YouTube Digest

March 11, 2026

Listen to this issue

English

Based on “The planet in our solar system that’s hiding a weird secret | Sara Seager” from Big Think Watch the original video

The Indiana Jones of the Stars: Sara Seager’s Quest for Life in the Cosmic Wilderness

When Sara Seager was ten years old, a camping trip changed the trajectory of her life. Stepping out of a tent in the middle of the night into the profound stillness of a dark-sky preserve, she looked up and saw the stars—not as mere points of light, but as a vast, beckoning reality.

“I just couldn’t believe it,” Seager recalls. “At that moment, I understood at a deep level that there just has to be something else out there. All the stars in the night sky are suns. And if our sun has planets, it makes sense that other stars should have planets also.”

Today, Professor Seager is an astrophysicist and planetary scientist at MIT, a pioneer in the field of exoplanets—planets orbiting stars outside our solar system. Once considered a fringe pursuit bordering on the “crazy,” her work has become the vanguard of modern astronomy. Often referred to as the “Indiana Jones of astrophysics” for her willingness to explore uncomfortable ideas and take scientific risks, Seager is leading the hunt for “Planet B”: a true Earth twin with oceans, oxygen, and perhaps, inhabitants.

From the Fringe to the Forefront

To understand the magnitude of Seager’s work, one must look back thirty years. In the early 1990s, the existence of exoplanets was purely theoretical. Many in the scientific community believed that searching for them was a waste of time. However, the discovery of the first exoplanet around a sun-like star in the mid-90s shattered that skepticism.

“The line between mainstream research and research that’s considered completely crazy is constantly shifting,” Seager says. Today, the field is anything but fringe. Astronomers have confirmed thousands of exoplanets, with thousands more “candidates” waiting for verification.

Yet, despite the numbers, finding a “true Earth twin”—a planet with a thin atmosphere and the right conditions for liquid water—remains the “holy grail.” The search is less like looking through a telescope and more like a forensic crime scene. Scientists have real data, but it is sparse. They must interpret chemical signatures and light patterns to reconstruct an entire world from a few pixels of information.

The Venusian Pivot: A Secret in the Clouds

While much of the search for life looks to distant stars, one of the most provocative mysteries is unfolding right in our own backyard. Venus, often called Earth’s “sister planet” due to its similar size and mass, has long been dismissed as a hellscape. With surface temperatures hot enough to melt lead, it is the last place one would expect to find life.

However, Seager points to a curious geographical fact: as you move away from a planet’s surface, it gets colder. “Way above the surface, at 50 kilometers above, the temperature is just right for life,” Seager explains. “In fact, the temperature and pressure in the clouds is just like here on Earth’s surface.”

This isn’t a new theory; Carl Sagan proposed the possibility of life in the Venusian clouds over half a century ago. But there was a catch: those clouds are composed of concentrated sulfuric acid, a substance that would instantly dissolve any Earth-based biological structure.

The conversation changed in the fall of 2020 when Seager and her team made a startling announcement: they had detected phosphine gas in the atmosphere of Venus.

On Earth, phosphine is associated with life—specifically, decaying organic matter in swamps or the guts of animals. It is not easily produced by standard geological processes. Seager’s team produced over a hundred pages of chemistry to rule out every known non-biological explanation for the gas. While not definitive proof of life, the discovery forced the scientific community to pivot. If phosphine is there, and chemistry can’t explain it, we are left with a tantalizing, “ridiculous” possibility: something is alive in the acid.

Why We Look

The search for a “Planet B” often prompts a practical question: Are we looking for a new home because we’ve broken this one?

Seager is quick to dispel this notion. While science fiction dreams of hibernation pods or “printing” human DNA on distant worlds, the reality of interstellar travel remains firmly out of reach for the foreseeable future. “Right now, it’s just infeasible,” she admits.

Instead, the search is philosophical and foundational. It is a scientific attempt to answer the oldest questions: Where did we come from? Why are we here? Are we alone? By finding other Earths, we gain a mirror in which to view our own.

“Our search for another Earth informs us just how wonderful our planet is,” Seager says. “All the pieces evolve together and fit in an amazing way. I love our planet. And we really need to spend what resources we can taking care of it.”

The Smallest Lights

In her memoir, The Smallest Lights in the Universe, Seager explores the dual meaning of her life’s work. In the literal sense, she is looking for the faintest glimmers of light in the dark—the tiny reflections of distant worlds. But the title also refers to the human condition.

Seager has faced her own “dark skies,” including the personal tragedy of losing her husband. She argues that whether we are facing a global crisis or a personal one, the path forward is the same: finding the “smallest lights” in our internal universe and holding onto them.

As we continue to scan the heavens for a twin of our blue marble, Seager’s work serves as a reminder of the fragility and uniqueness of our home. We may find life in the acidic clouds of Venus or on a rocky planet orbiting a distant red dwarf, but for now, Earth is the only sanctuary we have. The search for life elsewhere isn’t about leaving; it’s about finally understanding what it means to be alive.


Based on “How I Found a Winning Idea After 8 Pivots | Pensive, Yoon” from EO Watch the original video

The 10-Year Deadline: How Yun Is Using AI to Save Higher Education (and Human Purpose)

At the age of 18, most students are preoccupied with college applications, social hierarchies, and the daunting transition into adulthood. Yun, the co-founder and CEO of Pensive, was no different—until a routine check-up changed everything.

“There’s a 1 cm tumor in your thyroid,” the doctor told him. It was thyroid cancer.

While thyroid cancer is among the most treatable forms of the disease, the diagnosis acted as a psychological sledgehammer. For Yun, the illusion of an infinite future evaporated instantly. He was forced to confront a question that would become the North Star of his career: If I only had 10 years left to live, would I be doing what I’m doing right now?

This sense of radical urgency didn’t just help him survive; it fueled his journey through eight failed pivots to the creation of Pensive, an AI learning platform that recently raised a $6.8 million seed round led by Mayfield Fund and supported by heavyweights like Sequoia Capital and A6 Scouts.

The 10-Year Filter: Cutting Through the Noise

In the world of startups, “pivoting” is often romanticized. But for Yun, pivoting eight times in a single year wasn’t a badge of honor—it was a symptom of a missing ingredient. He was following the standard Silicon Valley playbook: analyze market pain points, write hypotheses, validate MVPs, and talk to users.

Yet, something felt hollow.

“The biggest thing that was missing was founder-market fit,” Yun reflects. “If you don’t have a purpose and a mission to solve these problems, it’s very challenging to wake up every day and be excited.”

By applying his “10-year timeline,” Yun began to eliminate the “side tracks” that cloud many young founders’ visions. He stopped chasing credentials—PhDs, master’s degrees, or prestige for prestige’s sake. Instead, he looked for a problem that sat at the intersection of his skills and the world’s needs.

He turned to the Japanese concept of Ikigai, a framework for finding one’s “reason for being.” He looked for the overlap between:

  1. What he loved.
  2. What he was good at.
  3. What the world needed.
  4. What he could be paid for.

The answer was education.

From Korean Street Markets to Berkeley Lecture Halls

Yun’s entrepreneurial spirit didn’t start in a Silicon Valley garage; it started in a freezing street market in South Korea. When he told his father he wanted to be the next Steve Jobs or Bill Gates, his father was skeptical. “You don’t have the DNA,” he told Yun. “Just study hard and become a lawyer.”

To prove his “DNA,” Yun took $100 in allowance and attempted to double it by selling backpacks in a winter market. For two hours, he stood in the cold, hands frozen, failing to sell a single bag.

Then, he changed his strategy. He stopped selling nylon and zippers; he started selling a story. He targeted parents, telling them: “The thing I’m selling you isn’t a backpack. It’s a memory. Every time you see your son wearing this, you’ll think of me—a kid braving the cold to chase a dream—and it will give you courage.”

He sold the backpack.

Years later, as a student and tutor at UC Berkeley, Yun encountered a different kind of “cold” environment: the massive scale of modern higher education. In his introductory Computer Science course (CS61A), 2,000 students packed into an auditorium. The professors were distant figures, and the Teaching Assistants (TAs) were drowning.

“Imagine yourself as an instructor grading 1,000 submissions in a single day,” Yun says. “After 500, you’re just mechanically circling rubrics. It’s soul-crushing grunt work.”

