YouTube Digest
Listen to this issue
English
Based on “Robert Herjavec: The hidden reason smart people stop growing | Big Think+” from Big Think Watch the original video
Beyond the Guru: Robert Herjavec’s Unconventional Truths About Mentorship and Lifelong Growth
For anyone aspiring to climb the career ladder or launch a successful venture, the quest for a mentor often feels like a non-negotiable step. We envision a sage figure, perhaps a titan of industry or a celebrity entrepreneur, who will take us under their wing and guide us to greatness. But according to Robert Herjavec, the renowned investor from Shark Tank, this common perception of mentorship is not only flawed but might be precisely why many smart people unknowingly halt their own growth.
Herjavec, frequently bombarded with requests like, “Robert, can you be my mentor?”, offers a resounding “No.” Not out of malice, but out of a deep understanding that the traditional view of mentorship is riddled with myths. His insights challenge us to rethink how we seek guidance and, more importantly, how we cultivate a mindset of continuous learning.
Myth #1: You Need a Celebrity Mentor
The biggest misconception, Herjavec argues, is the belief that a mentor must be a household name or a “guy from Shark Tank.” This star-struck approach misses the fundamental purpose of mentorship. “The entire role of a mentor is to teach you something that helps you at this point in your career,” he explains. You don’t need a universally recognized leader; you need someone who possesses the specific knowledge or experience relevant to your current challenges. Learning, Herjavec emphasizes, can come from anyone, anywhere.
Myth #2: Mentorship is a Lifelong, Constant Relationship
Life is dynamic, and so too should be our sources of wisdom. Herjavec debunks the idea that mentorship is a static, enduring relationship. The advice a 23-year-old starting their career needs is vastly different from what a 45-year-old senior vice president requires. “Your mentors, like your life, will change over time,” he states. This fluidity means we should embrace different mentors for different phases and needs, rather than searching for a single, all-encompassing guru.
Myth #3: Mentorship Requires a Formal Contract
Forget invisible contracts or formal agreements. Some of Herjavec’s most profound mentors had no idea they were guiding him. He cites Warren Avis, the founder of Avis Rent A Car, as one such figure. “If he was still alive today and you asked him, ‘Were you a mentor to Robert?’ He would say, ‘I have no idea what you’re talking about.’”
Mentorship, Herjavec reveals, is often a product of observation, learning through “osmosis,” and simply taking it all in. This informal approach is more accessible than ever before. “You can just learn, read, engage in many, many different ways, especially today with YouTube,” he notes, pointing out the vast trove of knowledge available, including his own thousands of hours of video content.
The Hidden Growth Killer: Ego
The true “hidden reason smart people stop growing,” according to Herjavec, isn’t a lack of opportunity or talent, but something far more insidious: ego. “Ego has killed more careers than anything else in the world,” he asserts. Regardless of how brilliant your ideas or how impressive your achievements, a dose of humility is essential.
“The first key to it is being open to the idea that you may not have all the answers,” Herjavec advises. Confidence is a virtue, but arrogance is a barrier. If you believe the world has nothing left to teach you, you effectively shut yourself off from learning. Mentorship, at its core, is about being a constant learner. Herjavec illustrates this with a surprising anecdote: “One of the greatest lessons I ever learned was from a janitor, who had an incredible work discipline and a really fanatical drive for perfection and constant improvement.” Wisdom, he reminds us, is truly all around us.
The “Red Car Theory” of Opportunity
To truly embrace learning and find mentors, one must first be open to seeing them. Herjavec shares a powerful analogy, the “red car theory,” taught to him by one of his own mentors. If asked how many red cars you saw on the way to work yesterday, you likely wouldn’t know. But if instructed to count them tomorrow, you would.
“Opportunity is like that, and mentorship is like that,” Herjavec explains. When you are consciously open to it, actively looking for it, opportunities for learning and guidance become visible everywhere. The world doesn’t change, but your perception does. So, be open, look for it, and then, crucially, “never be afraid to ask.” Herjavec’s general experience is that “98 percent of the people I’ve encountered are super nice. And if you ask for help, most of the time, people will help you.”
Mastering the Mentor-Mentee Dynamic
While informal learning is vital, some direct engagement is inevitable. For those seeking to maximize a mentor-mentee relationship, Herjavec offers practical, often surprising, advice:
- Respect Their Time: This is paramount. Herjavec, a “very, very busy guy,” makes it clear: if he has free time, he wants to spend it with his kids, “probably don’t want to hang out with you.” Mentors are often successful precisely because they manage their time meticulously. Understand this and make every interaction count.
- Know How to Communicate: Different people prefer different communication channels. Herjavec uses Mark Cuban as an example: “Mark Cuban doesn’t talk on the phone ever. He hates talking on the phone. But if I emailed him at 10:30 tonight, he’d get back to me within three minutes.” Herjavec himself dislikes phone calls because they disrupt his day, preferring email for its asynchronous nature. It’s your responsibility to learn your mentor’s preferred method.
- Understand Boundaries: The idea that “we’re one big family” at work is often a myth. “My company was not my family. It was a business. My family was my family,” Herjavec states. He advises understanding where a mentor draws the line between professional and personal life. Some are highly social; others, like Herjavec, prefer not to be called on weekends. However, he’s an early riser, often up at 5 AM, and suggests that reaching out between 5 and 9 AM might be a “great time for me.” The onus is on the mentee to understand the mentor’s rhythm and preferences to make the most effective connection.
In essence, Herjavec’s philosophy on mentorship isn’t about finding a singular, famous individual to light your path. It’s about cultivating an open, humble, and strategic mindset. It’s about recognizing that every interaction, every observation, and every conscious effort to learn can contribute to your growth. By shedding the myths and embracing these unconventional truths, we can unlock our full potential and ensure that our learning journey never truly stops.
Based on “How Linear Turned AI Agents Into First-class Users” from Every Watch the original video
The Quiet Revolution: How Linear Turned AI Agents into First-Class Citizens
In a world clamoring for instant AI integration, where every company scrambled to affix a chatbot to its product, Linear, the esteemed project management tool, took a different path. While the market declared “SaaS is dead” and frantically chased token counts, Linear’s co-founder and CEO, Ki, cultivated a patient, thoughtful approach that has now positioned his company as a quiet leader in the age of AI agents.
Linear, a name synonymous with taste, craft, and an almost whispered reverence among its early adopters, began its journey pre-AI. Long before large language models became household names, Linear earned a reputation for quality and a deliberate pace of development. The company intentionally kept its doors somewhat closed, avoiding the hyper-growth pressure to instead build something enduring. This foundational philosophy, as Ki explains, would become the very bedrock of their successful transition into the AI era.
“We just want Linear to be the best product in this category and like helping companies move work forward,” Ki states, emphasizing that the core mission hasn’t changed, only been amplified by AI. “Our goal was always that Linear can take more of the burden of running product teams… and let the product teams and individuals actually build the things. And now, it’s also like they build it with the AI, or the AI builds it.”
Understanding Before Acting: Navigating the AI Wave
When the first wave of AI hype hit, many companies rushed to integrate chatbots, often without a clear use case. Linear, with its design-first philosophy, resisted this urge. “I come from a design background,” Ki explains. “A lot of times my way of approaching things is first I’m trying to understand them… what happens in the tech world a lot of times is people don’t try to understand things. They often like jump into the ‘oh I can do this so I’ll do it now’ but like did you think should you do it or does it actually help you?”
This critical self-reflection led Linear to experiment internally with chatbots, only to find them “not really that useful.” The crucial question became: “What is the workflow where you would actually need this or use this?”
Instead of chasing trends, Linear spent years understanding potential AI workflows. This led to a prescient insight: “I don’t think there’s going to be one agent but everyone will have many agents, and companies will build their own agents.” Recognizing this multi-agent future, Linear released an open agent platform with robust documentation, allowing external coding agents – including OpenAI’s Codex and homegrown solutions from customers like Coinbase – to integrate seamlessly. Linear positioned itself not as the sole agent, but as “a system for guiding the agents and building this context.”
Beyond Vanity Metrics: The True Measure of AI Success
The host noted a common market sentiment: “The stock market thinks that SaaS is dead.” This narrative often pushes companies to demonstrate AI integration at any cost, leading to what Ki calls “vanity metrics.”
“The biggest like vanity metric is like how much of your code is agent written or how many PRs are you merging,” Ki observes. “I think like that’s like not the right metric. It’s measures output, but what does that output do? Does it actually generate value? Is it improving the product?”
Linear’s approach is far more nuanced. While they encourage their team – now about 120 people, half on product – to use AI tools, they emphasize classic metrics: profit, revenue, and user love. Crucially, they focus on quality. Ki challenges the industry: “Why do you even have bugs in your product? Like you should be like there’s no excuse for it anymore.” Linear internally enforces a “zero bugs” policy, with a one-week SLA for fixes. Here, AI agents play a vital role, performing the first pass on bug fixes, then alerting engineers for review and finalization. It’s a testament to how AI can elevate quality, not just speed.
The “Slow to Decide, Fast to Execute” Philosophy
Linear’s adoption of AI tools internally reflects a core philosophy: “I don’t want the problem finding to be fast. Like you should take the time to find the right problem and like the right approach for the problem and then once you decide that then you can go faster on it.”
This means AI is not used to rush the initial conceptualization, but to accelerate execution once a clear direction is set.
- Product Development: Linear’s internal “Linear Way skill” (fed with internal docs and blog posts) acts as an AI product teammate. When a feature request like “multiple assignees per issue” arises, the AI synthesizes customer feedback, identifies underlying needs, and offers product direction recommendations. This helps Ki and his team quickly understand complex problems without extensive manual research.
- Design: While Ki personally prefers the “slowness” of manual design for exploratory conceptual work (“Every time you draw something you have to kind of like check on yourself: why am I drawing it this way?”), the broader design team leverages AI for rapid prototyping and testing. This allows them to iterate quickly on established concepts.
- Engineering: AI significantly shortens development loops. Conversations in Slack can be instantly turned into actionable issues by the Linear Agent, bypassing meetings and manual assignment. Bug fixes are faster, and small tasks can be automated.
This strategic application of AI ensures that speed serves thoughtful decision-making, rather than replacing it.
From Hub to Core: The Strategic Evolution of Linear Agent
Initially, Linear thrived as the central hub for external AI agents. The host even mused, “This is the perfect business for this era because it’s still SaaS… you’re the one who has the sort of sticky interface… but you don’t have to pay for any of the actual tokens.”
However, Linear recognized a limitation: when they weren’t in control of the agents, they couldn’t fully optimize the end-to-end workflow or implement their own innovative ideas. This led to a significant strategic shift: the development of Linear Agent.
“One of the reasons we are doing this coding agent is like we actually think we see this like a lot more smoother end-to-end workflow,” Ki explains. The Linear Agent, with its inherent context of an organization’s work, can understand and act more naturally than external agents that require explicit prompting. It can turn a task into an issue, start writing code, show diffs, and facilitate reviews, all within Linear’s environment. This aims to create a “product context or the product memory platform” rather than a generic agent platform.
This move does introduce a new business model consideration: token costs. While basic Q&A functionality will be included, coding agents will likely operate on a usage-based billing model. It’s a calculated risk, prioritizing deeper integration and a superior user experience over maintaining a purely token-cost-free model.
A Glimpse into the Future
Ki offered a preview of Linear Agent in action:
- “Understand Problem” Skill: Using a custom skill trained on Linear’s internal knowledge, the agent can analyze a complex feature request like “multiple workspaces.” It synthesizes customer activities, identifies underlying needs (e.g., unified billing/governance with divisional autonomy), explains the problem’s complexity, and even suggests product directions.
- Coding Agent: A simple request like “make a new dark theme, just black” can be delegated to the coding agent. It creates an issue, spins up a sandbox environment, and starts writing code, with the changes (diffs) visible online for review.
These examples highlight how Linear Agent leverages the platform’s inherent context, allowing for more natural, powerful interactions that streamline complex product development.
The Enduring Value of Human Judgment
Linear’s journey into AI is a masterclass in thoughtful adaptation. It demonstrates that in an era of rapid technological change, core values like patience, quality, and a deep understanding of user needs remain paramount. By moving beyond superficial integrations and focusing on how AI can genuinely enhance human creativity and efficiency, Linear is not just surviving the “AI revolution” but actively shaping its most productive future.
As Ki concludes, the goal isn’t just more output, but meaningful output. “We shouldn’t be tied into the past experience, like the past product we have, but like see like what the future product should be.” In a world where AI can build anything, Linear is ensuring that we first build the right things, with clarity, context, and unwavering commitment to quality.
Based on “23 AI Trends keeping me up at night” from Greg Isenberg Watch the original video
The AI Awakening: 23 Trends Keeping Greg Isenberg Up at Night
The world of artificial intelligence is moving at a dizzying pace, transforming industries and redefining what’s possible. For entrepreneur and investor Greg Isenberg, this isn’t just a technological shift; it’s a profound re-architecture of business, creativity, and daily life – a phenomenon so potent it keeps him awake, churning with both excitement and a healthy dose of apprehension.
In a recent deep dive, Isenberg outlined 23 AI trends that dominate his thoughts, presenting a compelling vision of the opportunities, the looming challenges, and the radical shifts awaiting those ready to build in this new era. It’s a call to action, a warning, and a roadmap for navigating the most asymmetric building window of our time.
The Dawn of Hyper-Speed Entrepreneurship: Your Company, Today.
