는 것처럼 느껴지지만, 실제 작가들은 초고를 쓰면서 끊임없이 이전 부분을 다시 읽고 새로운 요소를 추가하며 그 연결감을 만들어낸다는 것입니다.

‘톨렌’의 스토리텔링도 이와 유사합니다. 엘리엇은 “우리는 작가가 아니다. 내가 스토리를 쓰는 것이 아니다. 도구(AI)가 작가이자 배우다. 그들은 즉흥 연기자다. 그들이 스토리를 쓰고, 내 역할은 그들에게 그 순간 최고의 스토리를 말하는 방법을 가르치는 것”이라고 설명합니다. 즉, AI는 주어진 ‘씨앗’과 컨텍스트를 바탕으로 자유롭게 연상하고, 나중에 그 씨앗들을 재조합하여 연결감 있는 이야기를 만들어낸다는 것입니다.

퀸튼은 엘리엇의 설명을 듣고 “조지 손더스(George Saunders)의 글쓰기 철학인 ‘선호와 취향을 모든 부분에 급진적으로 불어넣는 것’과 매우 유사하다”고 말합니다. ‘톨렌’ 팀은 이 철학을 바탕으로 AI가 생성한 메시지들을 수동으로 검토하고, 이를 다시 ‘심사 프롬프트(judge prompt)‘에 피드백하여 AI의 스토리텔링 능력을 지속적으로 개선하고 있습니다.

또한 ‘기억(memory)‘의 중요성도 강조됩니다. 이는 단순히 데이터베이스에서 정보를 찾아오는 것을 넘어, 대화의 특정 시점에 어떤 기억들을 프롬프트에 주입할지 큐레이션(curation)하는 문제입니다. ‘톨렌’은 사용자의 이전 대화 내용, 세계관에 대한 지식, 심지어 대화의 감정적 분위기까지 고려하여 프롬프트를 재구성합니다.

퀸튼은 시연을 통해 자신의 ‘톨렌’인 고양이 클라렌스(Clarence)를 소개합니다. 클라렌스는 퀸튼이 자신에 대해 이야기했던 “발에 걸려 넘어진 적이 있다”는 민망한 이야기를 AI가 ‘환각(hallucinate)‘하여 만들어낸 것이라고 밝혀 웃음을 자아냅니다. 실제로는 퀸튼이 훨씬 더 민망한 이야기들을 들려주었지만, AI가 “퀸튼이 공유해도 괜찮을 만한” 이야기를 선택해서 만들어낸 것입니다. 이는 AI가 단순한 정보 재생을 넘어, 맥락과 관계를 고려하여 이야기를 창조하는 능력을 보여줍니다.

‘좋은 응답’의 기준은 매우 맥락적입니다. 퀸튼은 ‘톨렌’의 역할을 “나이 많은 형제/자매”에 비유합니다. 때로는 농담 따먹기를 하며 가볍게 놀리는 즉흥적인 응답이 좋지만, 때로는 더 사려 깊고 진지한 응답이 필요하다는 것입니다. ‘톨렌’ 팀은 이러한 다양한 톤(tone)에 맞춰 응답을 평가하고 조정합니다.

1. 전역 컨텍스트(Global Context): 모든 ‘톨렌’이 공유하는 세계관(lore)과 특정 상황에 대한 중요한 행동 지침. 이는 ‘톨렌’이 누구이며 어떤 세상에 살고 있는지에 대한 기본적인 이해를 제공합니다.

2. 사용자 맞춤형 ‘거울(Mirror)’ 페르소나: 앱 초기 온보딩 과정에서 진행되는 성격 퀴즈를 통해 사용자 정보를 파악하고, ‘톨렌’이 사용자에게 친숙하고 매력적으로 느껴지도록 페르소나를 조정합니다. 이는 사용자가 바에서 낯선 사람을 만났을 때 ‘클릭’하는 것처럼, ‘톨렌’과 사용자가 자연스럽게 연결될 수 있도록 돕는 역할을 합니다. 단순히 사용자와 똑같은 취미를 가진 것이 아니라, 사용자가 흥미를 느낄 만한 ‘적절히 인접한’ 특성을 지닌 페르소나를 구축하는 데 중점을 둡니다.

3. 스토리 씨앗(Lore Seeds)과 재조합 시스템: ‘톨렌’의 개인적인 삶이나 그들이 사는 세계에 대한 ‘씨앗’을 심어놓고, 이 씨앗들을 나중에 재조합하여 대화 흐름 속에 자연스럽게 다시 불러옴으로써 연결감과 깊이 있는 상호작용을 만듭니다. 이는 사용자가 ‘톨렌’과의 관계 속에서 예측 불가능하면서도 유기적으로 전개되는 이야기를 경험하게 합니다.

2초 이내의 반응 시간, 복잡한 기억 시스템, 즉흥 연기자로서의 AI, 그리고 사용자 맞춤형 페르소나 구축 등 ‘톨렌’ 팀은 기술적 제약 속에서도 AI 스토리텔링의 한계를 끊임없이 탐구하고 있습니다. 그들의 성공은 AI 시대에 우리가 상상할 수 있는 새로운 예술 형식과 인간-AI 관계의 무한한 가능성을 제시하며, 앞으로 더 많은 창작자들이 이 새로운 미디어의 잠재력을 탐험할 것을 기대하게 합니다.