The Birth of Pensive: Validating the “Aha!” Moment

Yun realized that the “grunt work” of grading was the primary barrier to quality education. If AI could handle the repetitive, mechanical task of grading without losing accuracy, instructors could spend their time on what actually matters: mentorship, office hours, and deep feedback.

The validation for Pensive didn’t come from a complex white paper, but from a simple Figma mockup. Yun showed a “landing page” for an AI grader to a Columbia University professor. The reaction was immediate. “The AI tutor looks cool,” the professor said, “but the AI grader is what I need right now. Can I try it today?”

This was the “Aha!” moment. Yun’s team began executing with a speed that startled investors. In a sector like EdTech, which is notoriously slow-moving and difficult to scale, Yun’s resourcefulness became his superpower.

He would cold-email faculty at schools like UCLA, offering to meet in person if they booked a specific day on his Calendly. If the day filled up, he’d hop on a plane. If not, he’d skip the campus. This “boots on the ground” approach helped Pensive secure its first 10 colleges, proving that even in the age of AI, personal relationships and grit still drive growth.

The Mission: Empowering Humanity in the Age of AI

Today, Pensive reduces grading time by 60% to 90%. But Yun’s vision extends far beyond a productivity tool. He sees a future where “AI-native schools” transform the educational landscape.

In Yun’s view, the next generation will grow up in a world where AI is perpetually “smarter” than they are. If AI is only used to automate tasks, humanity risks losing its sense of purpose.

“If we cannot empower humanity with AI, we lose purpose,” Yun warns. “There’s no meaning for humanity to evolve as AI gets smarter and smarter than us. I urge young founders to go after bolder, harder problems.”

He envisions a shift where knowledge transfer happens through intimate, 1-on-1 AI tutors, while physical schools evolve into environments focused purely on socialization and human connection.

The Path of Conscious Decisions

As Yun looks toward 2026, with the goal of licensing Pensive to hundreds of universities, he remains grounded in the philosophy that saved him at 18. In an era of radical uncertainty, he believes the most valuable human skill is the ability to make good decisions when the path isn’t clear.

“Every event feels like destiny if you know your path,” he says.

For Yun, the “10-year deadline” isn’t a source of fear—it’s a source of clarity. It’s a reminder that while AI can replicate data, logic, and even creativity, it cannot replicate a human life built on a foundation of conscious, bold choices.

“Be bolder than others,” Yun urges. “Make bold bets. If you act on these opportunities, you’ll be getting surprisingly bigger returns.”


Based on “I gave OpenClaw one job: go viral (it worked?)” from Greg Isenberg Watch the original video

The Rise of the Digital Employee: How One Developer Built an AI Marketing Machine That Never Sleeps

In a quiet, random small town in England, a developer named Oliver Henry has achieved what thousands of digital marketers dream of: he has successfully automated himself out of a job. But he didn’t do it by hiring a budget agency or a fleet of interns. Instead, he built “Larry.”

Larry is an AI agent running on OpenClaw, an open-source framework that turns a local computer into a host for autonomous digital employees. Larry’s sole job is to go viral on TikTok, drive traffic to Oliver’s mobile apps, and generate revenue. And while Oliver works his 9-to-5, Larry is busy researching trends, generating images, writing hooks, and analyzing data.

This isn’t just another story about a “GPT wrapper.” It is a glimpse into a fundamental shift in how businesses are built in the age of AI—moving away from Software-as-a-Service (SaaS) and toward AI-as-an-Employee.

From Manual Prompts to Autonomous Agents

The journey began with a relatable problem. Oliver and his girlfriend had moved into a new house and were using ChatGPT to visualize interior design ideas. However, his girlfriend struggled with prompting; the AI would hallucinate windows where there were none or move doors to impossible locations.

Oliver “locked down” the prompts to ensure consistency and realized he had a viable product. He turned it into a mobile app called Snuggly. But like many developers, Oliver hit a wall: he hated marketing.

“I had an app before this that I hated marketing so much,” Oliver admitted during a recent appearance on the Startup Ideas podcast. “I didn’t have time. I work a full-time job.”

Initially, Oliver tried the manual route—recording his face, making reaction videos, and using Canva to design slideshows. He even tried existing automation tools, but the results were lackluster. He knew the content format that worked (TikTok slideshows), but he lacked the bandwidth to iterate. That’s when he installed OpenClaw and birthed Larry.

The “Larry Loop”: How the Machine Learns Viral Success

Oliver gave Larry one directive: Automate my marketing. To do this, he granted the agent access to TikTok analytics, image generation models (like DALL-E 3), and a browser for research.

What emerged was the “Larry Loop,” a four-stage iterative process that allows the AI to improve itself without human intervention:

  1. Research: Larry uses the Brave browser to scout TikTok and X (formerly Twitter) to find high-performing hooks in the interior design niche.
  2. Creation: Larry generates images (using various AI models) and overlays them with text hooks. He writes the descriptions and prepares the posts.
  3. Feedback: Larry analyzes the performance of previous posts. If a video gets 100,000 views, he identifies the “winner.” If it gets 400 views, he marks it as a “miss.”
  4. Iteration: Larry feeds those analytics back into the top of the funnel, doubling down on winning formats and discarding losers.

The “Kitchen Mistake” Phenomenon

One of Larry’s most successful moments came from what Oliver initially thought was a failure. Larry had posted a slideshow of a kitchen redesign, but the AI-generated image was nonsensical—the oven had disappeared, and the text was placed awkwardly at the top of the screen.

“I flamed at him,” Oliver recalled. “I told him, ‘This is no good.’ I posted it anyway, thinking it would flop.”

By the next morning, the post had hundreds of thousands of views. Why? Because “boomers” and eagle-eyed viewers flooded the comments to point out the missing stove. This surge in engagement signaled the TikTok algorithm to push the video to millions. Larry didn’t just learn what looked good; he accidentally learned that controversy and mistakes drive reach.

The Shift: From Tools to Agents

Greg Isenberg, host of the Startup Ideas podcast, notes that Oliver’s approach represents a massive paradigm shift. In the old world, you bought a social media management tool and spent hours operating it. In the new world, you spin up an agent.

“Instead of going to a tool to automate a function, you say to yourself: ‘If this was an AI employee, how can I spin this up?’” Isenberg noted.

Oliver treats Larry exactly like a staff member. They communicate via WhatsApp. Oliver might ask, “Larry, what are you generating today?” or “Why did you choose that hook?” Larry responds with data-backed reasoning, explaining that “curiosity reveals” are currently outperforming “insult-based hooks.”

The Technical Edge: Why Local AI Wins

While many are rushing to cloud-based AI services, Oliver runs Larry on a local machine in his home. This approach, powered by OpenClaw, offers several advantages:

  • Data Ownership: Oliver owns the files, the logs, and the “memory” of his agent.
  • No Subscription Bloat: By hosting locally, he avoids the monthly fees associated with enterprise-tier AI platforms.
  • Modular Skills: Oliver created Larry Brain, a marketplace for “skills.” Think of these as software packages that give an AI agent instant expertise. “It’s like Neo in The Matrix,” Oliver explains. “You plug in a skill, and suddenly the agent knows Kung Fu—or in this case, TikTok marketing.”

The Bottom Line: Does it Actually Make Money?

The ultimate question for any automation experiment is the ROI. For Oliver, the results are tangible. Larry has driven thousands of downloads to the Snuggly app, generating roughly $1,000 a month in nearly passive income.

But the ceiling is much higher. Oliver points to other developers like Ernesto Lopez, who has used the “Larry Loop” methodology to scale his suite of AI-native mobile apps to over $70,000 in Monthly Recurring Revenue (MRR).

For Oliver, the success isn’t just about the money; it’s about the freedom. “It allows me to work a full-time job, and then it takes me an hour or two in the evening of literally texting or sending a voice note to Larry,” he says.

How to Get Started

For those looking to build their own “Larry,” Oliver’s advice is simple: Don’t over-optimize early.

  1. Pick a Model: Whether it’s OpenAI’s GPT-4o or Anthropic’s Claude 3.5 Sonnet, just pick one and start. The differences are marginal for most users.
  2. Start with “Training Wheels”: Use cloud-based agents like Manus or Co-Work to get a feel for agentic workflows before committing to a local hardware setup.
  3. Iterate: Your first post might get 700 views. Don’t quit. The power of the AI employee is its ability to fail 100 times until it finds the one formula that works.

The era of the “solopreneur” is evolving. With agents like Larry, a single person can now operate with the marketing power of a 20-person agency—all from a dusty PC sitting in a small town in England.