Forget the months-long process of traditional startup incubation. Isenberg envisions a world where the “one-hour company stack” is not just a dream, but a reality. “You grab an idea, you vibe code something, you build a landing page, you add a Stripe, and you can get first customers,” he exclaims, highlighting the mind-blowing speed at which ideas can now materialize into revenue-generating entities.
The traditional timeline of building a company – months for development, a year for first revenue – is obsolete. In 2026, Isenberg predicts, you could have an idea by 9 AM, a product built by 9:45 AM, first customer by 10 AM, and be iterating by lunch. This isn’t “vibe coded slop,” he asserts, but the result of powerful agent engineering platforms like Claude Code, Google AI Studio, and Code Interpreter, which enable comprehensive, rapid development. The caveat? You still need an audience or distribution, a challenge AI is also poised to help solve.
This acceleration leads to the rise of “ambient businesses” – autonomous entities that run with zero or very low daily human input. Imagine agents monitoring markets, identifying opportunities, executing trades, and handling customer service, allowing founders to check in only every few days. Isenberg believes these autonomous businesses will soon generate seven or eight figures, driven by “the arrow of progress” pushing us towards self-steering, agent-driven enterprises.
The Agent Economy: A New Digital Frontier
The evolution of the digital landscape is clear: from the “App Store Era” (2009-2015) where humans operated apps, to the “API Economy” (2015-2024) where developers wired systems together. Now, Isenberg declares, we are entering the “Agent Economy” (2025-2030). This is a world where AI agents discover and hire other agents on the fly, dissolving fixed tasks and creating entirely new marketplaces.
This vision sparks a fascinating startup idea: “The Glassdoor of AI agents.” Just as platforms like Glassdoor rate human employees and companies, a similar system will be needed to evaluate the reputation and efficacy of AI agents. Isenberg points to a Gartner stat predicting 20% of commerce by 2030 will be agent-to-agent, machine-to-machine. With 31,000 agent skills already on marketplaces (albeit many of them “garbage”), the opportunity to build high-quality skills and agents, from CEO agents to dev agents to sales agents, is massive. He cites open-source technologies like Paperclip, which allow agents to spin up subtasks and shut them down when complete, as a glimpse into this serverless, agent-orchestrated future.
The Vertical AI Gold Rush: Bigger Than SaaS
YC (Y Combinator) predicts over 300 unicorns in vertical AI this decade, a testament to the immense opportunity in specialized niches. Isenberg draws a sharp contrast between “Vertical SaaS” and “Vertical AI.” While Vertical SaaS captures a fraction of IT spend, selling licenses for tools humans operate, Vertical AI taps directly into a company’s labor P&L.
“You’re building basically an agent as a software because you’re doing that thing that companies are hiring human beings to do,” Isenberg explains. This means the total addressable market for Vertical AI is an order of magnitude larger than Vertical SaaS, as it replaces headcount rather than just augmenting it. Outcomes are sold, agents do the work, and the potential for billion-dollar businesses is significantly higher.
Where are these “boring goldmine verticals”? Think industries still bogged down by phone calls, faxes, and outdated processes: insurance (with its 30-year actuary tables), legal, logistics, elder care, government, accounting, and construction. The key, Isenberg advises, is to pick a sub-niche within these boring categories, ideally one with less red tape and competition.
The Shift to Outcome-Based Pricing: Paying for Results
This agent-driven paradigm also heralds a fundamental shift in pricing models. The old “per-seat” licensing model, prevalent in SaaS, is dying. Isenberg attributes the recent stock market corrections for SaaS companies partly to this, as investors fear both reduced seat counts and the ease of “vibe coding” solutions.
The evolution moved from per-seat to usage-based, and now, rapidly towards “outcome-based” or “pay-per-result” pricing. Gartner predicts 40% of enterprise SaaS will shift to outcome-based pricing by 2030. Companies like Zendesk are already embracing this, and 83% of AI-native SaaS have made the switch.
Isenberg believes someone will build a billion-dollar business simply by converting legacy SaaS companies to outcome pricing. But even more compelling, he argues, is the opportunity to build outcome-based startups from scratch. “Why even help them when you could be incubating yourself?” he challenges, emphasizing the first-mover advantage and inherent appeal of paying only for delivered results.
The SaaS Graveyard and the Scarcity Flip
Not all software will survive this AI revolution. Isenberg foresees a “SaaS graveyard” for generic tools:
- Generic CRMs: Agents will do it better.
- Basic analytics dashboards: AI generates insights on demand.
- Template marketplaces: AI generates custom templates instantly.
- Scheduling tools: Agents will handle calendars natively.
- Basic customer support chatbots: Already being replaced by advanced AI.
What will survive? Vertical workflow tools that pivot to agent companies, infrastructure providers, and data moats.
This leads to a profound “scarcity flip.” AI commoditizes generic content, basic design, data entry, and routine analysis. What becomes scarce and premium?
- Creative judgment: The ability to make discerning, insightful decisions.
- Human-made crafts: Authenticity and the touch of a human hand.
- Physical experiences: When digital is infinite, scarcity shifts to real-world presence and interaction.
- Original weird thinking: LLMs aren’t good at being truly unique or eccentric; human idiosyncrasy becomes valuable.
- Proprietary data: Unique datasets that power specialized AI.
Isenberg even envisions luxury brands adopting “No AI Involved” certification labels, akin to “organic” for food, to signify premium, human-made quality. The “premium stack” will range from “human-made” (most premium) to “AI-assisted but human-led” (human taste with AI speed) to “fully AI service” (commodity, racing to zero pricing). This explains his interest in incubating “IRL stuff” – karaoke bars, escape rooms, immersive theater – where human presence and experience are paramount.
Redefining the Founder and Team: The Director and the Ghost Crew
The role of the founder is also evolving. Gone is the sole focus on “founder-market fit.” Isenberg introduces “founder-agent fit”: “Can you orchestrate a fleet of agents towards your goal?” He compares the modern founder to a film director – not holding the camera, acting, or writing the score, but getting the best performances out of their “actors,” who are now machines. The ability to manage and optimize agents for a particular niche becomes an unfair advantage.
This leads to the concept of the “ghost team org chart.” In the future, a company’s “About Us” page might feature a couple of human founders alongside a roster of AI agents: sales agents, content agents, customer support agents. These agents might even have names, personalities, images, and the ability to video chat or send voice notes, blurring the lines between human and machine colleagues. This vision excites Isenberg, suggesting a future with more holding companies owning multiple AI-native agent businesses, run by these ghost teams.
The classic “1000 true fans” concept also gets an AI-era update: it’s now “100 true fans.” With agents drastically cutting costs, 100 customers paying even $50 a month (for a total of $60,000 profit for a single person running with agents) can constitute a real business. This “micro-monopoly math” allows for the incubation of multiple profitable ventures, especially for those adept at building media and distribution.
The Shadows of Progress: Agent Attack Surface
Amidst the opportunities, Isenberg doesn’t shy away from the darker side. The “agent attack surface” is a significant concern. Prompt injections, poison context windows, malicious service providers, agent-to-agent manipulation, and compromised training data pose serious cybersecurity risks. “I would be lying to you if I said this didn’t freak me out, that bad things are going to happen,” he admits, noting that Palo Alto Networks has already documented real-world agent injection attacks.
He contrasts agent injection with traditional phishing. While phishing tricked humans into clicking bad links, targeting email inboxes and relying on human judgment as defense, agent injection tricks AI agents via hidden instructions, targeting context windows and web content. The agent’s autonomy becomes the vulnerability, and the potential for damage, especially when agents have system access and make autonomous decisions, is “far bigger than phishing.”
This necessitates a new form of “digital hygiene”: the “agent permission stack.” Just as we review app access, we’ll need to regularly cleanse and review what our agents can access (files, emails, bank accounts), remember (conversations, personal data), do (send emails, make purchases, modify code), and share (with other agents or third parties). This is a nascent but critical area for new cybersecurity startups.
Seizing the Asymmetric Opportunity: Build Now
Isenberg stresses that we are in an “asymmetric window” for builders. The build cost is effectively zero, niches are wide open, and audiences are underpriced. He warns that this won’t last forever: “I think that there’s 12 months where competition starts catching up. Some of the best niches get claimed. Some of the tools get crowded. I think there’s probably 24 months where the window narrows.”
His advice is clear: “Things are not settling down. This is the new normal.” The opportunity is asymmetric because with just an API key, some prompts, a tweet, and a small niche audience (100-5000), one can create a 24/7 business with 60-95% margins, compounding distribution, and zero or few employees. “This is the most asymmetric time to be building a startup,” he asserts.
Finally, Isenberg advocates for “building in public,” despite warnings about inviting competition. He believes the benefits outweigh the cons, especially if your audience is also your customer. Sharing what you’re building, allowing the community to vote on features, and shipping updates rapidly (in days, not months) fosters trust and compounds distribution. This “flywheel” creates a strong moat, particularly in a world where “forking a business” (copying others quickly) will become common.
In an era defined by rapid change and unprecedented potential, Greg Isenberg’s insights serve as both a thrilling invitation and a sober reminder. The future of AI is not just coming; it’s already here, keeping some awake with its boundless possibilities and pressing challenges. The question is, what will you build with it?
Based on ""We’re Not Writing Code by Hand Anymore. That’s Over.” | Owen Jennings & David Haber - The a16z Show” from a16z Watch the original video
The Code Whisperers: How Block’s Radical AI Overhaul Declared Manual Programming ‘Over’ and Reshaped its Workforce
In the fast-evolving landscape of technology, seismic shifts often happen quietly, then erupt into public view. For Block (formerly Square), the parent company behind Square, Cash App, and Afterpay, that eruption came in the form of a dramatic 40% reduction in force. But this wasn’t just another layoff in a challenging economic climate; it was a strategic, forward-looking decision, explicitly driven by the profound and immediate impact of artificial intelligence.
Owen Jennings, Business Lead at Block, who previously helmed Cash App during its critical scaling period, shared a candid and revealing account of this transformation on the a16z Show. His message was unequivocal: the era of writing code by hand is over, and the very structure of tech companies is being fundamentally rewritten by AI.
The Binary Shift: When Everything Changed
For decades, there was a clear, almost sacred correlation between the number of people at a company and its output. More engineers, designers, and product managers typically meant more features, more products, and more growth. According to Jennings, that correlation “basically broke” in the first week of December.
“What we were seeing,” Jennings explained, “is that one or two engineers, or a designer and an engineer who was on the tools, is able to be 10, 20, 100x more productive.”
This wasn’t a gradual evolution; it was a “binary change.” Block had been investing in “agentic development” – systems where AI agents assist or even autonomously perform development tasks – for two to three years, launching their first agent harness, “Goose,” in early 2024. They saw “meaningful progress” throughout 2024 and 2025 (in the transcript, this appears to be a misstatement, likely referring to an earlier period given the late November/early December “binary change” context).
However, late November and the first week of December brought a sudden, qualitative leap. The emergence of foundational models like Opus 46 and Codex 53 fundamentally shifted the capabilities of these AI tools. Initially, these models were “pretty good at writing code especially for new ventures and kind of like green space.” But almost overnight, they became “incredibly capable working with existing complex code bases.”
This meant that AI could now not only generate new code but also effectively navigate and contribute to the dense, intricate web of legacy systems that characterize large, established software products. “We’re not writing code by hand anymore. That’s over. That’s done,” Jennings declared, underscoring the finality of this paradigm shift within Block.
Block’s Bold Restructuring: A Response to AI, Not Overhiring
The decision to reduce Block’s workforce by over 40% was not a mere response to the “ZIRP (Zero Interest Rate Policy) hangover” or overhiring from the 2021 boom, as some might assume. Jennings was adamant: “If you thought it was like croft and bloat… then like this riff would have accrued to the operational teams… Those were really, really meaningful cuts on the development side.”
The cuts were disproportionately larger on the development side – engineering, product, and design – precisely because these were the areas where AI tools had delivered the most dramatic productivity gains. Sales and account management teams, for instance, saw “fairly de minimis” reductions.
The executive team, led by CEO Jack Dorsey, spent the first quarter of the year grappling with the implications of this new reality. “What does this mean fundamentally? What does this mean in terms of how we’re going to build products, how we’re going to build software for customers, and then also how we’re going to run a company?” These were the questions that ultimately led to the substantial restructuring.
Executing with Principle and Empathy
Such a drastic change, even if strategically motivated, carries significant human impact. Block approached the reduction with core principles:
- Reliability (P00): Ensuring no outages or service disruptions.
- Customer Trust & Compliance: Navigating complex regulatory environments without compromise (e.g., compliance teams were largely untouched).
- Durable Growth: Continuing to build on the roadmap and make long-term bets, albeit with smaller teams.
From an execution standpoint, Block aimed for transparency and generosity. They offered substantial severance packages, didn’t instantly cut technology access, and held an all-hands meeting where Jack Dorsey and the executive team explained the decision directly to the entire company.
Culturally, the impact was profound. While the initial days brought “shock” and “ambiguity,” the company quickly adapted. A key outcome was a massive reduction in meetings, “probably like 70 or 80%,” freeing up time for “building and work.” Weekly all-hands with Jack Dorsey became a new norm, fostering direct communication and a sense of shared purpose. The new Block is “smaller, leaner, [with] fewer layers, larger spans,” and a renewed focus on building.
Inside Block’s AI-Powered Engine: Goose, G2, and Builderbot
Block’s internal AI infrastructure is the engine driving this transformation. At its core is Goose, an “agent harness” that is model-agnostic. This means Block can swap in different foundational models (Anthropic, OpenAI, open-source models – they use over 120 internally) depending on the specific task. Goose provides the framework for AI agents to operate.
Building on Goose, Block developed G2, an internal “agentic operating system.” G2 allows anyone within the company to automate “any deterministic workflow,” empowering employees across various functions.
Perhaps most illustrative of the shift in software development is Builderbot. This tool “autonomously merging PRs and actually like building features to 100%.” While it often completes 85-90% of a complex feature, with a human providing the final 10% context and polish, its impact is undeniable. The ability to go from “idea to like this is in the hands of 100,000 or a million customers has been compressed massively since December.”
A New Way of Working
The organizational structure has been completely reshaped. Gone are the classic hierarchical, functional teams of eight server engineers, four client engineers, a PM, and a designer working linearly. Instead, Block now operates with “small squads” of one to six people, offering “way more flexibility and fluidity.” These squads can fluidly move between products, iterating rapidly.
Layers have been dramatically reduced, by “50 or 60% on the development side,” meaning information flows more freely. The workflow itself has become non-linear: instead of sequentially submitting a Pull Request (PR), getting a review, and making changes, engineers now manage “14 agents who are building PRs on my behalf right now.” The human role shifts to “context switching between all of those,” checking on work, nudging it, and committing it.
This agent-driven workflow extends beyond engineers to product managers and growth marketers, who now manage “countless agents running right now.” The role of the human is increasingly to oversee, guide, and refine the output of these intelligent systems.
Beyond core development, AI is automating “anytime there’s a deterministic workflow.” This includes customer support, where chatbots and AI phone support handle the majority of inquiries. In product operations, risk operations, and compliance operations, AI models are increasingly making decisions. While a “human in the loop” is currently critical, Jennings predicts that “over time it’s like pretty obvious that these systems are just going to be so much better than like having a thousand humans who are doing that work.”
Generative UI: Reshaping Products and User Experience
Block’s AI transformation isn’t just internal; it’s profoundly changing the products customers interact with. The company, which has functionalized its structure to build a unified financial platform across Square, Cash App, and Afterpay, is leveraging its agentic infrastructure to create dynamic, personalized user experiences.
“For the past 10 or 15 years, everyone’s used to a static UI, a rigid UI,” Jennings noted. “That’s going to fundamentally change in the next like six months. Generative UI is here.”
This means your Cash App, for example, will look “really different” from someone else’s, not just through basic personalization but through on-the-fly generation. Moneybot, Block’s “CFO in your pocket” within Cash App, can generate charts and visualizations dynamically based on a user’s query (“How have I been spending my money?”). These visualizations are not hardcoded; they are generated in real-time by the AI.
On the Square side, ManagerBot allows a multi-location restaurant owner to “build me an app where I can manage scheduling for these two locations and like automatically fire off texts… to my employees.” The app itself, and its specific look and feel, are generated on demand and are not part of the application’s source code.
This generative UI offers “way more control” and promises “higher engagement” and “better product.” However, it presents new challenges, particularly in quality assurance (QA) for “non-deterministic outputs.” Block is also heavily investing in “proactive intelligence,” understanding that customers may not know the right prompts. The AI needs to anticipate needs and offer relevant insights and actions.
The New Moat: Understanding the Unseen
In a world where code can be generated by AI, traditional moats like network effects, regulatory licenses, and even hardware (which is harder to “vibe code”) remain important in the near term. But in the long term, Jennings argues, defensibility will hinge on something far more abstract: “the extent to which the company understands something that is pretty hard for other companies to understand.”
Block is building towards becoming an “intelligent system itself,” sitting on a “rich data and deep insight” about how sellers and buyers participate in the economy. The key question for any company will be: “how quickly can you iterate to improve that understanding over time?”
This involves building “world models” internally and externally – understanding customers, but also understanding how Block itself operates. Companies will need a “markdown file” of who they are, their values, metrics, and priorities, combined with a feedback loop from their unique “signal” (that deep understanding) and powerful AI tools like Builderbot or Clawed Code.
This loop – “This is what I’m seeing. This is what’s happening. Great. This is our markdown file for Block… and then you have a gentic system, so you can just build stuff” – will iterate at unprecedented speeds. What once took months to build a feature with humans, now takes a week or two. In the future, this loop could run “hundreds, thousands of times a day,” with humans acting more as editors or nudgers.
The ultimate warning is stark: “If your answer to that [what you deeply understand] is ‘I don’t know,’ then you maybe could get vibe coded away.”
A Glimpse into the Future
Block’s radical transformation offers a compelling glimpse into the future of the tech industry. While the immediate implication might be fewer engineers, designers, and PMs for a given product or roadmap, Jennings suggests a “Jevons paradox” effect. The increased efficiency might lead to a “superset of things that can be built,” potentially resulting in more tech companies or the expansion of development into entirely new sectors.
The message is clear: AI is not merely a tool for optimization; it’s a force for fundamental restructuring. Companies that embrace this shift, understand its implications, and build their operations and products around intelligent systems will lead the next era of innovation. For Block, the future of programming is not about writing code by hand; it’s about orchestrating intelligence.
Based on “Today’s Mission to the Moon” from New York Times Podcasts Watch the original video
Our Next Giant Leap: Why Humanity is Returning to the Moon
Nearly six decades after the United States made history by landing men on the moon, a new era of lunar exploration is dawning. This time, the mission isn’t just about planting a flag; it’s about building a future, a permanent human presence beyond Earth. Today marks a pivotal moment in this ambitious endeavor, as four astronauts prepare to embark on a journey that will swing them around the moon and back, a critical test for humanity’s sustained return to our celestial neighbor.
The mission, known as Artemis 2, is the latest step in NASA’s grand plan to not only put humans back on the moon but to establish a long-term base there. As Ken Chang, a science reporter for The New York Times, explains, the Artemis program is designed in carefully orchestrated phases, acknowledging the immense challenges of lunar exploration.
“The goal this time in this program, which is called Artemis, is to get people back to the moon, but also stay there this time,” Chang notes. “And so this program is broken up into pieces because it’s really hard to land on the moon. You don’t want to do all the hard things all at once.”
The journey began with Artemis 1 in 2022, an uncrewed flight that sent the Orion spacecraft around the moon for several weeks, successfully proving the basic machinery. Now, Artemis 2 is taking the next monumental leap: sending a crew of four astronauts to test the critical life support systems. Their primary objective, as Chang starkly puts it, is “to not die.” This mission will ensure that the spacecraft can sustain human life for an extended period, producing carbon dioxide, water, and other waste that must be managed. If successful, Artemis 2 will pave the way for Artemis 3, which aims to land astronauts on the lunar surface within a few years.
Beyond the Footprint: A Vision for a Permanent Lunar Outpost
But why return to the moon now, after so many decades? The rationale extends far beyond simply revisiting a past achievement. NASA’s vision for the Artemis program is to establish a sustainable presence, a permanent lunar base that will serve multiple purposes.
Initially, this base would function as a scientific research station, akin to the outposts currently operating in Antarctica. Scientists could study the moon’s unique environment, its geological history, and its potential resources in unprecedented detail.
Beyond pure science, the moon base holds significant commercial and economic potential. There’s growing interest in harnessing the moon’s resources. One of the most intriguing prospects is the mining of Helium-3, a light isotope of helium that is exceptionally rare on Earth but more prevalent on the moon’s surface. Helium-3 is considered a potential fuel for future fusion reactors, offering a clean and abundant energy source. It’s also believed to be useful for advanced quantum computers, which could revolutionize fields like artificial intelligence. With Helium-3 currently valued at roughly $3 million a pound on Earth, even a small quantity from the moon could yield substantial profits, attracting private companies already planning such ventures.
The moon also presents unique opportunities for groundbreaking scientific endeavors. Imagine a giant radio telescope built on the far side of the moon. This shielded location, protected by the moon’s entire mass from Earth’s constant radio noise – generated by TV, cell phones, and podcasts – would offer an unparalleled vantage point to “listen to the universe.” Such a telescope could detect faint signals from just after the Big Bang billions of years ago, offering humanity a chance to “hear the ancient sounds of really the dawn of time itself,” the echoes of creation itself.
Furthermore, the moon is envisioned as a testing ground for future missions to Mars. Many of the challenges of living and working on the moon – developing nuclear power plants, constructing habitats, and refining life support systems – are directly applicable to a Mars mission. The moon’s gravity and environmental conditions provide a crucial proving ground before humanity attempts the even more ambitious journey to the Red Planet.
Finally, there’s the romantic notion of humanity spreading out into the solar system, transcending our confinement to a single planet. A moon colony represents the first tangible step towards becoming a multi-planet species. Underlying all these aspirations is a compelling geopolitical competition, particularly with China. Being the first to establish a significant presence on the moon means setting the rules for space commerce, securing prime locations, and controlling potential resources. “Being first means being the one who’s in charge,” Chang emphasizes.
The Crew: Four Pioneers on a 10-Day Odyssey
At the heart of this monumental mission are four individuals, carrying the weight of these dreams on their shoulders. Their 10-day journey is a meticulously planned sequence of events, designed to push the boundaries of human spaceflight.
The commander of Artemis 2 is Reid Wiseman, a former fighter pilot in the U.S. Navy who served two deployments in the Middle East. Before his assignment to this historic mission, he led NASA’s astronaut office. Wiseman reflects on the profound nature of their journey: “I was outside last night and I was looking up at the moon thinking the next time I see this sight there’s a good chance that we will have been around the far side and back.” A poignant detail about Wiseman is that his wife passed away a few years ago, meaning he will be leaving his daughters behind for this critical 10-day period.
Joining him is Victor Glover, also a former naval aviator. Glover made history as the first Black man to serve an extended stay on the International Space Station, and with Artemis 2, he will become the first Black man to travel to the moon. He speaks to the collaborative spirit of the mission: “10 years from now when the next challenging thing happens, maybe we can look back on this and go, ‘Hey guys, remember we did that. We had our own wing shot, remember?’ And it was global. It was international. And we did it together.”
Christina Koch, an electrical engineer, holds the record for the longest single space flight by a woman, at 328 days. She worked on NASA missions on the ground before being selected as an astronaut. Koch emphasizes the collective nature of their achievements: “All these things that we talk about first are really not about any one individual’s accomplishment, but more about celebrating where we are at.”
Rounding out the crew is Jeremy Hansen, a Canadian astronaut, making him the first non-American to venture into deep space. Hansen possesses a dry sense of humor, quipping that “If something goes wrong on this mission, then NASA can blame Canada.” More seriously, he notes, “We don’t know the ripple effects of what we’re about to do, but we can do it well to the best of our ability and have joy while we’re doing it. And hopefully that will make a small contribution.”
Their journey begins with a long day even before launch, waking eight hours prior. After donning their space suits, they’ll be driven to the launchpad, ascend a massive tower, and enter the Orion spacecraft. They’ll spend four hours inside before the countdown reaches zero, and just eight minutes later, they’ll be in space.
The first two days in orbit will be spent meticulously checking every system on Orion through two looping orbits around Earth. Only once everything is confirmed to be working perfectly will they fire the engines to propel them towards the moon, a quarter-million-mile journey that will take approximately four days. Life inside the Orion capsule, which offers about as much space as two minivans, will be a unique experience, made more comfortable by the ability to float and utilize all available volume.
Day six marks the mission’s climax: the lunar flyby. As they approach the moon, it will appear roughly the size of a basketball held at arm’s length. The moon’s gravity will then gracefully pull them around its far side, an event that will cause a temporary loss of radio communication with Earth for about 40 minutes. During this silent period, the astronauts will make critical observations, seeing parts of the far side of the moon that no human eyes have ever witnessed in actual daylight. Previous missions passed over these areas in darkness, leaving them unseen by human eyes in their full illumination. The trajectory is so precise that the moon’s gravity will effectively “sling them around and throw them right back toward Earth without them doing much of anything,” a testament to the power of celestial mechanics.
After the lunar flyby, the astronauts will endure another three days of travel back to Earth – what might be the “boring part of the trip,” as Chang describes it. On the final day, Earth’s gravity will guide them into the atmosphere for a splashdown in the Pacific Ocean off San Diego. They’ll be recovered by ship, flown to shore for medical checks, and finally return to Houston, bringing Artemis 2 to a triumphant close.
A Shifting Landscape: NASA and the New Space Age
Artemis 2 represents a significant moment for NASA, a “triumph for them in that the old ways of doing things succeeded,” as Chang notes. This mission is largely an “old school NASA” production, with the agency designing and operating the entire spacecraft. However, it also marks the end of an era.
For the subsequent mission, Artemis 3, the landscape of space exploration will dramatically shift. Private companies like SpaceX, founded by Elon Musk, and Blue Origin, started by Jeff Bezos, will become integral partners. These “new space” companies are developing the lunar landers that will eventually transport astronauts to the moon’s surface. Artemis 2, in this sense, is the last major mission where NASA operates independently from these burgeoning private space giants, paving the way for a collaborative future where government agencies and commercial enterprises work hand-in-hand to explore the cosmos.
A Message of Hope in Turbulent Times
The launch of Artemis 2 comes at a time of global turbulence, reminiscent of the era when humanity first ventured to the moon. In 1968, a year marked by assassinations, social unrest, and war, the Apollo 8 mission provided a profound moment of unity and hope. That Christmas Eve, as the three Apollo 8 astronauts orbited the moon, they took turns reading from the Book of Genesis, sharing a message of creation and peace with a world in turmoil.
“In the beginning, God created the heaven and the earth,” they broadcast. “And God said, ‘Let there be light.’ And there was light.” This simple act resonated deeply, with some even suggesting that Apollo 8 “saved 1968.”