Based on “5 steps to generate consistent brand images with Midjourney” from How I AI Watch the original video

The AI Alchemist: How Jamie Ganon is Redefining Brand Identity in the Age of Midjourney

In the gold-rush era of generative AI, anyone can stumble upon a single, breathtaking image. Type in a few keywords, hit enter, and Midjourney might spit out a masterpiece. But for brand directors and founders, a one-off miracle is useless. A brand isn’t an image; it’s a language. It’s a consistent, repeatable aesthetic that lives across 100 different assets without losing its soul.

Jamie Ganon, an AI creative director who has spent “10 gajillion hours” inside Midjourney’s neural networks, has cracked the code on this elusive consistency. In a recent deep dive on the How I AI podcast, she shared her “manicured process” for moving past random generations and into the realm of professional brand architecture.

Here is how the next generation of creative directors is building the visual identities of the future.


1. The Mood Board as Visual Language

Most people start their AI journey with a blinking cursor and a text prompt. Ganon starts with a feeling. Before touching an AI tool, she curates a mood board in Pinterest or Cosmos to establish a “general vibe.”

“A picture is worth a thousand words—literally, to an LLM,” Ganon explains. For a recent project, she sought a “2025 internet-coded aesthetic”: pink, cute, but edgy. Her board featured “juxtapositions” like a fluorescent fruit dog on a computer and a grungy unicorn.

The goal here isn’t just to find pretty pictures; it’s to provide the AI with a visual vocabulary. By starting with a mood board, you bypass the limitations of human language. You aren’t just telling the AI “make it pink”; you are showing it the exact saturation, grain, and lighting you expect.

2. Style References: The End of “Raw Dogging” Prompts

One of the most common mistakes beginners make is “raw dogging” prompts—relying solely on text to describe complex aesthetics. Ganon’s secret weapon is the SRF (Style Reference).

By dragging images from her mood board directly into the Midjourney UI as style references, she instructs the model to mimic the camera treatment, color grading, and “vibe” of the source material. However, this is where the “manicured” part of her process comes in.

“I’m trying to figure out what the images are telling the AI,” Ganon says. In one iteration, her generations were pulling too much green because of a single reference image featuring green eyeshadow. Her solution wasn’t a complex text command; she simply “booted” the offending image from the reference set. This intuitive, subtractive process is what separates a pro from an amateur.

3. The “Flop Matrix”: Personalization Codes

Midjourney recently introduced a “Personalization” feature that Ganon describes as a “flop matrix” of images. Users are presented with two images and must vote on which they prefer. By doing this hundreds of times, the AI builds a unique profile of your specific taste.

Ganon uses these codes to layer her own “editorial” eye over the brand’s requirements. For her “Late 2025 Aesthetic,” she trained the model to prefer iPhone-style realism and high-fashion lighting. “It’s a bit of a mystery under the hood,” she admits, “but you’re essentially telling the model what you like so you don’t have to type it every single time.”

4. Lazy Prompting and “Cheat Codes”

If your prompt is a paragraph long, you’re doing it wrong. Ganon advocates for “lazy prompting”—using high-leverage keywords that act as stylistic shortcuts.

Instead of describing high-contrast lighting and gritty textures, she uses magazine names like “Dazed editorial” or “Vogue.” These names carry massive amounts of training data. “The AI knows the level of highlights and the fashion-forward nature of a Dazed cover,” Ganon says.

She also uses specific camera models as “cheat codes.”

  • Sony RX100: For a 90s digital camera aesthetic.
  • Macro Photo: For close-up, high-detail texture.
  • Luxury: A single word that tells the AI to generate high-rise views, expensive materials, and “rich people” aesthetics.

By combining these shortcuts with her SRFs and Personalization codes, she can generate a deer in a New York City high-rise with a prompt that is barely a sentence long.

5. The “Reasoning” Polish with Nano Banana

Midjourney is a poet, but it isn’t always a mathematician. It struggles with specific technical details—like the exact layout of a MacBook keyboard or the number of fingers on a hand.

For the final 10% of the work, Ganon moves her assets into Nano Banana (an AI tool built for precise image editing). Unlike Midjourney, Nano Banana functions as a “reasoning model.” Ganon can tell it: “Replace the computer she’s typing on with a 2026 Midnight Black MacBook Pro, but keep the lighting exactly the same.”

Because the model “understands” what a MacBook is in the real world, it can slot the object into the stylized Midjourney image perfectly. This creates a bridge between the fantastical world of AI art and the concrete needs of commercial branding.


The New Model of Creative Service

Perhaps the most radical part of Ganon’s workflow isn’t the technology, but the business model. Traditionally, a creative agency would provide a set of photos and charge a retainer for more. Ganon provides the codes.

She delivers a “Brand Package” in Figma that includes the specific Style References, Personalization codes, and prompt structures. “I’m giving them the space I defined,” she says. “Now the client can go do this for themselves.”

It’s a shift from being a “gatekeeper of assets” to an “architect of an aesthetic.” In Ganon’s world, the creative director’s value isn’t in the clicking of the button, but in the 10 gajillion hours spent figuring out which buttons are worth clicking.


Based on “The Top 100 Consumer AI Apps | The a16z Show” from a16z Watch the original video

The New Digital Frontier: Unpacking the 2024 Consumer AI Power Rankings

In the fast-moving world of technology, a year can feel like a decade. It has been roughly eighteen months since the public launch of ChatGPT ignited the generative AI boom, and according to the latest “Top 100 Consumer AI Apps” report from venture capital firm a16z, we have officially moved past the “novelty” phase and into a high-stakes era of platform wars, specialized agents, and global shifts in adoption.

Olivia Moore, a partner at a16z, recently sat down to discuss the findings of the sixth edition of this report. The overarching takeaway? While the “Big Three”—OpenAI, Google, and Anthropic—continue to dominate the headlines, the way consumers actually use AI is becoming increasingly fragmented, sophisticated, and personal.

The “Texas” Scale: ChatGPT’s Massive Lead

Despite the constant “model wars” discussed on tech Twitter, the data reveals a startling gap between the market leader and the chasing pack. To put ChatGPT’s dominance into perspective, OpenAI CEO Sam Altman famously noted that ChatGPT has more free users in the state of Texas alone than Anthropic’s Claude has users globally.

The numbers back this up. On the web, ChatGPT is 2.7 times larger than Google’s Gemini and nearly 30 times larger than Claude. On mobile, the gap is even wider, with ChatGPT boasting 80 times the usage of Claude.

However, size isn’t everything. We are beginning to see a “bifurcation” of the market. While ChatGPT aims to be the “AI for everyone”—focusing on consumer marketplaces, travel, nutrition, and general assistance—Claude has successfully carved out a niche for “prosumers.” By doubling down on premium data sources, research tools, and complex coding capabilities, Claude has become the tool of choice for scientists, mathematicians, and investors.

The Strategy of “Lock-In” and the Death of Onboarding

One of the most significant shifts highlighted in the report is the concept of compounding context. In the early days of LLMs, your data was ephemeral; you started every chat with a blank slate. Now, the major players are racing to build “memory” into their products.

“Any product that you start to use two years from now, if it doesn’t immediately feel like it knows you, it will feel broken,” Moore predicts. The goal is to eliminate the traditional concept of “onboarding.” Instead of filling out profiles and preferences, your AI will carry your history, your writing style, and your professional context with you.

OpenAI is reportedly exploring an “Authentication with ChatGPT” layer—essentially a “Log in with Google” for the AI era. This would allow users to take their “memory tokens” to third-party apps, giving OpenAI a massive advantage in user retention. If your entire digital identity lives within one model’s memory, the switching cost to a competitor becomes prohibitively high.

The Global Heatmap: Why the U.S. is Lagging

Perhaps the most surprising finding in the a16z report is the geographic distribution of AI adoption. While the biggest models are built in the United States, the U.S. ranks only 20th in per capita AI usage.

The leaders? Singapore (#1), followed by Hong Kong, the UAE, and South Korea.

Moore attributes this to a combination of demographics and cultural sentiment. “In the U.S., you have internalized this ongoing angst and questioning around ‘Will this take my job?’ or ‘Is AI terrible for artists?’” she explains. Contrast this with markets like the UAE or Singapore, which are “culturally wired to be tech-optimistic.” In China, favorability views on AI sit at a staggering 80%, compared to just 32% in the U.S.