While we don’t know what message the Artemis 2 crew might share upon their return, the imagery alone – of Earth seen from a quarter-million miles away – has the potential to offer a similar moment of calm and perspective. It’s a reminder that despite our differences, we are all inhabitants of the same fragile planet, a “good earth” suspended in the vastness of space.
Artemis 2 is more than just a mission; it’s a testament to human ingenuity, resilience, and our innate drive to explore. It’s a blend of scientific ambition, economic opportunity, geopolitical strategy, and the enduring romantic quest to reach for the stars. As the four astronauts prepare for their journey, they carry with them not just the hopes of a nation, but the aspirations of humanity for a future among the stars.
한국어
“Robert Herjavec: The hidden reason smart people stop growing | Big Think+” — Big Think 기반 기사 원본 영상 보기
똑똑한 당신이 성장을 멈추는 이유? 로버트 허재벡이 밝히는 ‘진짜 멘토링’의 힘
누구나 한 번쯤 “저 사람처럼 성공하고 싶다”는 생각을 해본 적이 있을 겁니다. 그리고 그 바람을 실현하기 위해 멘토를 찾곤 합니다. 하지만 인기 TV 프로그램 ‘샤크 탱크’의 스타 사업가 로버트 허재벡(Robert Herjavec)은 많은 사람이 멘토십(mentorship)에 대해 오해하고 있으며, 이 때문에 똑똑한 사람들조차 성장의 기회를 놓치곤 한다고 지적합니다. 그가 말하는 멘토십의 숨겨진 진실은 무엇일까요?
멘토십에 대한 세 가지 흔한 오해
허재벡은 사람들이 멘토십에 대해 가지고 있는 세 가지 큰 오해를 꼬집습니다. 이러한 오해는 우리가 진정한 성장의 기회를 놓치게 만드는 주된 원인이 됩니다.
1. 멘토는 유명인이어야 한다는 환상 “로버트, 제 멘토가 되어줄 수 있나요?” 허재벡이 가장 많이 받는 질문 중 하나입니다. 그의 대답은 단호하게 “아니요, 저는 너무 바쁩니다”입니다. 그는 멘토십에 대한 가장 큰 오해가 “샤크 탱크”에 나오는 유명인이나 모두가 아는 리더가 멘토가 되어야 한다고 생각하는 것이라고 말합니다. 하지만 멘토의 진정한 역할은 현재 당신의 경력에 도움이 될 무언가를 가르쳐주는 것입니다. 배움은 누구에게서든 얻을 수 있으며, 유명세와는 상관없습니다.
2. 멘토십은 영원불멸하다는 믿음 멘토십이 평생 지속되는 불변의 관계라는 생각은 사실이 아닙니다. 23세에 조직에 처음 발을 들였을 때 필요한 조언과 45세에 고위 부사장(Senior Vice President)이 되었을 때 필요한 조언은 완전히 다릅니다. 당신의 삶이 시간이 지남에 따라 변하듯이, 당신의 멘토 또한 필요에 따라 자연스럽게 변화해야 합니다. 한 명의 멘토가 당신의 모든 성장 단계를 책임질 수는 없습니다.
3. 멘토십은 공식적인 계약 관계라는 착각 많은 사람이 멘토십을 어떤 보이지 않는 계약을 맺는 것과 같은 공식적인 과정으로 여깁니다. 하지만 허재벡은 자신의 인생에서 가장 큰 멘토 중 일부는 자신이 멘토였다는 사실조차 몰랐다고 말합니다. 에이비스 렌터카(Avis Rent A Car)를 창업한 워렌 에이비스(Warren Avis)가 대표적인 예입니다. 허재벡은 그에게서 많은 것을 배웠지만, 에이비스는 자신이 멘토 역할을 했다는 것을 알지 못했을 것이라고 합니다. 멘토십은 때로는 그저 지켜보고, 배우고, 자연스럽게 흡수하는 간접 학습(osmosis)을 통해 이루어지기도 합니다. 특히 오늘날 유튜브와 같은 온라인 플랫폼에서는 수많은 정보를 통해 간접적으로도 충분히 배울 수 있습니다.
진정한 학습과 멘토를 찾는 핵심
그렇다면 어떻게 해야 진정으로 배우고 성장하며, 자신에게 맞는 멘토를 찾을 수 있을까요? 허재벡은 세 가지 핵심 요소를 강조합니다.
1. 성장을 위한 첫걸음: 열린 마음과 겸손 허재벡은 “자만심(ego)은 세상의 어떤 것보다 많은 경력을 망쳤습니다”라고 단언합니다. 당신이 아무리 뛰어나고 당신의 아이디어가 아무리 훌륭해도, 약간의 겸손은 반드시 필요합니다. 모든 답을 가지고 있지 않을 수도 있다는 가능성에 마음을 여는 것이 중요합니다. 자신감(confidence)은 훌륭하지만, 오만함(arrogance)은 나쁩니다. 세상이 당신에게 가르쳐줄 것이 아무것도 없다고 느낀다면, 당신은 세상으로부터 아무것도 배울 수 없습니다. 멘토십은 결국 끊임없이 배우는 삶의 자세이며, 당신은 누구에게서든 배울 수 있습니다. 허재벡은 심지어 한 청소부에게서 놀라운 근면함과 완벽주의에 대한 광적인 집착, 그리고 끊임없는 개선 의지를 배운 적이 있다고 고백합니다. 배움은 당신 주변 어디에나 있습니다.
2. 기회를 포착하는 ‘빨간 차 이론’ 허재벡의 멘토 중 한 명이 그에게 ‘빨간 차 이론’을 들려주었습니다. 오늘 출근길에 빨간 차를 몇 대 봤는지 묻는다면, 아마 정확히 대답하지 못할 것입니다. “몇 대 본 것 같긴 한데…” 정도겠죠. 하지만 내일 출근길에 빨간 차 수를 세어보라고 한다면, 당신은 분명 그렇게 할 것입니다. 기회와 멘토십도 이와 같습니다. 당신이 마음을 열고 주의를 기울이면, 그것들은 당신 주변에 늘 존재합니다. 핵심은 ‘인식’입니다. 주변에 얼마나 많은 빨간 차가 있는지 알기 위해서는 먼저 빨간 차를 의식하고 찾아야 합니다.
3. 도움을 청하는 용기 허재벡은 자신의 경험상, 물론 세상에는 타인에게 불친절한 사람들도 존재하지만, 대다수(98%)의 사람들은 매우 친절하다고 말합니다. 그리고 당신이 도움을 요청하면, 대부분의 경우 사람들은 기꺼이 당신을 도울 것입니다. 그러니 도움을 청하는 것을 결코 두려워하지 마세요.
성공적인 멘토-멘티 관계를 위한 현명한 접근법
마지막으로, 허재벡은 멘토와 멘티(mentee) 관계를 성공적으로 이끌기 위한 실질적인 조언을 전합니다.
1. 시간 존중: 멘토의 가장 소중한 자산 멘토와의 관계에서 가장 중요한 것은 시간을 존중하는 것입니다. 허재벡은 “저는 매우 바쁜 사람입니다. 그리고 자유 시간이 있다면 아이들과 시간을 보내고 싶지, 아마 당신과 시간을 보내고 싶지는 않을 겁니다”라고 솔직하게 말합니다. 멘토의 시간을 존중하는 것은 기본 중의 기본입니다.
2. 현명한 소통 방식: 마크 큐반의 예시 올바른 질문을 올바른 시기에 하는 것만큼이나 중요한 것이 올바른 소통 방식을 찾는 것입니다. 그는 마크 큐반(Mark Cuban)의 예를 듭니다. 큐반은 전화 통화를 극도로 싫어하지만, 밤 10시 30분에 이메일을 보내면 3분 안에 답장을 받을 수 있습니다. 허재벡 자신도 마찬가지입니다. 그는 전화 통화가 하루의 흐름을 방해하기 때문에 싫어하지만, 이메일은 자신의 일정에 맞춰 적절한 시간에 답장할 수 있습니다. 허재벡은 매일 아침 5시에 일어나는데, 오전 5시부터 9시 사이에 이메일을 확인합니다. 이때 연락하는 것이 그에게는 가장 좋은 시간입니다. 멘티인 당신이 멘토의 작업 방식을 이해하고 그에 맞춰 소통해야 최대한의 도움을 받을 수 있습니다.
3. 명확한 경계 설정: 일과 삶의 균형 직장에서 “우리는 한 가족”이라는 말을 자주 듣지만, 허재벡은 현실은 그렇지 않다고 말합니다. 그의 회사는 사업이었고, 그의 가족은 그의 가족이었습니다. 그는 밤이나 특정 시간에 가족에게서 시간을 빼앗으려는 시도에 불편함을 느꼈다고 합니다. 모든 사람은 다르므로, 멘토의 경계를 명확히 이해해야 합니다. 어떤 사람들은 주말에도 연락하는 것을 좋아하지만, 허재벡은 그렇지 않습니다. 멘토의 개인적인 공간과 시간을 존중하는 것이 관계 유지에 필수적입니다.
로버트 허재벡의 통찰은 멘토십이 결코 거창하거나 복잡한 것이 아님을 보여줍니다. 중요한 것은 유명한 멘토를 찾는 것이 아니라, 겸손한 자세로 끊임없이 배우고, 주변의 기회를 포착하며, 현명하게 관계를 구축하는 것입니다. 이처럼 ‘진짜 멘토링’의 힘을 이해하고 적용한다면, 당신의 성장은 결코 멈추지 않을 것입니다.
“How Linear Turned AI Agents Into First-class Users” — Every 기반 기사 원본 영상 보기
Linear, AI 에이전트를 ‘일등 시민’으로: 신중함과 깊은 이해로 혁신을 이끌다
인공지능(AI)이 전 세계 산업을 재편하고 있는 가운데, 수많은 기업들이 저마다의 AI 전략을 내놓으며 격변의 시대를 헤쳐나가고 있습니다. 특히 SaaS(Software as a Service) 기업들에게 AI는 기회이자 동시에 거대한 도전으로 다가오고 있습니다. 이러한 흐름 속에서 프로젝트 관리 및 이슈 트래킹 도구인 Linear는 서두르지 않으면서도 깊이 있는 이해를 바탕으로 AI 시대를 선도하는 독특한 접근 방식을 보여주고 있습니다.
Every 채널의 인터뷰에서 Linear의 공동 창업자이자 CEO인 카리 (Karri)는 AI 에이전트를 ‘일등 시민(First-class Users)‘으로 대하는 Linear의 철학을 소개하며, 단순한 기술 도입을 넘어선 근본적인 제품 전략과 조직 문화의 변화를 이야기했습니다.
AI 시대, SaaS의 생존 전략 – Linear의 통찰
최근 주식 시장에서는 ‘SaaS는 끝났다’는 인식이 확산되고 있습니다. 많은 SaaS 기업들이 AI 열풍에 휩쓸려 단순한 챗봇 기능을 추가하는 데 급급했지만, 이는 시장의 기대에 미치지 못하거나 실제 가치를 창출하지 못하는 경우가 많았습니다. 카리는 이러한 단순한 관점에 대해 반박하며, AI 시대에 SaaS 기업이 직면한 본질적인 변화를 지적합니다.
“SaaS 기업으로서 미래 현금 흐름에 대한 불확실성이 커진 것은 사실입니다. 하지만 ‘사람들이 자기만의 CRM 도구를 코딩할 것’이라는 식의 이야기는 너무 단순합니다.” 카리는 AI가 모든 것을 직접 코딩하게 만들지는 않을 것이라고 말합니다. 대신, 그는 기존의 대형 공개 기업들이 가장 큰 타격을 입을 것이라고 예상합니다. “대부분의 공개 기업들은 가장 유연하거나 견고한 솔루션이 아닙니다. 이들은 대기업이 사용하는 거대한 솔루션이며, 그 안에 일종의 관성(inertia)이 존재합니다. 그들의 해자(moat)가 일종의 방식으로 사라지고 있는 셈이죠.”
Linear는 이러한 변화에 대응하기 위해 스스로 ‘데이 원(Day One)의 세계’에 살고 있다고 선언합니다. 이는 과거의 성공이나 결정에 얽매이지 않고, 매일매일 새로운 관점에서 문제를 바라보고 제품을 재구상해야 한다는 의미입니다. AI 에이전트가 제품 개발 프로세스에 통합될 때 어떤 새로운 문제들이 발생하는지, 그리고 Linear가 어떻게 이를 도울 수 있을지에 집중하는 것입니다. 이는 수십 년간 존재해 온 대기업들에게는 훨씬 어려운 과제이며, 스타트업이나 성장 기업들이 더 잘 해낼 수 있는 부분이라고 카리는 강조합니다.
서두르지 않는 혁신: ‘이해’에서 시작된 AI 통합
Linear의 AI 접근 방식의 핵심은 ‘이해(understanding)‘에 있습니다. GPT-3가 처음 출시되었을 때, Linear는 서둘러 AI 기능을 제품에 통합하지 않았습니다. 카리는 디자인 배경을 가진 사람으로서 문제를 해결하기 전에 먼저 깊이 이해하려 노력한다고 말합니다. “기술 세계에서는 사람들이 종종 뭔가를 ‘할 수 있다’는 이유만으로 뛰어들곤 합니다. 하지만 ‘해야 하는가’, ‘실제로 도움이 되는가’를 먼저 생각해야 합니다.”
이러한 철학은 초기 AI 챗봇 열풍 당시 Linear의 경험에서도 드러납니다. 많은 기업들이 너도나도 “우리는 이제 AI 회사”라며 챗봇을 통합했지만, Linear는 내부적으로 테스트해본 결과 “이것이 정말 유용한가? 실제로 어떤 워크플로우에서 필요한가?”라는 의문을 가졌습니다. 그래서 Linear는 지난 몇 년간 사용자들이 AI를 실제로 어떻게 활용하고 싶어 하는지, 어떤 워크플로우가 가치를 창출할지 이해하는 데 시간을 보냈습니다.
그 결과 Linear는 독특한 전략을 채택했습니다. 첫째, 개방형 에이전트 플랫폼을 구축하여 다른 코딩 에이전트들이 Linear와 쉽게 통합될 수 있도록 했습니다. 잘 정리된 문서(docs)를 제공하여 에이전트들이 자체적으로 통합을 구축하도록 유도했고, 그 결과 OpenAI의 코덱스(Codex)와 같은 주요 에이전트는 물론, 코인베이스(Coinbase)나 브램(Bram) 같은 Linear 고객사들이 자체 개발한 코드 에이전트들도 Linear와 연동되었습니다.
이러한 접근 방식에서 Linear는 에이전트들을 위한 ‘가이드 시스템(system for guiding the agents)‘이자 ‘맥락 구축자(building this context)‘의 역할을 수행합니다. 카리는 “하나의 에이전트만 존재하지 않을 것이며, 모든 사람이 수많은 에이전트를 가질 것이고, 기업들은 자체 에이전트를 구축할 것”이라고 예측합니다. Linear는 이 생태계의 모든 것을 소유하려 하기보다, 다른 기업들과 협력하며 가치를 창출하는 데 집중했습니다.
Linear Agent의 등장: 통제권 확보와 심층 통합의 필요성
하지만 외부 에이전트와의 통합만으로는 한계가 있었습니다. 카리는 “우리가 직접 통제하지 못할 때 어렵다”며, Linear가 가진 아이디어를 직접 구현할 수 없다는 점을 지적했습니다. 이러한 한계를 극복하기 위해 Linear는 자체적인 ‘Linear Agent’를 개발하고 있습니다. 이 에이전트는 Linear의 제품 기획 워크플로우는 물론, 디자인, 엔지니어링 전반에 걸쳐 더욱 매끄러운 엔드투엔드(end-to-end) 경험을 제공할 것입니다.
Linear Agent는 사용자의 업무 맥락과 조직의 맥락을 이해하여 다양한 방식으로 활용될 수 있습니다. 예를 들어, 디자이너는 Linear Agent를 통해 문제의 본질을 더 깊이 이해할 수 있고, 개발자는 에이전트와 함께 코드를 작성할 수 있습니다. 특히 코딩 에이전트의 경우, 온라인에서 코드 변경 사항(diffs)을 확인하고 가이드하며, 샌드박스 환경에서 버그 수정과 같은 작은 작업을 자동화하여 백그라운드에서 실행할 수 있습니다.
카리는 이러한 자체 에이전트 개발의 가장 큰 이점으로 ‘맥락(context)’ 관리의 효율성을 꼽습니다. “클로드(Claude)나 챗GPT(ChatGPT) 같은 도구를 사용할 때, 에이전트에게 어떤 맥락을 가져와야 할지 명시적으로 알려줘야 합니다. 하지만 Linear에서는 맥락이 이미 존재하며, 이를 작업 흐름의 일부로 영리하게 주입할 수 있어 훨씬 자연스럽습니다.” 이는 컨텍스트 윈도우(context window)를 불필요하게 스팸하지 않고도 에이전트가 관련성 높은 정보를 바탕으로 작업할 수 있게 합니다.
비즈니스 모델 측면에서도 중요한 변화가 예상됩니다. Linear Agent의 도입은 토큰(token) 사용 비용 발생으로 이어지기 때문입니다. 카리는 코딩 에이전트의 경우 사용량 기반(usage-based) 과금을 도입해야 할 정도로 비용이 발생할 수 있다고 언급했습니다. 반면, 질문에 답변하는 기본 Linear Agent 기능은 시스템에 포함될 가능성이 높습니다. Linear는 범용적인 에이전트 플랫폼을 지향하기보다는, ‘제품 컨텍스트 또는 제품 메모리 플랫폼’으로서의 역할을 강화하고, 제품 개발 워크플로우에 특화된 에이전트 활용을 목표로 합니다.
AI가 바꾼 Linear 내부의 일하는 방식
Linear는 AI 에이전트를 외부 고객뿐만 아니라, 내부 팀의 업무 방식 혁신에도 적극적으로 활용하고 있습니다.
1. 팀 문화와 성과 측정: 초기에는 팀원들이 새로운 AI 도구를 사용하는 데 익숙해지도록 장려하는 시간이 필요했습니다. 하지만 이제는 대부분의 엔지니어와 일부 디자이너, PM들이 에이전트 코딩 도구를 사용합니다. 카리는 여기서 중요한 점은 ‘어떤 지표를 측정하는가’라고 강조합니다. “사람들이 이제 ‘코드의 몇 퍼센트가 에이전트가 작성한 것인가’, ‘얼마나 많은 PR(Pull Request)을 병합했는가’를 최고의 허영 지표(vanity metric)로 삼습니다. 하지만 이는 올바른 지표가 아닙니다.”
그는 토큰 사용량이나 생산량 같은 단순한 지표보다, **‘실제로 가치를 창출하는가?’, ‘제품이 개선되고 있는가?’, ‘버그가 줄었는가?’**와 같은 본질적인 지표에 집중해야 한다고 말합니다. 특히 ‘버그’는 매우 측정 가능한 지표이며, Linear는 ‘제로 버그(Zero Bugs)’ 정책을 통해 모든 버그를 일주일 내에 수정하는 것을 목표로 합니다. AI 코딩 에이전트가 버그 수정의 첫 단계를 자동화하고 엔지니어에게 검토를 위임함으로써, 이 프로세스를 훨씬 빠르게 진행할 수 있게 되었습니다.
2. 제품 기획(PM) 워크플로우: 카리는 개인적으로 Linear Agent의 ‘스킬(Skill)’ 기능을 활용하여 제품 기획 프로세스를 혁신하고 있습니다. 그는 Linear의 내부 문서와 블로그 게시물을 학습시켜 ‘Linear 제품 팀원처럼 행동하며 문제의 근본적인 필요를 이해하고 솔루션을 제안하는’ 개인 스킬을 만들었습니다.
예를 들어, 수백 명의 사용자가 요청한 ‘이슈당 여러 담당자 지정’과 같은 기능 요청이 있을 때, 카리는 이 스킬을 사용하여 해당 요청의 다양한 배경과 고객의 활동을 분석하게 합니다. 에이전트는 문제의 본질적인 필요(underlying need)를 파악하고, 여러 추천 사항을 제시하여 팀이 해당 기능을 지금 다룰지, 나중에 다룰지, 혹은 아예 다루지 않을지 결정하는 데 도움을 줍니다. 이는 과거에 직접 여러 사람에게 묻거나 자료를 찾아봐야 했던 과정을 매우 빠르게 단축시킵니다.
3. 디자인 및 엔지니어링 워크플로우: 카리는 개인적으로는 여전히 피그마(Figma)에서 수동으로 디자인하며 ‘느림의 미학’을 통해 깊이 있는 탐색을 즐긴다고 말합니다. 하지만 디자인 팀 전체적으로는 AI를 활용하여 더 많은 프로토타입을 빠르게 구축하고 테스트하는 데 주력하고 있습니다. Linear는 강력한 빌드 시스템을 갖추고 있어, 프로토타입을 직접 앱에 통합하여 실시간으로 테스트할 수 있습니다.
엔지니어링 측면에서는 다른 많은 기업들과 유사하게, 문제가 식별되면 AI를 통해 해결 속도를 크게 높이고 있습니다. 특히 Linear는 슬랙(Slack) 에이전트를 활용하여 팀 대화에서 결정된 사항을 즉시 Linear 이슈로 전환하여 추적하고 실행 가능하게 만듭니다. 이는 회의를 통해 프로젝트를 시작하고 사람을 할당하는 전통적인 방식보다 훨씬 빠르게 작업을 시작할 수 있게 합니다.
카리는 이 모든 변화의 공통적인 패턴이 “일종의 루프(loop)를 단축시켜 더 빠르게 만들고, 기다리지 않고 즉시 실행할 수 있게 하는 것”이라고 요약합니다.
속도와 사려 깊음의 조화: ‘문제 발견’은 느리게, ‘해결’은 빠르게
AI는 의심할 여지 없이 작업 속도를 엄청나게 가속화시킵니다. 하지만 카리는 이러한 속도가 항상 좋은 것만은 아니라고 강조하며, ‘속도’와 ‘사려 깊음’ 사이의 중요한 균형을 제시합니다.
“우리는 결정을 내리거나 결정을 ‘스피드런(speedrun)‘하는 데 있어서는 빠르게 움직이지 말아야 합니다. 어떤 사람들은 아이디어를 내면 바로 구현해 버리는데, 나중에 보면 아무도 왜 이 아이디어가 존재하는지 모르는 경우가 많습니다.” 모든 새로운 프로토타입이나 아이디어가 유용해 보일 수 있지만, 다른 우선순위와 비교하여 얼마나 유용한지, 지금 이 아이디어에 전념할 가치가 있는지 신중하게 판단해야 한다는 것입니다.
카리는 Linear가 많은 프로세스를 가지고 있지는 않지만, 일단 어떤 작업이나 프로젝트에 전념하기로 결정하면 빠르게 진행되기를 원합니다. “문제 해결에 대한 루프는 빨라야 하지만, 문제 발견은 빨라서는 안 됩니다. 올바른 문제를 찾고, 문제에 대한 올바른 접근 방식을 찾는 데 시간을 투자해야 합니다. 일단 결정되면, 그 다음에는 빠르게 진행할 수 있습니다.”
그는 디자인 과정의 ‘개념 작업(conceptual work)‘을 예로 듭니다. 때로는 디자인의 결과물이 당장 출시할 제품이 아니라, ‘콘셉트 카’처럼 아이디어를 탐색하고 이해를 돕기 위한 개념일 수 있습니다. 이러한 과정을 통해 사람들은 당장의 생산성이나 제약에 대한 두려움 없이 새로운 아이디어를 자유롭게 탐색하고, 나중에 그 가치를 평가하여 다음 단계로 나아갈 수 있습니다.
물론, AI 시대에는 LLM(대규모 언어 모델)의 비결정성(non-deterministic) 때문에 모든 것을 내부적으로 완벽하게 생각한 후 출시하기는 어렵습니다. 카리는 “어느 시점에는 내부적으로 시도해보고, 더 나아가 고객과 함께 베타 테스트를 해야 합니다. 그리고 각 단계마다 명확한 목표를 가져야 합니다. 베타 테스트의 목표는 ‘사용자 워크플로우를 이해하고 어떻게 개선할지 배우는 것’이지, ‘가능한 한 빨리 출시하는 것’이 아닙니다.”라고 말하며, 신중한 목표 설정의 중요성을 강조합니다.
결론: 인간의 장인정신과 AI의 시너지
Linear의 사례는 AI 시대의 성공이 단순히 최신 기술을 도입하는 것을 넘어선다는 것을 명확히 보여줍니다. 이는 제품의 본질적인 가치와 사용자 워크플로우에 대한 깊은 이해, 그리고 장기적인 관점에서 품질을 추구하는 인내심에 달려 있습니다.
카리의 말처럼, AI는 인간의 ‘장인정신(craft)‘과 ‘사고(thinking)‘가 필요한 영역에서 부담을 덜어주고, 자동화를 통해 인간이 더 고차원적인 문제 해결과 창의적인 작업에 집중할 수 있도록 돕는 도구입니다. Linear는 AI 에이전트를 단순히 작업을 수행하는 도구로 보는 것을 넘어, 제품 개발이라는 복잡한 여정에서 인간과 협력하는 ‘일등 시민’으로 대하며, 신중함과 이해를 바탕으로 진정한 혁신을 이끌어 나가고 있습니다. 이는 AI 시대에 모든 기업이 고민해야 할 중요한 질문과 통찰을 던져줍니다.
“23 AI Trends keeping me up at night” — Greg Isenberg 기반 기사 원본 영상 보기
AI가 선사하는 잠 못 이루는 밤: 지금 당장 뛰어들어야 할 23가지 혁신 기회
최근 몇 년간 인공지능(AI)의 발전은 전례 없는 속도로 진행되며 수많은 사람의 밤잠을 설치게 하고 있습니다. 이는 단순히 기술적 경이로움을 넘어, 비즈니스 모델, 창업 방식, 그리고 우리가 일하고 살아가는 방식 자체를 근본적으로 뒤흔드는 거대한 변화의 물결입니다. 유명 투자자이자 창업가인 그렉 아이젠버그(Greg Isenberg) 또한 이러한 AI 트렌드에 대한 깊은 고민과 함께 엄청난 기회에 대한 흥분으로 밤잠을 설친다고 고백합니다. 그는 AI가 가져올 미래에 대한 두려움과 동시에, 지금 이 순간이야말로 역사상 가장 비대칭적인(Asymmetric) 기회의 창이 열려 있다고 강조합니다.
이 글은 그렉 아이젠버그가 말하는 AI 시대의 핵심 트렌드와 그에 따른 기회, 그리고 잠재적 위험 요소를 심층적으로 분석하여, 현재의 변화를 이해하고 미래를 준비하려는 독자들에게 실질적인 통찰을 제공하고자 합니다.