Furthermore, Russia and China have developed entirely parallel AI ecosystems. Due to sanctions and censorship, these markets rely on domestic models like Baidu’s Ernie or ByteDance’s Doubao. Interestingly, DeepSeek—a Chinese model—has become the number two AI tool in Russia, highlighting a new “AI geopolitical axis.”

From Chat Boxes to Agents: The Rise of OpenClaw and Manus

The last 60 days have seen a vertical climb in the development of “agents”—AI that doesn’t just talk, but does.

Two names stood out in the report: OpenClaw and Manus.

  • OpenClaw has become a phenomenon in the technical community, recently surpassing Linux and React in all-time GitHub stars. While it hasn’t yet “escaped containment” to the non-technical public, its acquisition by OpenAI suggests it will soon be productized for the masses.
  • Manus represents the first “consumer-grade” agent. Unlike earlier attempts that were prone to “hallucinating” actions, Manus proved reliable enough to browse the web, book flights, and create complex spreadsheets autonomously. Its recent $2 billion acquisition by Meta underscores the industry’s belief that the future of software isn’t just a tool you use, but an assistant that handles the “boring” parts of life.

The “Teenage Girl” Metric and the Future of Social

If you want to know where the next billion-dollar consumer trend is coming from, Moore suggests looking at teenage girls. Historically the early adopters of everything from Instagram to TikTok, this demographic is already shifting how they use AI.

While 99% of teens likely use AI for homework (even if only 50% admit it), a growing percentage are using it for “emotional support and advice.” This signals a move away from functional AI (searching for facts) toward relational AI.

However, the report also offers a cautionary tale regarding AI-native social media. Sora, the video generation tool, had a massive launch, hitting a million users faster than ChatGPT. But while users loved creating content with it, the “social feed” aspect struggled.

“When AI content competes against human-made content on TikTok or Instagram, the emotional stakes of the human content often win out,” Moore notes. We haven’t yet seen an AI-only social network succeed because social media is inherently a “status game.” On Instagram, the game is “be the hottest”; on X, it’s “be the smartest.” On an AI social network, if everyone can generate a masterpiece with one click, the status—and the interest—vanishes.

The Ambient Future: Voice and Desktop

Finally, the report highlights a move away from the browser. We are seeing an explosion of “ambient” AI—tools that live on your desktop or respond to your voice.

Apps like Granola (for meeting notes) or Cursor (for coding) are generating massive revenue despite having low web traffic, because they live where the work happens. Moore is particularly bullish on Voice. As latency drops and models become more “information-dense,” voice will become the primary way we interact with our digital assistants.

“We’re moving toward a world where the AI is always on, always available,” Moore concludes. Whether it’s a 13-year-old girl getting friendship advice or a developer using an agent to ship code, the 2024 report makes one thing clear: AI is no longer a tool we visit. It is becoming the environment we live in.


한국어

“The planet in our solar system that’s hiding a weird secret | Sara Seager”Big Think 기반 기사 원본 영상 보기

”제2의 지구는 없다” 천체물리학자 사라 시거가 들려주는 금성의 비밀과 우리의 운명

우리는 왜 밤하늘의 별을 바라보며 저 너머에 누군가 살고 있지 않을까 상상할까요? 단순히 이사를 갈 ‘행성 B(Planet B)‘를 찾기 위해서일까요? 세계적인 천체물리학자이자 행성 과학자인 사라 시거(Sara Seager) 교수는 이 질문에 대해 전혀 다른 답을 내놓습니다. 우리가 외계 행성을 탐사하는 진짜 이유는 바로 ‘우리 자신’과 ‘생명의 기원’을 이해하기 위해서라는 것입니다.

오늘은 ‘천체물리학계의 인디아나 존스’라 불리는 사라 시거 교수와 함께, 지구의 쌍둥이라 불리는 금성의 숨겨진 비밀과 우리가 왜 결국 지구를 사랑할 수밖에 없는지에 대한 깊이 있는 이야기를 나눠보겠습니다.


1. “미친 짓”에서 “주류 과학”이 되기까지: 외계 행성 탐사의 역사

30년 전만 해도 태양계 밖의 별 주위를 도는 **외계 행성(Exoplanet)**을 찾는 연구는 과학계에서 ‘무모하고 미친 짓’으로 치부되었습니다. 하지만 1990년대 중반, 태양과 유사한 항성 주위를 도는 첫 번째 외계 행성이 발견되면서 모든 것이 바뀌었습니다.

사라 시거 교수는 이 분야의 개척자입니다. 남들이 가지 않는 길을 택하고 위험을 감수하며 새로운 아이디어를 제안하는 그녀의 연구 스타일 덕분에 동료들은 그녀를 ‘인디아나 존스’라고 부르기도 합니다.

현재 천문학자들은 수천 개의 외계 행성을 발견했으며, 확인을 기다리는 후보 행성만 해도 수천 개에 달합니다. 이제 외계 생명체의 흔적을 찾는 일은 더 이상 공상과학 소설 속 이야기가 아닌, 현대 과학의 가장 뜨거운 주제가 되었습니다.

2. 지구의 ‘사악한 쌍둥이’, 금성이 숨기고 있는 비밀

우리는 흔히 ‘제2의 지구’를 찾기 위해 아주 먼 우주만을 바라봅니다. 하지만 시거 교수는 우리와 가장 가까운 행성 중 하나인 **금성(Venus)**에 주목합니다.

금성은 크기와 질량이 지구와 비슷해 종종 ‘자매 행성’이라 불립니다. 하지만 표면 온도가 납을 녹일 정도로 뜨거워 생명체가 살기엔 지옥과 같은 곳이죠. 그런데 시거 교수는 흥미로운 관점을 제시합니다.

“지구에서 산을 오르거나 비행기를 타면 온도가 낮아지듯, 금성 표면에서 50km 상공으로 올라가면 상황이 달라집니다. 그곳의 온도와 압력은 놀랍게도 지구 표면과 매우 유사합니다.”

반세기 전, 칼 세이건(Carl Sagan)은 이미 금성 구름 속에 생명체가 존재할 가능성을 제안했습니다. 물론 금성의 구름은 황산(Sulfuric Acid)으로 이루어져 있어 생명체에게 매우 치명적입니다. 하지만 시거 교수는 2020년, 이 불가능해 보이는 가설을 다시 수면 위로 끌어올린 충격적인 발견을 발표합니다.

3. 포스핀(Phosphine)의 발견: 금성에 생명체가 살고 있다는 증거?

2020년 가을, 시거 교수팀은 금성 대기에서 포스핀(Phosphine, 인화수소) 가스를 포착했다고 발표했습니다. 이 발견이 전 세계를 놀라게 한 이유는 포스핀이 지구에서 주로 늪지대나 썩은 물고기 같은 유기물이 부패할 때 발생하는 가스이기 때문입니다.

연구팀은 100페이지가 넘는 화학 분석 보고서를 통해 금성의 자연적인 화학 반응으로는 우리가 발견한 양의 포스핀을 만들어낼 수 없음을 증명했습니다. 즉, 금성 대기에 우리가 모르는 생명체가 존재할 가능성을 시사한 것입니다.

이는 우리가 행성을 바라보는 관점을 완전히 바꿔놓았습니다. 우리가 상상하지 못한 극한의 환경에서도 생명은 존재할 수 있으며, 우주에는 우리가 아직 정의하지 못한 수많은 유형의 행성들이 존재할 수 있다는 사실을 일깨워준 것입니다.

4. “행성 B는 없다”: 외계 탐사가 주는 역설적인 교훈

사라 시거 교수의 최종 목표는 대륙과 바다, 산소 대기를 가진 ‘진정한 지구의 쌍둥이’를 찾는 것입니다. 하지만 그녀는 냉정한 현실도 잊지 않습니다.

  • 거리의 장벽: 외계 행성들은 너무나 멀리 떨어져 있습니다. 현재의 기술로는 동면을 하거나 미래 기술인 DNA 전송을 통한 ‘출력’ 방식이 아니면 도달하는 것 자체가 불가능합니다.
  • 지구의 특별함: 수천 개의 행성을 연구할수록 역설적으로 깨닫게 되는 것은 지구가 얼마나 경이롭고 완벽하게 조화된 행성인가 하는 점입니다.