I. 초고속 비즈니스 구축의 시대: ‘1시간 기업’과 ‘앰비언트 비즈니스’
AI 시대의 가장 눈에 띄는 변화 중 하나는 비즈니스 구축 속도의 혁신적인 단축입니다. 과거에는 아이디어를 구체화하고 MVP(최소 기능 제품)를 개발하며 첫 고객을 확보하기까지 수개월에서 1년 이상의 시간이 필요했습니다. 하지만 이제는 모든 것이 하루, 심지어 몇 시간 안에 가능해졌습니다.
아이젠버그는 이를 **‘1시간 기업 스택(1-Hour Company Stack)‘**이라고 부릅니다. 아이디어 브라우저(ideabrowser.com) 같은 플랫폼에서 검증된 아이디어를 얻고, 아이디어를 즉시 코드로 구현하는 ‘바이브 코딩(Vibe Coding)’ 도구를 활용해 제품을 만들며, 랜딩 페이지를 구축하고, 결제 시스템(Stripe)을 연동하여 몇 시간 안에 첫 고객을 맞이할 수 있게 된 것입니다. 클로드 코드(Claude Code), 코덱스(Codeex), 구글 AI 스튜디오(Google AI Studio)와 같은 **에이전트 엔지니어링 플랫폼(Agent Engineering Platform)**의 발전 덕분입니다. 이는 과거 개발자를 고용하고 MVP를 만드는 데 수개월이 걸리던 구식 타임라인을 완전히 뒤집어 놓았습니다. 단, 이러한 초고속 비즈니스 구축이 성공하려면 사전에 확보된 이메일 리스트나 타겟 오디언스 같은 ‘유통 채널(Distribution)‘이 필수적입니다. AI를 활용한 유통 채널 구축 또한 밤잠을 설치게 하는 중요한 주제입니다.
나아가 아이젠버그는 ‘앰비언트 비즈니스(Ambient Business)’ 또는 **‘자율 비즈니스(Autonomous Business)‘**의 등장을 예고합니다. 이는 인간의 개입이 거의 또는 전혀 없이 운영되는 사업 모델을 의미합니다. AI 에이전트들이 시장을 모니터링하고, 기회를 식별하며, 스스로 실행하고, 고객 서비스를 처리하는 등 사업의 거의 모든 측면을 자율적으로 관리합니다. 창업자는 며칠에 한 번씩 사업을 확인하는 정도로 충분하며, 이러한 자율 비즈니스가 연간 78자릿수(수백억수천억 원)의 수익을 창출할 수 있는 시대가 올 것이라고 전망합니다. 현재는 초기 단계이며 ‘AI 슬롭(AI Slop)‘이라 불리는 저품질 결과물을 내는 경우가 많지만, 진보의 화살표는 명확히 이 방향을 향하고 있습니다.
II. 에이전트 경제의 도래: ‘에이전트 to 에이전트’ 비즈니스
기술 경제의 진화는 앱 스토어 시대(2009-2015년)에서 API 경제(2015-2024년)를 거쳐 이제 **‘에이전트 경제(Agent Economy)‘**의 시대로 접어들고 있습니다(2025-2030년). 앱 스토어 시대에는 사람들이 앱을 다운로드하고 인간이 직접 조작했으며, API 경제에서는 개발자들이 API를 연결하여 서비스를 구축했습니다. 하지만 에이전트 경제에서는 AI 에이전트들이 다른 에이전트들을 스스로 발견하고 고용하여 작업을 수행하게 됩니다. 이는 고정된 업무(Fixed Tasks)의 개념을 해체시킬 것입니다.
아이젠버그는 이러한 에이전트 경제에서 엄청난 스타트업 기회를 발견합니다. 예를 들어, **‘AI 에이전트판 글래스도어(Glassdoor)‘**를 만드는 것입니다. 에이전트들의 평판을 구축하고, 누가 어떤 에이전트를 고용해야 할지 알려주는 마켓플레이스 말입니다. 이는 과거 소셜 네트워크로 메타(Meta)에 2억 달러에 인수된 ‘몰드북(Moldbook)’ 같은 사례를 AI 에이전트 영역에서 재현하는 것과 같습니다. 가트너(Gartner)의 예측에 따르면, 2030년까지 전체 상거래의 20%가 에이전트 대 에이전트(Agent-to-Agent), 즉 기계 대 기계(Machine-to-Machine) 방식으로 이루어질 것이라고 합니다.
현재 마켓플레이스에는 3만 1천 개 이상의 에이전트 스킬이 존재하지만, 대부분은 품질이 좋지 않습니다. 이는 곧 고품질의 스킬과 에이전트를 구축할 수 있는 엄청난 기회가 있다는 뜻입니다. CEO 에이전트, 영업 에이전트, 개발 에이전트, 마케팅 에이전트 등 다양한 역할을 수행하는 에이전트들이 서로를 고용하고 관리하는 세상이 올 것입니다. 그는 오픈 소스 기술인 ‘페이퍼클립(Paperclip)‘을 예로 들며, 에이전트들이 하위 작업을 생성하고 완료되면 종료하는 **‘서버리스 함수(Serverless Function)‘**와 같은 조직도를 형성할 것이라고 설명합니다. 이는 인간에게 ‘작업을 수행해달라’고 요청하는 대신, 작업을 완료할 에이전트들을 고용하고 관리하게 될 것임을 의미합니다.
III. 버티컬 AI의 거대한 물결: 숨겨진 금광을 찾아서
YC(Y Combinator)는 향후 10년 동안 버티컬 AI(Vertical AI) 분야에서 300개 이상의 유니콘 기업이 탄생할 것이라고 예측했습니다. 이는 특정 산업이나 틈새시장에 특화된 AI 솔루션을 의미합니다. 과거 컨스텔레이션 소프트웨어(Constellation Software)가 다양한 버티컬(Vertical) 산업에서 수백 개의 SaaS(Software as a Service) 기업을 인수한 것처럼, 이제는 버티컬 AI 컨스텔레이션 소프트웨어를 구축할 기회가 왔습니다.
아이젠버그는 버티컬 SaaS와 버티컬 AI의 차이점을 명확히 설명하며 후자의 잠재력을 강조합니다.
| 구분 | 버티컬 SaaS (Vertical SaaS) | 버티컬 AI (Vertical AI) |
|---|---|---|
| IT 예산 | IT 예산의 일부를 차지 | 기업의 인건비 손익계산서(Labor P&L)에 직접적으로 영향을 미침 |
| 판매 대상 | 소프트웨어 라이선스 판매 | ’소프트웨어로서의 에이전트(Agent as a Software)‘를 통해 결과 판매 |
| 사용 주체 | 인간이 도구를 조작 | 에이전트가 직접 작업을 수행 |
| 성과 규모 | 일반적으로 1천만~1억 달러 규모의 기업 | 평균적으로 더 큰 규모의 성과 기대 (인건비 대체로 10배 큰 TAM) |
즉, 버티컬 AI는 단순히 IT 예산을 차지하는 것을 넘어, 인력을 대체함으로써 **총 유효 시장(Total Addressable Market, TAM)**을 10배 이상 확장할 수 있습니다. 이는 AI가 인간이 수행하던 업무를 직접 처리하기 때문에 가능합니다.
그렇다면 어떤 ‘지루한 금광(Boring Goldmine)’ 버티컬에 주목해야 할까요? 아이젠버그는 여전히 전화 통화나 팩스 같은 구식 시스템에 의존하는 산업을 주목하라고 조언합니다. 보험(30년 된 보험 통계표 사용), 법률, 물류, 노인 돌봄(Elder Care), 정부, 회계, 건설 등이 대표적입니다. 중요한 것은 이러한 큰 카테고리 내에서 **‘매우 구체적인 서브 니치(Sub-Niche)‘**를 공략하고, 규제나 진입 장벽(Red Tape)이 적은 분야를 선택하는 것입니다. “지루할수록 좋고, 틈새시장일수록 좋다”는 것이 그의 조언입니다.
IV. 가격 모델의 혁신: 결과 기반(Outcome-Based) 모델로의 전환
SaaS 가격 모델은 좌석당 라이선스(Per-Seat Licensing)에서 사용량 기반(Usage-Based)을 거쳐 이제 결과 기반(Outcome-Based) 모델로 진화하고 있습니다. 이는 에이전트가 직접 작업을 수행하기 때문에 가능해진 변화입니다. 가트너(Gartner)는 2030년까지 기업용 SaaS의 40%가 결과 기반 가격 모델로 전환될 것이며, 좌석당 라이선스 모델은 21%에서 15%로 감소할 것이라고 예측합니다.
결과 기반 모델은 고객이 실제로 얻은 결과에 대해서만 비용을 지불하는 방식입니다. 예를 들어, 월 100달러의 좌석당 요금을 내고 사용 여부와 관계없이 지불하는 대신, 해결된 고객 문의 티켓당 1.50달러를 지불하는 식입니다. 이는 젠데스크(Zendesk)와 같은 대기업에서도 이미 도입하고 있으며, AI 네이티브 SaaS 기업의 83%가 이미 이러한 모델로 전환했습니다.
아이젠버그는 여기서 두 가지 큰 기회를 봅니다. 첫째, 기존의 레거시 SaaS 기업들이 결과 기반 가격 모델로 전환하도록 돕는 컨설팅 및 솔루션 사업입니다. 둘째, 아예 처음부터 결과 기반으로 설계된 AI 네이티브 스타트업을 구축하는 것입니다. 시장 선점자가 되면 고객에게 더 매력적인 제안을 할 수 있습니다.
이러한 변화는 기존 SaaS 기업들에게 위협이 될 수 있으며, 일부는 **‘SaaS의 무덤(SAS Graveyard)‘**으로 사라질 것입니다.
- 사라질 비즈니스:
- 범용 CRM(Generic CRM): 에이전트가 더 나은 방식으로 고객 관계를 관리할 수 있기 때문입니다. (세일즈포스나 허브스팟 같은 선두 기업들은 이미 AI로 전환하고 있습니다.)
- 기본적인 분석 대시보드: AI가 즉시 통찰력을 생성할 수 있기 때문입니다.
- 템플릿 마켓플레이스: AI가 맞춤형 템플릿을 즉석에서 생성할 수 있기 때문입니다.
- 스케줄링 도구: 에이전트가 캘린더를 기본적으로 처리할 수 있기 때문입니다.
- 기본적인 고객 지원 챗봇: AI가 이미 이를 대체하고 있기 때문입니다.
- 살아남을 비즈니스:
- 에이전트 기업으로 전환하는 버티컬 워크플로우 도구
- 인프라 및 데이터 모드(Data Moats)를 가진 기업
V. 희소성의 역전: 인간적 가치의 재발견
AI가 코드, 일반 콘텐츠, 기본 디자인, 데이터 입력, 일상적인 분석 등을 상품화(Commoditize)하면서, ‘희소성의 역전(Scarcity Flip)’ 현상이 발생하고 있습니다. 과거에는 희소했던 것들이 흔해지고, 과거에는 흔했던 것들이 희소해지는 것입니다. 가치는 ‘실행(Execution)‘에서 **‘판단(Judgment)‘**으로 이동할 것입니다.
- AI가 상품화하는 것: 일반적인 코드, 콘텐츠, 기본적인 디자인, 데이터 입력, 반복적인 분석.
- 희소해지고 프리미엄이 되는 것:
- 창의적인 판단(Creative Judgment): 독창적이고 예측 불가능한 사고.
- 인간이 만든 공예품(Human-Made Crafts) 및 물리적 경험(Physical Experiences): 디지털이 무한하고 AI가 생성하는 세상에서, 인간과의 물리적 상호작용은 더욱 가치 있게 됩니다. 노래방, 방탈출 게임, 몰입형 극장, 코워킹 스페이스, 라이브 음악 등 ‘경험 경제(Experience Economy)‘는 더욱 가속화될 것입니다.
- 독창적인 기이한 사고(Original Weird Thinking): AI는 아직 ‘기이함’을 잘 모방하지 못합니다. 자신만의 독특한 관점과 경험에서 나오는 ‘이상함’이 오히려 가치를 가질 것입니다.
- 독점 데이터(Proprietary Data): 특정 분야의 독점적인 데이터는 여전히 강력한 해자(Moat)가 될 것입니다.
아이젠버그는 **‘프리미엄 스택(Premium Stack)‘**의 개념을 제시합니다.
- 최고의 프리미엄: 인간이 만든(Human-Made) - 포르쉐(Porsche)가 100% 인간이 만든 광고 캠페인을 통해 ‘AI 무관(AI-free)’ 로고를 내세웠던 사례처럼, 명품 브랜드는 ‘AI 무관’을 인증 라벨(예: 유기농 식품의 ‘유기농 인증’)처럼 활용할 것입니다.
- 프리미엄: AI의 도움을 받지만 인간이 주도하는(AI-Assisted but Human-Led) - 인간의 취향과 AI의 속도가 결합된 형태입니다.
- 상품(Commodity): 완전히 AI 서비스(Fully AI Service) - 가격 경쟁이 심화되어 ‘제로 가격(Race to Zero Pricing)‘으로 향할 가능성이 큽니다.
VI. 미래의 창업가 역량: ‘파운더-에이전트 적합성’과 ‘고스트 팀’
과거 실리콘밸리에서는 창업가가 고객과 시장을 얼마나 잘 이해하는지, 즉 **‘파운더-시장 적합성(Founder-Market Fit)‘**을 강조했습니다. 하지만 AI 시대에는 **‘파운더-에이전트 적합성(Founder-Agent Fit)‘**이 더욱 중요해질 것입니다. 이는 창업가가 자신의 목표를 향해 에이전트들을 얼마나 잘 조율하고 지휘할 수 있는지를 의미합니다.
아이젠버그는 이를 영화감독에 비유합니다. 영화감독은 카메라를 들거나 직접 연기하거나 음악을 작곡하지 않습니다. 대신 배우들로부터 최고의 연기를 이끌어냅니다. 이제 그 ‘배우’의 역할이 사람에서 기계(에이전트)로 바뀌는 것입니다. 특정 니치(Niche)에서 에이전트를 구축하고, 관리하며, 최대한의 성과를 이끌어내는 능력은 창업가에게 비대칭적인 우위(Unfair Advantage)를 제공할 것입니다.
이러한 변화는 기업의 조직도(Org Chart)에도 혁명적인 변화를 가져올 것입니다. 미래의 기업은 **‘고스트 팀(Ghost Team) 조직도’**를 가질 수 있습니다. 팀 페이지에 몇 명의 인간과 수많은 AI 에이전트(영업 에이전트, 콘텐츠 에이전트, 고객 지원 에이전트 등)가 나란히 소개되는 것입니다. 이 에이전트들은 이름과 개성을 가질 수 있으며, 심지어 이미지를 가질 수도 있습니다. 궁극적으로는 사람처럼 비디오 채팅을 하거나 음성 메시지를 보내는 등 인간과 거의 유사한 방식으로 상호작용하게 될 것입니다. 아이젠버그는 이러한 AI 네이티브 에이전트 비즈니스를 여러 개 소유하는 지주회사(Holding Company) 모델이 더욱 많아질 것이라고 예측합니다.
VII. 마이크로 독점 시대: ‘100명의 진정한 팬’과 비대칭적 기회
케빈 켈리(Kevin Kelly)의 ‘1000명의 진정한 팬(Thousand True Fans)’ 개념은 AI 시대에 **‘100명의 진정한 팬(Hundred True Fans)‘**으로 축소될 수 있습니다. AI 에이전트가 운영 비용을 극적으로 절감하기 때문에, 100명의 고객만으로도 충분히
""We’re Not Writing Code by Hand Anymore. That’s Over.” | Owen Jennings & David Haber - The a16z Show” — a16z 기반 기사 원본 영상 보기
손으로 코딩하는 시대는 끝났다: 블록(Block)의 40% 인력 감축, AI가 이끈 파격 변신과 미래 전략
기술 산업 전반에 걸쳐 인공지능(AI)의 영향력이 심화되고 있는 가운데, 핀테크 기업 블록(Block)이 단행한 40%에 달하는 대규모 인력 감축은 단순한 비용 절감을 넘어선 근본적인 변화의 신호탄으로 받아들여지고 있습니다. “더 이상 손으로 코딩하지 않는다. 그 시대는 끝났다.” 블록의 비즈니스 리드인 오웬 제닝스(Owen Jennings)의 이 대담한 발언은 AI가 기업의 운영 방식과 생산성 개념을 어떻게 재정의하고 있는지를 단적으로 보여줍니다.
최근 a16z 쇼에 출연한 제닝스는 스퀘어(Square), 캐시 앱(Cash App), 애프터페이(Afterpay) 등 블록의 주요 사업 전반에 걸쳐 AI 전환을 주도한 경험을 공유하며, 블록이 이 파격적인 결정을 내리게 된 배경과 그로 인해 발생한 조직 및 제품 혁신, 그리고 미래 전략에 대해 심도 깊은 통찰을 제시했습니다.
40% 인력 감축: AI가 이끈 ‘이진적 변화’의 결과
블록의 대규모 인력 감축은 단순히 과잉 고용에 대한 반작용이 아니었습니다. 제닝스는 이 결정의 뿌리가 2~3년 전 잭 도시(Jack Dorsey) CEO의 선견지명에서 시작되었다고 설명합니다. 도시는 일찍이 ‘에이전트 기반 개발(agentic development)‘의 중요성을 인지하고 있었고, 블록은 2024년 초 첫 에이전트 하네스(agent harness)인 ‘구스(Goose)‘를 출시하며 소프트웨어 개발 방식을 혁신하기 시작했습니다.
결정적인 전환점은 2023년 11월 말에서 12월 초에 찾아왔습니다. 당시 ‘오푸스 46(Opus 46)‘과 ‘코덱스 53(Codex 53)‘과 같은 도구와 기반 모델들이 기존의 복잡한 코드베이스와도 놀랍도록 잘 작동하기 시작하면서, 기술의 발전은 ‘이진적 변화(binary change)‘를 맞이했습니다. 제닝스는 “수십 년 동안 기업 내 인력 수와 생산량 사이에는 상관관계가 있었지만, 12월 첫째 주에 그 관계는 근본적으로 깨졌다”고 단언합니다. 이제 1~2명의 엔지니어 또는 디자이너와 엔지니어 한 명이 도구를 활용하면 10배, 20배, 심지어 100배 더 높은 생산성을 낼 수 있게 된 것입니다.
이러한 변화는 블록의 인력 감축이 단순히 재정적 동기에서 비롯된 것이 아님을 시사합니다. 제닝스는 “만약 단순한 과잉 고용이나 비효율성 때문이었다면, 운영팀에서 훨씬 더 큰 감축이 있었을 것”이라며, 실제로 개발 부문에서 가장 큰 규모의 감축이 이루어졌다는 점을 강조했습니다. 이는 블록이 기술과 도구가 근본적으로 변화했음을 인식하고, ‘더 이상 손으로 코딩하지 않는’ 새로운 시대에 맞춰 조직을 재편했음을 의미합니다.
AI 전환의 실행과 조직 문화의 재편
블록은 인력 감축을 실행하는 과정에서 몇 가지 핵심 원칙을 세웠습니다. 첫째, **안정성(reliability)**을 최우선으로 하여 대규모 조직 변화에도 불구하고 시스템 장애가 발생하지 않도록 했습니다. 둘째, **고객과의 신뢰 구축 및 규제 준수(compliance)**를 비협상 불가능한 원칙으로 삼아, 규제 환경이 복잡한 핀테크 산업의 특성을 고려했습니다. 실제로 규제 관련 팀은 거의 영향을 받지 않았습니다. 셋째, **지속 가능한 성장(durable growth)**을 위해 기존 로드맵의 기능을 계속 구축하고 장기적인 투자도 이어갔습니다.
이러한 원칙 아래 블록은 조직을 ‘제로 베이스’에서 다시 구축했습니다. 특히 개발 부문은 완전히 새로운 모습으로 바뀌었지만, 규제나 영업과 같은 일부 영역은 이전과 유사한 형태를 유지했습니다. 인력 감축 과정에서도 블록은 퇴직 패키지 제공에 관대했으며, 기술 접근을 즉시 차단하지 않고 모든 직원에게 결정의 배경을 투명하게 설명하는 전사 회의를 가졌습니다.
인력 감축 이후 블록의 업무 방식은 극적으로 변화했습니다. 회의는 7080%가량 줄어들어 직원들이 ‘빌딩(building)‘에 집중할 시간을 확보했습니다. 또한, 잭 도시를 비롯한 경영진과 매주 12시간의 전사 회의를 통해 소통을 강화했습니다. 제닝스는 “더 작고, 더 효율적이며, 계층이 적고, 더 넓은 관리 범위를 갖게 되었다”며, “다시 빌딩에 집중하는 분위기”라고 전했습니다. AI는 단순히 도구가 아니라 조직의 변화를 강제하는 ‘강제 기능(forcing function)‘으로 작용하며, 선형적인 워크플로우에서 벗어나 여러 AI 에이전트를 관리하는 방식으로 업무가 진화하고 있음을 보여줍니다.
AI 기반 인프라와 제품의 혁신
블록의 AI 전환은 내부 인프라와 외부 제품 모두에 걸쳐 광범위하게 이루어졌습니다.
내부 시스템 및 조직 변화:
- 에이전트 인프라: ‘구스(Goose)‘는 모델에 구애받지 않는(model agnostic) 에이전트 하네스로, 다양한 AI 모델(Anthropic, OpenAI, 오픈소스 모델 등 120여 개)을 활용할 수 있게 합니다. 내부 에이전트 운영 체제인 ‘G2’는 모든 결정론적 워크플로우(deterministic workflow)를 자동화할 수 있습니다.
- 개발 자동화: ‘빌더봇(Builderbot)‘은 자율적으로 PR(Pull Request)을 병합하고 기능을 85~100%까지 개발하는 데 활용됩니다. 이는 아이디어에서 수십만 고객에게 도달하는 시간이 획기적으로 단축되었음을 의미합니다.
- 조직 구조 개편: 기존의 계층적, 기능별 구조에서 벗어나 계층을 50
60% 줄이고, 16명으로 구성된 소규모 스쿼드(squad) 중심으로 유연하게 운영됩니다. 이는 정보 흐름을 원활하게 하고, 팀들이 여러 제품에 유연하게 기여할 수 있도록 합니다. - 개발 외 영역 자동화: 고객 지원 챗봇, AI 전화 지원은 문의의 대부분을 자동 처리하며, 제품 운영, 리스크 운영, 규제 준수 운영 등 결정론적 워크플로우 전반에 걸쳐 AI가 인간보다 더 나은 결정을 내릴 수 있음을 보여줍니다. 현재는 ‘인간 개입(Human-in-the-loop)‘이 중요하지만, 장기적으로는 시스템이 인간보다 훨씬 뛰어날 것이라는 전망입니다.
제품 혁신과 생성형 UI: 블록은 과거 사업부 중심(스퀘어, 캐시 앱, 애프터페이 각각 별도 CEO)에서 벗어나, 18개월 전부터 엔지니어링, 디자인, 제품 등 핵심 기능을 통합하는 ‘기능별 조직(functionalized company)‘으로 전환했습니다. 이는 브랜드에 구애받지 않는 기술 인프라를 구축하고, 스퀘어, 캐시 앱, 애프터페이 전반을 연결하는 생태계 전략의 일환입니다.
- AI 기반 제품: 캐시 앱의 ‘머니봇(Moneybot)‘은 사용자를 위한 ‘주머니 속 CFO’처럼 작동하며, 스퀘어의 ‘매니저봇(Managerbot)‘도 유사한 기능을 제공합니다. 이들은 구스(Goose) 플랫폼 위에 구축되어 있습니다.
- 생성형 사용자 인터페이스(Generative UI): 제닝스는 향후 6개월 내에 앱의 사용자 인터페이스(UI)가 근본적으로 변할 것이라고 예측합니다. 기존의 정적이고 고정된 UI와 달리, 사용자의 앱은 개인화되어 다르게 보일 것입니다. 더 나아가, 머니봇은 사용자의 지출 패턴을 분석하여 실시간으로 차트나 시각화를 생성해 보여주며, 매니저봇은 식당 주인이 스케줄 관리 앱을 요청하면 그 자리에서 앱을 생성해 주는 등 ‘코드에 없는’ 동적인 UI를 제공합니다.
- 선제적 인텔리전스(Proactive Intelligence): 고객이 직접 AI 도구에 적절한 프롬프트를 입력하기 어려울 수 있음을 고려하여, 블록은 고객에게 의미 있는 제안을 ‘선제적으로’ 제공하는 데 집중하여 가치를 창출하고 있습니다.
블록(Block)의 해자(Moat)와 AI 시대의 미래
블록의 주가는 지난 몇 년간 정체되어 있었지만, 제닝스는 시장의 주기성을 언급하며 장기적인 관점에서 ‘빌딩(building)‘에 집중하고 있다고 말합니다. 그는 AI 시대에 블록의 경쟁 우위, 즉 ‘해자(moat)‘를 다음과 같이 설명했습니다.
-
단기 및 중기 해자:
- 유통 및 네트워크 효과: 캐시 앱처럼 수천만 명의 월간 활성 사용자를 보유한 플랫폼은 단기간에 ‘바이브 코딩(vibe coded)‘될 수 없습니다.
- 라이선스 및 규제: 핀테크 산업의 복잡한 라이선스와 규제 환경은 진입 장벽으로 작용합니다.
- 하드웨어: 스퀘어의 하드웨어처럼 물리적인 제품은 AI 도구만으로는 쉽게 복제될 수 없습니다.
-
장기적 해자:
- ‘이해하기 매우 어려운 무언가를 이해하는 능력’: 가장 중요한 장기적 해자는 다른 기업들이 쉽게 파악하기 어려운 심층적인 통찰력을 보유하는 것입니다. 블록의 경우, 판매자와 구매자가 경제 활동에 참여하는 방식에 대한 깊은 이해가 여기에 해당합니다.
- 지능형 시스템으로서의 블록: 블록은 고객을 이해하고 블록 자체의 운영 방식을 이해하는 ‘세계 모델(world models)‘을 구축하여 ‘지능형 시스템’으로 진화하고 있습니다. 이는 풍부한 데이터와 심층적인 통찰력을 기반으로 합니다.
- 피드백 루프: 기업은 ‘신호(signal)‘(깊이 이해하는 것)와 ‘도구(tool)‘(빌더봇 등)를 연결하여 이해도를 지속적으로 개선하는 피드백 루프를 반복해야 합니다. 제닝스는 과거 몇 달이 걸리던 기능 개발이 이제 1~2주로 단축되었고, 미래에는 이 루프가 하루에 수백, 수천 번 실행될 수 있으며 인간은 ‘편집자(editors)’ 역할로 전환될 것이라고 예측합니다.