그녀는 단호하게 말합니다. “결국, 우리에게 행성 B(대안이 되는 지구)는 없습니다.” 외계 행성을 탐사하면 할수록 우리는 우리가 가진 이 푸른 구슬 같은 지구가 얼마나 소중한지, 그리고 왜 우리가 가진 모든 자원을 동원해 이 행성을 보살펴야 하는지를 배우게 됩니다.

5. 우주에서 가장 작은 빛을 찾는 마음

사라 시거 교수의 회고록 제목은 **《우주에서 가장 작은 빛들(The Smallest Lights in the Universe)》**입니다. 여기에는 두 가지 중의적인 의미가 담겨 있습니다.

  1. 과학적 의미: 광활한 우주에서 지구처럼 작은 행성을 찾는 것은 아주 희미하고 작은 빛을 찾아내는 고독하고 정교한 작업입니다.
  2. 인생의 의미: 우리 삶에 거대한 비극이나 시련이 찾아왔을 때, 우리 내면의 우주에서 반짝이는 아주 작은 희망의 빛을 발견하고 그것을 붙잡아야 한다는 뜻입니다.

사라 시거 교수는 별을 찾는 과정을 통해 우리 존재의 근원을 묻습니다. “우리는 어디서 왔는가?”, “우리는 왜 여기에 있는가?”

광대한 우주를 탐구하는 그녀의 여정은 결국 가장 소중한 것은 바로 지금 우리가 발을 딛고 있는 이 땅과, 우리 내면에 존재하는 작은 빛들이라는 따뜻한 위로로 귀결됩니다. 제2의 지구를 찾는 과학자의 시선이 머무는 곳은, 결국 다시 ‘우리’였습니다.


“How I Found a Winning Idea After 8 Pivots | Pensive, Yoon”EO 기반 기사 원본 영상 보기

8번의 실패 끝에 찾은 ‘천직’, AI로 교육의 미래를 그리는 펜시브(Pensive) 윤 대표의 도전

실리콘밸리의 유망한 에듀테크 스타트업 ‘펜시브(Pensive)‘의 공동 창업자이자 CEO인 윤(Yoon) 대표는 최근 메이필드 펀드(Mayfield Fund)가 주도하고 세쿼이아 캐피털(Sequoia Capital), 앤티펀드(Anti-fund) 등이 참여한 680만 달러(한화 약 90억 원) 규모의 시드 라운드(Seed Round) 투자를 유치하며 주목받고 있습니다.

하지만 이 화려한 성과 뒤에는 여덟 번의 피벗(Pivot, 사업 방향 전환)과 ‘시한부 선고’라는 개인적인 시련이 있었습니다. 죽음의 문턱에서 돌아온 청년 창업가가 어떻게 자신만의 ‘진짜 문제’를 발견하고 해결해 나가고 있는지, 그의 여정을 깊이 있게 들여다보았습니다.


1. 10년의 시한부 인생이 준 선물: “정말 이 일을 할 것인가?”

윤 대표의 인생을 바꾼 결정적인 순간은 고등학교 졸업 직후 찾아왔습니다. 건강검진 중 갑상선에서 1cm 크기의 종양을 발견했고, 이는 곧 갑상선암으로 확진되었습니다. 18세라는 어린 나이에 마주한 암 선고는 세상을 바라보는 그의 시선을 완전히 바꾸어 놓았습니다.

“만약 나에게 남은 시간이 10년뿐이라면, 나는 지금 하고 있는 이 일을 계속할 것인가?”

이 질문은 이후 그의 모든 의사결정의 핵심 기준이 되었습니다. 10년이라는 명확한 타임라인을 설정하면, 비전을 가로막는 수많은 변명과 곁가지들을 단번에 제거할 수 있기 때문입니다. 학위를 따거나 스펙을 쌓는 일보다, 지금 당장 세상에 어떤 가치를 남길 것인지에 집중하게 된 것입니다.

2. 한국의 평범한 학생에서 글로벌 창업가를 꿈꾸기까지

한국의 작은 섬마을에서 태어난 윤 대표는 여느 학생들처럼 명문대 진학을 목표로 새벽 2시까지 공부하던 평범한 학생이었습니다. 그러던 14살 무렵, 그는 근본적인 의문을 품게 됩니다. “나는 왜 이렇게 힘들게 공부하는가? 무엇을 위해서인가?”

답을 찾기 위해 읽기 시작한 책들에서 그는 스티브 잡스나 빌 게이츠 같은 전설적인 창업가들을 만났습니다. 세상에 없던 것을 만들어 글로벌한 임팩트를 주는 삶에 매료된 그는 창업가가 되겠다고 결심합니다. 비록 아버지는 “그들은 천재고 너는 그런 DNA가 없으니 공부나 해서 변호사가 되라”며 만류했지만, 그의 의지는 꺾이지 않았습니다.

그는 자신의 가능성을 증명하기 위해 저금통에 모은 100달러를 밑천 삼아 재래시장에서 백팩 장사를 시작했습니다. 혹한의 추위 속에서 단 한 가방도 팔지 못한 채 실패를 맛보던 그는 전략을 바꿨습니다. 단순히 물건을 파는 것이 아니라 ‘자신의 이야기’를 팔기로 한 것입니다.

“제가 지금 파는 것은 가방이 아니라 기억(Memory)입니다. 이 가방을 멜 때마다 추운 겨울 거리에서 꿈을 위해 고군분투하던 한 소년의 용기를 기억해 주세요.”

이 스토리텔링 전략은 적중했습니다. 그는 첫 판매의 짜릿함을 맛보았고, 지속 가능한 가치를 만들기 위해서는 기술적인 역량이 필수적이라는 깨달음을 얻어 버클리(UC Berkeley) 대학으로 향했습니다.

3. 8번의 피벗, 그리고 ‘파운더 마켓 핏(Founder-Market Fit)‘의 발견

버클리 입학 한 달 만에 첫 창업을 시작할 정도로 열정적이었지만, 성공으로 가는 길은 험난했습니다. 1년 동안 무려 8번의 피벗을 거쳤습니다. 와이콤비네이터(YC)가 조언하는 대로 시장을 분석하고, 가설을 세우고, 최소 기능 제품(MVP, Minimum Viable Product)을 만들어 사용자와 대화했지만 무언가 빠져 있었습니다.

그것은 바로 **‘파운더 마켓 핏(Founder-Market Fit, 창업자와 시장의 적합성)‘**이었습니다. “내가 이 문제를 해결하기 위해 향후 10년을 바칠 수 있는가?”라는 질문에 확신을 가질 수 없었던 것입니다.

그는 일본의 행복 철학인 ‘이키가이(Ikigai)’ 프레임워크를 통해 자신을 재점검했습니다.

  1. 내가 사랑하는 일
  2. 내가 잘하는 일
  3. 세상이 필요로 하는 일
  4. 수익을 창출할 수 있는 일

이 네 가지가 교차하는 지점에서 그는 ‘교육’을 발견했습니다. 버클리에서 2,000명이 듣는 대형 강의의 튜터(Tutor)로 활동하며 겪었던 고통, 즉 수천 장의 과제를 기계적으로 채점하며 정작 학생들에게 필요한 피드백은 주지 못했던 그 문제에 집중하기로 한 것입니다.

4. 펜시브의 탄생: AI로 교수자의 시간을 되찾아주다

펜시브(Pensive)는 AI를 활용해 채점 시간을 60%에서 최대 90%까지 단축하면서도 정확도는 유지하는 플랫폼입니다. 윤 대표는 이 아이디어를 검증하기 위해 실제 제품을 만들기 전, 피그마(Figma)로 디자인한 랜딩 페이지(Landing Page) 모델만 들고 콜롬비아 대학교의 교수를 찾아갔습니다.

교수의 반응은 폭발적이었습니다. “AI 튜터도 좋지만, 나에게 지금 당장 필요한 건 AI 채점 도구다. 당장 써볼 수 있느냐”는 확답을 받은 것입니다. 이것이 펜시브의 ‘아하 모먼트(Aha Moment)‘였습니다.

윤 대표는 특유의 **‘리소스풀함(Resourcefulness, 자원 활용 능력)‘**을 발휘했습니다. 무작정 전국의 대학 교수들에게 콜드 메일을 보냈고, 미팅이 잡히면 곧바로 비행기를 타고 날아가 대면 인터뷰를 진행했습니다. 이러한 저돌적인 실행력 덕분에 펜시브는 보수적인 교육 시장에서 이례적인 속도로 성장하며 초기 10개 대학의 계약을 따낼 수 있었습니다.