제닝스는 이러한 변화가 전 세계 엔지니어, 디자이너, PM의 수를 반드시 줄이지는 않을 것이라고 덧붙였습니다. 마치 제번스의 역설(Jevons Paradox)처럼, 효율성이 증가하면 소비가 늘어나듯이, AI로 인해 ‘더 많은 것’을 만들 수 있는 가능성이 열려 새로운 기술 기업들이 생겨나거나 기존에 기술이 적용되지 않던 분야로 확산될 수 있다는 전망입니다.
결론: AI 시대, ‘이해’가 곧 경쟁력이다
블록의 사례는 인공지능이 더 이상 단순한 생산성 도구가 아니라, 기업의 조직 구조, 운영 방식, 제품 개발, 그리고 궁극적으로는 존재 방식 자체를 근본적으로 재정의하는 시대가 도래했음을 명확히 보여줍니다. 40%의 인력 감축은 고통스러운 결정이었지만, 이는 AI가 이끄는 새로운 패러다임에 맞춰 선제적으로 변화하려는 블록의 대담한 의지를 반영합니다.
미래 기업의 성공은 AI를 얼마나 잘 활용하여 ‘다른 사람들이 이해하기 매우 어려운 무언가를 심층적으로 이해’하고, 이를 기반으로 끊임없이 혁신하는 능력에 달려 있을 것입니다. 만약 그 질문에 대한 답이 “모르겠다”라면, 제닝스의 경고처럼 당신의 기업은 AI가 만들어내는 새로운 흐름 속에서 ‘감각적으로 밀려날(vibe coded away)’ 수 있다는 점을 명심해야 할 것입니다. 블록의 파격적인 변신은 AI 시대의 새로운 조직 모델과 경쟁 전략에 대한 중요한 질문을 우리 모두에게 던지고 있습니다.
“Today’s Mission to the Moon” — New York Times Podcasts 기반 기사 원본 영상 보기
달을 넘어, 인류의 새로운 시대를 열다: 아르테미스 2호 임무의 심층 분석
인류가 달에 발자국을 남긴 지 반세기 이상이 지난 지금, 미국은 다시 한번 달을 향한 원대한 여정을 시작합니다. 단순한 과거의 재현이 아닌, 인류의 미래를 재편할 혁신적인 비전을 품고서 말이죠. 뉴욕타임스 팟캐스트는 아르테미스(Artemis) 프로그램의 핵심 단계인 아르테미스 2호 임무를 심층적으로 다루며, 이 역사적인 비행이 왜 지금 다시 시작되는지, 그리고 그 너머에 어떤 꿈이 숨겨져 있는지 조명했습니다.
왜 다시 달인가? 아르테미스 프로그램의 궁극적 목표
1969년 아폴로 11호의 달 착륙은 인류에게 큰 발자국을 남겼지만, 그 이후 달은 오랫동안 잊힌 존재였습니다. 하지만 아르테미스 프로그램은 단순히 과거의 영광을 되찾는 것을 넘어, 달에 영구적인 거주지를 건설하고, 궁극적으로는 화성 탐사를 위한 전초 기지로 활용하겠다는 야심 찬 목표를 가지고 있습니다.
아르테미스 프로그램은 여러 단계로 나뉘어 진행됩니다.
- 아르테미스 1호(Artemis 1): 2022년 발사된 무인 시험 비행으로, 우주선을 달 궤도에 보내 수 주간 머무르게 하며 기본적인 장비와 시스템의 작동 여부를 확인했습니다.
- 아르테미스 2호(Artemis 2): 현재 진행될 임무로, 유인 비행을 통해 달 궤도를 선회한 후 지구로 귀환합니다. 이 임무의 가장 중요한 목표는 바로 우주 비행사들이 탑승한 상태에서 생명 유지 시스템(life support systems)이 제대로 작동하는지 시험하는 것입니다. 우주 공간에서는 이산화탄소, 물, 배설물 등 인간 활동으로 발생하는 모든 것을 처리해야 하므로, 실제 사람이 탑승하여 시스템을 검증하는 것이 필수적입니다.
- 아르테미스 3호(Artemis 3): 아르테미스 2호가 성공하면, 몇 년 안에 우주 비행사들을 달 표면에 착륙시키는 것을 목표로 합니다.
궁극적으로 미국 항공우주국(NASA)은 달에 기지를 건설하여, 일회성 방문이 아닌 장기 체류가 가능한 환경을 조성하려 합니다. 이는 과학 연구를 위한 전초 기지이자, 달의 자원을 활용하기 위한 첫걸음이 될 것입니다.
미지의 달을 향한 여정: 아르테미스 2호의 임무와 승무원
아르테미스 2호 임무에는 네 명의 우주 비행사가 탑승합니다. 이들은 인류의 새로운 달 탐사 시대를 여는 중요한 역할을 맡게 됩니다.
- 리드 와이즈먼(Reid Wiseman) 사령관: 전직 미 해군 전투기 조종사이자 중동 지역에 두 차례 파병된 베테랑입니다. 우주비행사 사무실 책임자를 역임했으며, 그의 아내가 몇 년 전 사망한 후 두 딸을 10일간 남겨두고 달로 향하는 개인적인 사연도 안고 있습니다.
- 빅터 글로버(Victor Glover): 역시 전직 해군 조종사로, 국제 우주정거장(ISS)에 장기 체류한 최초의 흑인 남성입니다. 이제 그는 달로 향하는 최초의 흑인 남성이라는 역사를 쓰게 됩니다.
- 크리스티나 코크(Christina Koch): 전기 기술자 출신으로, NASA 우주비행사로 선발되기 전 지상에서 NASA 임무를 수행했습니다. 여성으로서 단일 우주 비행 최장 기록인 328일을 보유하고 있습니다.
- 제레미 핸슨(Jeremy Hansen): 캐나다 출신으로, 미국인이 아닌 사람으로서는 최초로 심우주(deep space)로 향하는 인물이 됩니다. 그는 임무 중 문제가 발생하면 “NASA는 캐나다를 탓할 수 있을 것”이라며 유머 감각을 보여주기도 했습니다.
이들은 발사 8시간 전 기상하여 우주복을 착용하고 발사대로 이동, 오리온(Orion) 우주선에 탑승한 후 4시간의 카운트다운을 기다립니다. 발사 8분 후 우주에 진입하며, 첫째 날에는 곧바로 달로 향하지 않고 지구 궤도를 두 바퀴 돌며 우주선의 모든 시스템을 꼼꼼히 점검합니다. 이는 달로 향하기 전 마지막 안전 확인 절차입니다.
둘째 날 엔진을 점화하여 달로 향하며, 지구에서 약 25만 마일(약 40만 km) 떨어진 달까지는 약 4일이 소요됩니다. 이 기간 동안 우주 비행사들은 미니밴 두 대 크기의 오리온 캡슐 안에서 생활합니다. 지구에서라면 좁게 느껴질 공간이지만, 무중력 상태에서는 위아래 공간을 모두 활용할 수 있어 육상에서보다는 덜 답답할 것이라고 합니다.
임무 6일째, 이들은 달에 가장 가까이 접근하는 지점을 통과합니다. 이때 달은 팔을 뻗었을 때 농구공만 한 크기로 보일 것이라고 합니다. 달의 중력은 우주선을 달 주위로 끌어당겨 뒷면으로 이동하게 하는데, 이 40분 동안은 지구와의 모든 무선 통신이 두절됩니다. 이들은 이 시간 동안 달의 뒷면을 관측하게 되는데, 특히 인류 역사상 단 한 번도 인간의 눈으로 직접 목격되지 않았던 달의 뒷면 일부 지역을 보게 될 것입니다. 이는 과거 아폴로 우주 비행사들이 달의 뒷면을 지났을 때는 해당 지역이 어둠 속에 있었기 때문입니다.
달의 중력은 우주선을 지구로 되돌리는 슬링샷(slingshot) 효과를 제공하므로, 우주 비행사들은 엔진을 거의 사용하지 않고도 안전하게 지구로 귀환할 수 있습니다. 달 궤도 비행을 마친 후에는 다시 3일간 지구로 돌아오는 여정을 보내게 되며, 이는 임무 중 가장 지루한 부분이 될 것이라고 합니다. 마지막 날, 지구의 중력이 우주선을 대기권으로 끌어당기면, 샌디에이고 인근 태평양 해상에 착수(splashdown)하며 10일간의 임무를 마치게 됩니다.
달 기지 건설을 넘어: 인류의 새로운 미래를 꿈꾸다
아르테미스 프로그램은 단순한 달 탐사를 넘어 인류의 미래에 대한 거대한 비전을 제시합니다.
- 과학 연구 기지: 달 기지는 처음에는 남극의 연구 기지처럼 과학자들이 달에 장기 체류하며 심도 있는 연구를 수행하는 장소가 될 것입니다.
- 자원 채굴: 달 자원의 활용 가능성에 대한 논의도 활발합니다. 특히 ‘헬륨-3(Helium-3)‘는 지구에서는 매우 희귀하지만 달 표면에 상대적으로 풍부하게 존재합니다. 헬륨-3는 미래 핵융합 발전(fusion reactors)의 핵심 연료로 활용될 수 있으며, 양자 컴퓨터(quantum computers)나 인공지능(AI) 기술 개발에도 유용할 것으로 기대됩니다. 현재 지구에서 헬륨-3 1파운드(약 450g)는 약 3백만 달러에 달하는 가치를 가지고 있어, 달에서 소량만 채굴해도 상당한 수익을 창출할 수 있습니다.
- 심우주 관측소: 달의 뒷면에는 지구의 TV, 휴대폰, 팟캐스트 등 온갖 전파 소음이 차단된 ‘죽은 듯이 고요한’ 환경이 존재합니다. 이곳에 거대한 전파 망원경을 건설하면, 수십억 년 전 빅뱅(Big Bang) 직후의 미세한 우주 신호를 포착하여 우주의 기원을 연구하는 데 획기적인 도움이 될 수 있습니다. 이는 사실상 ‘태초의 소리’를 듣는 것과 같은 경험이 될 것입니다.
- 화성 탐사 시험장: 달은 화성 탐사를 위한 기술 시험장으로도 활용될 것입니다. 달의 환경과 중력은 화성과 유사한 점이 많으므로, 핵 발전소, 거주지, 생명 유지 시스템 등 화성에서 필요한 기술들이 달에서 먼저 테스트될 것입니다.
- 다행성 종족의 꿈: 인류가 하나의 행성에만 국한되지 않고 우주로 뻗어나가는 ‘다행성 종족(multi-planet species)‘이 되겠다는 낭만적인 비전도 있습니다. 언젠가 달에 인류의 식민지가 건설될 수도 있습니다.
- 지정학적 경쟁: 이 모든 비전 뒤에는 중국과의 지정학적 경쟁이라는 현실적인 동기도 있습니다. 먼저 달에 도착하여 거점을 확보하는 국가가 우주 상거래 규칙을 정하고 핵심 자원을 통제할 수 있는 주도권을 쥐게 될 것이기 때문입니다.
옛 방식의 승리, 그리고 새로운 시대의 도래
아르테미스 2호 임무는 여러모로 ‘올드 스쿨(old school) NASA’의 방식이 성공했음을 보여주는 중요한 시험대가 될 것입니다. 우주선 설계부터 운영까지 NASA가 주도적으로 진행하는 방식은 현대 우주 산업에서 민간 기업의 역할이 커지는 추세와는 대조적입니다.
하지만 아르테미스 3호부터는 스페이스X(SpaceX)의 일론 머스크(Elon Musk)와 블루 오리진(Blue Origin)의 제프 베이조스(Jeff Bezos)가 이끄는 민간 우주 기업들이 본격적으로 참여하게 됩니다. 이들은 우주 비행사들을 달 표면에 내려놓을 달 착륙선(Luna landers)을 개발하고 있습니다. 따라서 아르테미스 2호는 NASA가 대형 우주선과 로켓을 직접 설계하고 운영하는 시대의 사실상 마지막 대규모 임무가 될 것이며, 이후에는 민간 기업과의 협력이 더욱 중요해지는 새로운 우주 시대가 열릴 것으로 예상됩니다.
격동의 시대 속, 달이 전하는 메시지
아르테미스 2호 임무는 전 세계가 전쟁과 갈등으로 혼란스러운 시기에 시작됩니다. 하지만 우주 탐사는 인류 전체를 하나로 묶는 통합적인 경험이 될 수 있습니다. 1960년대, 특히 1968년은 마틴 루터 킹 목사 암살, 로버트 케네디 암살, 베트남 전쟁, 시카고 민주당 전당대회 폭동 등 미국 역사상 가장 격동적인 시기 중 하나였습니다.
그러나 그해 12월, 아폴로 8호가 발사되어 달 궤도에 진입했습니다. 우주 비행사들은 크리스마스 이브에 달 궤도에서 성경 창세기 구절을 낭독하며 지구로 메시지를 보냈고, 이는 당시의 혼란 속에서 인류에게 평온함과 희망을 전하며 “1968년을 구했다”는 평가를 받기도 했습니다.
아르테미스 2호의 우주 비행사들이 달에서 돌아오며 어떤 메시지를 전할지는 알 수 없지만, 달에서 바라본 지구의 모습은 여전히 인류에게 우리가 모두 같은 행성의 일부임을 일깨우는 강력한 이미지가 될 것입니다. 이 임무가 현재의 혼란스러운 시기에 평온함과 희망의 메시지를 전하며, 인류가 더 이상 격동의 시대에 갇히지 않을 것이라는 희망의 노래가 되기를 기대합니다. 아르테미스 2호는 단순한 우주 비행을 넘어, 인류의 미래를 향한 담대한 꿈과 도전, 그리고 희망을 상징하는 서사시가 될 것입니다.