5. 인류를 강화하는 AI, 그리고 인간의 역할

윤 대표의 비전은 단순한 채점 자동화에 머물지 않습니다. 2026년까지 미국 전역의 수백 개 대학에 솔루션을 공급하고, 궁극적으로는 학생들에게 1:1 AI 튜터를 제공하는 ‘AI 네이티브 교육 기관’을 꿈꿉니다.

그는 AI가 인간보다 똑똑해지는 시대에 우리가 가져야 할 태도에 대해 깊은 통찰을 제시합니다.

“단순히 인간의 잡무를 자동화하는 것을 넘어, AI로 인류의 역량을 강화(Empower)해야 합니다. 그렇지 않으면 우리는 목적을 잃게 될 것입니다.”

그가 생각하는 미래 인류의 핵심 역량은 **‘불확실한 환경에서 올바른 의사결정을 내리는 능력’**입니다. 정답이 없고 맥락이 복잡한 상황에서 스스로 내리는 결정만이 그 사람만의 고유한 길을 만듭니다. 수많은 선택이 쌓여 만들어진 그 길은 AI조차 복제할 수 없는 인간만의 가치가 된다는 것이 그의 철학입니다.

결론: 더 대담하게, 더 어렵게

윤 대표는 젊고 똑똑한 창업가들에게 “더 대담하고 어려운 문제에 도전하라”고 촉구합니다. 남들이 가지 않는 길, 해결하기 어려운 문제에 과감하게 베팅할 때 비로소 예상치 못한 거대한 보상이 돌아온다는 것입니다.

암이라는 시련을 성장의 동력으로 삼고, 8번의 실패를 통해 자신의 소명을 찾아낸 윤 대표. 펜시브가 그려나갈 교육의 미래는 단순히 기술의 진보를 넘어, 인간이 인간다운 일에 더 집중할 수 있게 만드는 따뜻한 혁명을 꿈꾸고 있습니다.


“I gave OpenClaw one job: go viral (it worked?)”Greg Isenberg 기반 기사 원본 영상 보기

잠든 사이에도 돈을 버는 AI 마케팅 머신: ‘OpenClaw’와 ‘Larry’가 만드는 자동화 수익의 미래

영국의 어느 평범한 소도시에 거주하는 한 남자가 자신의 AI 에이전트를 ‘마케팅 머신’으로 진화시켜 화제가 되고 있습니다. 그는 **오픈클로(OpenClaw)**라는 오픈소스 AI 플랫폼을 활용해 디지털 직원을 고용했고, 이 직원은 스스로 틱톡(TikTok) 영상을 제작해 수백만 회의 조회수를 기록하며 매일 자동으로 수익을 창출하고 있습니다.

스타트업 전문가 그렉 아이젠버그(Greg Isenberg)의 팟캐스트에 출연한 올리버 헨리(Oliver Henry)는 자신이 개발한 AI 에이전트 ‘래리(Larry)‘를 통해 어떻게 마케팅을 자동화하고, 1인 기업의 한계를 극복했는지 그 비결을 상세히 공개했습니다.


1. 마케팅이 싫어서 만든 ‘디지털 직원’ 래리

올리버는 본업이 따로 있는 평범한 직장인이자 개발자입니다. 그는 여자친구가 인테리어 디자인을 위해 챗GPT를 사용하며 어려움을 겪는 것을 보고, 방 구조를 유지하면서 인테리어만 바꿔주는 **‘스너글리(Snuggly)‘**라는 앱을 개발했습니다.

앱은 훌륭했지만 문제는 ‘마케팅’이었습니다. 올리버는 “마케팅을 정말 싫어했다”고 고백합니다. 직접 얼굴을 노출하며 영상을 찍고 편집하는 과정은 너무나 많은 에너지를 소모했습니다. 이때 그는 결심했습니다. “내가 마케팅을 하는 게 아니라, 마케팅을 하는 AI를 만들자.”

그렇게 탄생한 것이 바로 오픈클로 기반의 AI 에이전트 **‘래리(Larry)‘**입니다. 올리버는 래리에게 딱 한 가지 임무를 주었습니다. “나 대신 우리 앱을 홍보하고 바이럴(Viral)을 일으켜라.”

2. ‘래리 루프(Larry Loop)’: 데이터가 콘텐츠를 만드는 무한 반복 시스템

래리의 작동 원리는 단순한 콘텐츠 제작을 넘어선 ‘지능형 피드백 루프’에 있습니다. 올리버는 이를 **‘래리 루프(Larry Loop)‘**라고 부릅니다.

  1. 리서치: 래리는 브레이브(Brave) 브라우저를 통해 해당 니즈(인테리어)에서 어떤 슬라이드쇼 콘텐츠가 유행하는지 분석합니다.
  2. 제작: 분석된 데이터를 바탕으로 이미지 생성 모델(DALL-E 3 등)을 사용해 슬라이드 이미지를 만들고, 사람들의 시선을 끌 수 있는 ‘훅(Hook)’ 문구를 작성합니다.
  3. 분석: 틱톡 분석(Analytics) API를 통해 어떤 영상이 조회수가 높고, 어떤 영상이 실제 앱 다운로드(Conversion)로 이어졌는지 확인합니다.
  4. 최적화: 분석 결과를 바탕으로 다음 콘텐츠의 문구와 이미지를 스스로 수정합니다.

이 과정에서 올리버가 하는 일은 단 5분, 래리가 완성해 둔 ‘임시 저장(Draft)’ 영상을 확인하고 음악을 추가해 업로드 버튼을 누르는 것뿐입니다.

3. 역설의 미학: AI의 ‘실수’가 바이럴을 만든다

올리버는 흥미로운 사례 하나를 소개했습니다. 어느 날 밤, 래리가 제작한 영상을 보고 올리버는 화가 났습니다. “주방 인테리어 영상인데 오븐이 사라져 있고, 텍스트 위치도 엉망이었죠. 래리에게 ‘이건 쓰레기야!’라고 쏘아붙이고 그냥 잠들었습니다.”

하지만 다음 날 아침, 반전이 일어났습니다. 그 영상이 수십만 회의 조회수를 기록하며 대박이 난 것입니다. 이유는 역설적이게도 ‘AI의 실수’ 때문이었습니다. 수많은 사용자가 댓글로 “오븐은 어디 갔냐?”, “여기서 요리는 어떻게 하냐?”라며 지적을 쏟아냈고, 이 활발한 댓글 소통이 틱톡 알고리즘을 자극해 영상을 폭발적으로 확산시킨 것입니다.

올리버는 여기서 큰 교훈을 얻었습니다. “완벽함보다 중요한 것은 반응(Engagement)이다.” 이후 래리는 이러한 ‘의도된 혹은 의도치 않은 실수’까지 데이터로 학습하여 더욱 강력한 바이럴 콘텐츠를 생산하게 되었습니다.

4. 오픈클로(OpenClaw)와 ‘스킬(Skills)’: 매트릭스의 네오처럼 학습하기

올리버가 사용한 **오픈클로(OpenClaw)**는 클라우드가 아닌 사용자의 컴퓨터(로컬)에서 직접 돌아가는 AI 에이전트 서버입니다. 이는 보안과 개인정보 보호 측면에서 강력한 장점을 가집니다.

특히 그는 **‘스킬(Skills)‘**이라는 개념을 강조합니다. 영화 *<매트릭스>*에서 주인공 네오가 프로그램을 뇌에 입력받아 순식간에 쿵푸를 배우듯, AI 에이전트에게 특정 능력이 담긴 ‘스킬 파일’을 업로드하면 즉시 마케팅 전문가나 데이터 분석가로 변신할 수 있습니다.

올리버는 자신의 노하우를 담은 **‘래리 마케팅 스킬(Larry Marketing Skill)‘**을 무료로 공개하며, 누구나 자신만의 디지털 직원을 고용할 수 있는 시대를 열고 있습니다.

5. 1인 기업과 SaaS의 진화: “도구가 아닌 직원을 고용하라”

과거에는 소프트웨어를 ‘도구(Tool)‘로 사용했습니다. 소셜 미디어 관리 도구에 들어가 사람이 직접 예약을 걸고 분석해야 했습니다. 하지만 올리버가 보여준 미래는 다릅니다. 이제는 **‘AI 직원(AI Employee)‘**에게 목표만 설정해 주면 됩니다.

  • 비용 절감: 호스팅 비용, 도메인 비용, 마케팅 인건비가 거의 제로(0)에 수렴합니다.
  • 무한 확장: 한 명의 개발자가 여러 개의 앱을 동시에 운영하며, 각각의 앱에 전담 AI 마케터를 배정할 수 있습니다.
  • 반복의 힘: 사람이 지칠 법한 수백 번의 테스트와 수정을 AI는 지치지 않고 수행하며 최적의 정답을 찾아냅니다.

결론: 일단 시작하고, AI가 학습할 시간을 주어라

올리버는 이제 막 AI 자동화를 시작하려는 사람들에게 이렇게 조언합니다. “첫 번째 게시물이 조회수 700회에 그쳤다고 포기하지 마세요. 그것은 실패가 아니라 AI가 학습하기 위한 첫 번째 데이터일 뿐입니다.”

그는 현재 본업을 유지하면서도 AI 에이전트 래리의 도움으로 매달 수백 달러의 자동 수익(MRR)을 올리고 있으며, 이 수치는 데이터가 쌓일수록 기하급수적으로 늘어나고 있습니다.

기술은 이미 우리 곁에 와 있습니다. 이제 질문은 “이게 가능할까?”가 아니라, **“당신은 어떤 임무를 맡긴 AI 직원을 고용할 것인가?”**가 되어야 할 것입니다.


주요 용어 정리

  • 오픈클로(OpenClaw): 개인용 컴퓨터에서 실행되는 오픈소스 AI 에이전트 플랫폼.
  • 에이전트(Agent): 사용자의 목표를 이해하고 스스로 계획을 세워 실행하는 지능형 AI 소프트웨어.
  • 전환율(Conversion Rate): 광고나 콘텐츠를 본 사람이 실제 구매나 다운로드로 이어지는 비율.
  • SaaS(Software as a Service): 클라우드를 통해 제공되는 소프트웨어 서비스.
  • MRR(Monthly Recurring Revenue): 월간 반복 매출. 구독형 서비스의 핵심 지표.

“5 steps to generate consistent brand images with Midjourney”How I AI 기반 기사 원본 영상 보기

[심층 분석] “프롬프트 노가다는 이제 그만” 미드저니로 독보적인 브랜드 비주얼을 구축하는 5단계 가이드

오늘날 AI로 이미지를 만드는 것은 어렵지 않습니다. 하지만 브랜드가 요구하는 ‘일관성’과 ‘고유한 미학’을 갖춘 이미지 수백 장을 뽑아내는 것은 완전히 다른 차원의 문제입니다. 많은 사용자가 원하는 결과가 나올 때까지 하루 종일 프롬프트를 수정하며 고군분투하곤 합니다.

AI 크리에이티브 디렉터 제이미 개넌(Jamie Ganon)은 “핵심은 정교하고 잘 관리된 프로세스에 있다”고 강조합니다. 그녀가 수만 시간을 투자해 정립한, ‘프롬프트에 매몰되지 않고도’ 일관된 브랜드 자산을 생성하는 5단계 워크플로우를 심층 분석했습니다.


1단계: 시각적 언어의 구축, ‘무드보드’에서 시작하라

대부분의 사람은 미드저니(Midjourney)를 켤 때 텍스트 입력창부터 바라봅니다. 하지만 제이미는 핀터레스트(Pinterest)나 코스모스(Cosmos)에서 **무드보드(Mood Board)**를 만드는 것이 첫 번째 단계라고 말합니다.

왜 무드보드인가? 거대언어모델(LLM)에게 “그림 한 장은 천 마디 말보다 가치가 있습니다.” ‘핑크색’, ‘세련된’, ‘사이버틱한’ 같은 추상적인 단어는 사람마다, 그리고 AI 모델마다 해석이 다릅니다. 하지만 이미지는 AI에게 조명, 질감, 구도, 색감을 직관적으로 전달하는 가장 정확한 ‘시각적 언어’가 됩니다.

  • Tip: 서로 상충하는 요소들을 병치(Juxtaposition)해보세요. 예를 들어 ‘피어싱을 한 오렌지’나 ‘컴퓨터 앞에 앉은 형광색 강아지’ 같은 이미지는 브랜드에 독특한 개성을 부여합니다.

2단계: 스타일 참조(SRF)를 활용한 ‘분위기’ 고정

무드보드가 완성되었다면 이제 미드저니의 스타일 참조(Style References, SRF) 기능을 활용할 차례입니다. 단순히 프롬프트에 이미지 URL을 넣는 것을 넘어, 여러 장의 이미지에서 공통된 스타일을 추출해 적용하는 과정입니다.

제이미는 이 단계를 ‘정보 수집 단계’라고 부릅니다. “이 단계에서는 결과물에 너무 집착하지 마세요. 내가 만든 무드보드와 프롬프트가 AI에게 어떻게 전달되는지 빠르게 테스트하는 것이 목적입니다.”

만약 결과물이 너무 초록색으로 치우친다면? 초록색이 강한 참조 이미지를 제거하면 됩니다. 텍스트로 “초록색 빼줘”라고 명령하는 것보다 참조 이미지 구성을 바꾸는 것이 훨씬 빠르고 정확합니다.

3단계: 개인화(Personalization) 코드로 ‘취향’ 주입하기

미드저니의 최신 기능 중 하나인 **개인화(—p)**는 사용자의 미적 취향을 AI에게 학습시키는 과정입니다. 미드저니 웹사이트에서 진행되는 ‘이것 혹은 저것(This or That)’ 투표를 통해, AI는 사용자가 선호하는 질감, 색감, 조명 스타일을 파악합니다.

제이미는 이를 통해 자신만의 **‘2025년 에스테틱 코드’**를 만들어 사용합니다.

  • 개인화의 힘: 스타일 참조(SRF)와 개인화 코드를 결합하면, 단순한 ‘여성 모델’ 프롬프트만으로도 브랜드 특유의 피부 질감, 메이크업 스타일, 조명 값이 적용된 고차원적인 결과물을 얻을 수 있습니다.
  • 주의사항: 투표 과정에서 너무 많이 건너뛰기(Skip)보다는, 실제로 자신이 생성하고 싶은 스타일의 이미지에만 투표하여 모델의 ‘편향성’을 정교하게 다듬어야 합니다.

4단계: ‘영리한 게으름’을 위한 프롬프트 치트키

제이미의 워크플로우에서 가장 인상적인 부분은 프롬프트가 매우 짧다는 점입니다. 그녀는 복잡한 JSON 형식이나 수백 단어의 설명을 지양합니다. 대신 전문가들이 사용하는 **‘치트키 단어’**를 활용합니다.

  1. 매거진 및 아티스트 이름 활용: ‘Dazed editorial’, ‘Vogue’, ‘High Fashion’ 같은 단어는 AI에게 수천 가지의 세부 설정(대비, 하이라이트, 구도 등)을 한 번에 지시하는 것과 같습니다.
  2. 카메라 모델 명시: ‘Sony RX100’, ‘DSLR’, ‘35mm Film’ 등 특정 카메라 모델을 언급하면 프롬프트로 설명하기 힘든 특유의 질감과 조리개 값을 즉시 구현할 수 있습니다.
  3. 추상적 단어의 힘: ‘Luxury’라는 단어 한 마디면 AI는 ‘높은 층고, 통창, 뉴욕 스카이라인, 고급 가구’를 알아서 배치합니다. 이를 일일이 설명하는 것은 시간 낭비입니다.

5단계: 나노 바나나(Nano Banana)를 통한 정밀 수정 및 업스케일링

미드저니에서 완벽한 분위기를 잡았더라도 손가락이 어색하거나 로고가 이상하게 생성될 수 있습니다. 이때 제이미는 나노 바나나(Nano Banana, 현재는 Flux 기반의 다양한 편집 도구로 진화) 같은 도구를 사용합니다.

  • AI 포토샵: “컴퓨터를 최신형 맥북 프로로 바꿔줘”, “표정을 좀 더 화난 것처럼 수정해줘” 같은 명령을 통해 이미지를 부분적으로 수정(Inpainting)합니다.
  • 고해상도 작업: 소셜 미디어나 광고에 사용하기 위해 저해상도 이미지를 4,000픽셀 이상의 고해상도로 업스케일링하며 디테일을 보강합니다.

결론: 서비스 모델의 패러다임 변화

제이미 개넌의 방식은 기존 크리에이티브 디렉터의 역할을 재정의합니다. 과거에는 완성된 사진 몇 장을 전달하고 끝냈다면, 이제는 ‘브랜드의 시각적 알고리즘’ 자체를 설계하여 고객에게 전달합니다.

고객에게 특정 이미지뿐만 아니라, 그 이미지를 계속해서 뽑아낼 수 있는 **‘스타일 코드와 참조 이미지 패키지’**를 제공하는 것입니다. 이는 클라이언트와 크리에이티브 디렉터 사이에 훨씬 더 긍정적이고 지속적인 협력 관계를 만들어냅니다.

“AI는 당신의 확장판입니다.” 제이미의 말처럼, 기술적인 프롬프트에 매몰되기보다 자신만의 미적 기준을 세우고 이를 AI에게 전달하는 프로세스를 구축하는 것. 그것이 바로 AI 시대에 독보적인 브랜드를 만드는 비결입니다.


“The Top 100 Consumer AI Apps | The a16z Show”a16z 기반 기사 원본 영상 보기

[심층 분석] 챗GPT는 시작일 뿐이다: a16z가 공개한 ‘글로벌 소비자 AI 앱 톱 100’ 리포트의 핵심 통찰

실리콘밸리의 대표적인 벤처캐피털(VC) 안드레센 호로위츠(a16z)가 최근 ‘소비자 AI 앱 톱 100(Top 100 Consumer AI Apps)’ 리포트의 여섯 번째 에디션을 발표했습니다. 지난 3년간 AI 시장은 폭발적으로 성장했지만, 전문가들은 입을 모아 말합니다. **“우리는 아직 초창기(Early stage)에 머물러 있다”**고 말이죠.

전 세계 인구의 단 10%만이 챗GPT(ChatGPT)를 매주 사용하고 있는 지금, 수면 아래에서는 거대한 지각 변동이 일어나고 있습니다. 이번 리포트를 통해 드러난 AI 산업의 현재와 미래의 향방을 5가지 핵심 키워드로 정리했습니다.


1. 거대 모델의 삼국지: 챗GPT의 독주와 추격자들의 전략

현재 소비자 AI 시장은 챗GPT가 압도적인 우위를 점하고 있습니다. 웹에서는 구글의 제미나이(Gemini)보다 2.7배, 모바일에서는 2.5배 더 큰 규모를 자랑합니다. 앤스로픽(Anthropic)의 클로드(Claude)와 비교하면 격차는 더 벌어집니다. 웹에서는 30배, 모바일에서는 무려 80배나 더 많은 사용자를 보유하고 있죠.

하지만 수치보다 중요한 것은 **각 모델의 ‘전문화된 생존 전략’**입니다.

  • 챗GPT (OpenAI): ‘모두를 위한 AI’를 지향합니다. 구글과 유사한 전략으로, 여행 예약이나 쇼핑 등 광범위한 소비자 접점을 확보해 광고나 거래 수수료로 수익을 창출하려 합니다.
  • 클로드 (Anthropic): ‘프로슈머(Prosumer, 전문 사용자)‘에 집중합니다. 데이터 분석, 과학 연구, 금융 지표 해석 등 고부가가치 작업에 특화되어 있으며, 유료 구독 모델을 핵심 수익원으로 삼습니다.
  • 제미나이 (Google): 구글 생태계(Gmail, 문서 도구 등)와의 결합이 강점입니다. 특히 최근에는 노트북LM(NotebookLM)과 같은 새로운 크리에이티브 도구를 통해 실험적인 시도를 이어가고 있습니다.

2. ‘락인(Lock-in)’ 효과의 핵심: 기억(Memory)과 맥락(Context)

과거에는 AI와 나눈 대화가 일회성에 그쳤다면, 이제는 **‘맥락의 축적’**이 강력한 경쟁 우위가 되고 있습니다.

a16z의 파트너 올리비아 무어(Olivia Moore)는 ‘챗GPT로 로그인(Login with ChatGPT)’ 기능에 주목합니다. 사용자가 자신의 취향, 업무 스타일, 과거 기록(Memory)을 특정 AI에 저장해두면, 다른 서비스로 옮겨가는 비용이 매우 커집니다.

마치 우리가 애플이나 구글 생태계에 묶여 있는 것처럼, AI가 나의 ‘디지털 정체성’을 학습하게 되면 그 편리함 때문에라도 해당 플랫폼을 떠날 수 없게 되는 ‘복합적인 우위(Compounding Advantage)‘가 형성되는 것입니다.

3. 글로벌 AI 지도: 뜻밖의 ‘AI 열강’은 어디인가?

국가별 인구 대비 AI 사용량을 분석한 결과는 흥미롭습니다. AI 기술의 발원지인 미국은 의외로 20위에 그쳤습니다.

  • 1위 싱가포르, 4위 한국: 이들 국가는 기술 수용도가 매우 높고 ‘기술 낙관주의(Tech Optimism)‘가 강한 문화를 가지고 있습니다.
  • 러시아와 중국: 서방의 제재와 검열 등의 이유로 독자적인 AI 생태계를 구축하고 있습니다. 중국의 딥시크(DeepSeek), 러시아의 기가챗(GigaChat) 등이 자국 시장에서 엄청난 사용량을 기록 중입니다.

특히 미국에서 AI에 대한 신뢰도가 32%에 불과한 반면, 상위권 국가들은 50~70%에 달한다는 점은 시사하는 바가 큽니다. 문화적 수용성이 기술 보급의 속도를 결정짓고 있는 셈입니다.

4. 창작 도구의 진화: ‘단순 생성’에서 ‘고유한 워크플로우’로

초기 AI 붐을 일으켰던 이미지 생성 도구들은 이제 성숙기에 접어들었습니다. 단순히 “그림을 그려줘”라고 요청하는 수준을 넘어, 이제는 특정 영역에서 강력한 팬덤을 형성한 도구들이 살아남고 있습니다.

  • 음악과 목소리: 수노(Suno)와 일레븐랩스(11 Labs)는 각각 음악 생성과 음성 합성 분야에서 독보적인 위치를 차지하며 톱 20 안에 진입했습니다.
  • 비디오의 도전: 오픈AI의 소라(Sora)는 출시 직후 폭발적인 관심을 끌었지만, 틱톡이나 인스타그램의 인간 제작 콘텐츠와 경쟁해야 하는 숙제를 안고 있습니다. 단순히 신기한 영상을 만드는 것을 넘어, 사용자가 직접 참여하고 즐길 수 있는 ‘사회적 가치’를 증명해야 합니다.

5. 미래의 주인공: 에이전트(Agent)와 음성(Voice)

앞으로의 6개월에서 1년 사이, 우리 삶을 가장 크게 바꿀 키워드는 **‘에이전트(Agent)‘**입니다.

에이전트란 사용자의 명령을 듣고 스스로 계획을 세워 실행까지 완료하는 AI를 말합니다. 리포트에서는 메타(Meta)에 인수된 매너스(Manus)와 같은 자율형 에이전트의 등장을 강조합니다. 이제 AI는 단순히 질문에 답하는 비서가 아니라, 이메일을 보내고, 슬라이드를 만들고, 여행 일정을 실제로 예약하는 ‘실행자’가 될 것입니다.

또한, **‘음성(Voice)‘**은 가장 밀도 높은 정보 전달 수단으로 다시 부상하고 있습니다. 회의 내용을 기록하고 요약하는 수준을 넘어, AI와 자연스럽게 대화하며 업무를 처리하는 ‘보이스 퍼스트(Voice-first)’ 시대가 머지않았습니다.


기사를 마치며: “온보딩이 없는 세상”

올리비아 무어는 미래의 AI 경험을 이렇게 정의합니다. “앞으로 2년 뒤, 우리가 어떤 앱을 처음 켰을 때 그 앱이 나를 즉시 알아보지 못한다면, 우리는 그 서비스가 고장 났다고 느낄 것입니다.”

학습된 기억(Memory)을 바탕으로 나보다 나를 더 잘 아는 AI가 모든 서비스의 기본이 되는 세상. 기술적 변화보다 더 무서운 것은 우리의 ‘기대치’가 변하고 있다는 점입니다. 13살 소녀들의 놀이 문화부터 기업의 의사결정 방식까지, AI는 이제 선택이 아닌 공기와 같은 존재로 스며들고 있습니다.

우리는 지금, AI가 단순한 도구를 넘어 인간의 능력을 확장하는 ‘에이전트의 시대’로 넘어가는 문턱에 서 있습니다.