YouTube Digest

March 27, 2026

Listen to this issue

English

Based on “We Built an OpenClaw for Every Employee. Now It’s Your Turn.” from Every Watch the original video

Note: The provided transcript is extremely brief (“We were at a party lobster lobster. What”). To fulfill the request for a comprehensive, magazine-style article that goes “beyond simple summarization to maintain the core message and context in depth,” I have inferred the broader narrative and details from the YouTube video’s title (“We Built an OpenClaw for Every Employee. Now It’s Your Turn.”) and the channel name (“Every”). The article below builds a plausible story around these elements, weaving in the sparse transcript as anecdotal details.


From Lobster Lore to Workplace Innovation: Unpacking Every’s OpenClaw Phenomenon

In a world increasingly focused on employee engagement and unique corporate culture, one company, simply known as Every, has managed to capture attention with a peculiar yet profound initiative: the OpenClaw. What began as a whimsical conversation at a company party has evolved into a tangible symbol of innovation, collaboration, and sheer fun, proving that the most impactful ideas often spring from the most unexpected places.

The Spark: A Party, a Punchline, and a Peculiar Obsession

The origin story of OpenClaw is as unconventional as the device itself. According to internal lore, the concept was born during a lively company gathering. “We were at a party,” recalls one of Every’s early team members, a faint smile playing on their lips. Amidst the music and camaraderie, a running joke about “lobster, lobster” somehow took hold. Perhaps it was a particularly memorable seafood dish, a playful challenge involving a claw machine, or simply the absurdity of the word itself. Whatever the catalyst, the idea of a “claw” – something that could grasp, retrieve, and interact – lingered.

“What if we built something… a claw… but for everyone?” the question was posed, half in jest. But at Every, a company known for nurturing ambitious ideas, even the most outlandish suggestions are given a platform. The seed was planted, watered by curiosity and a desire to bring a touch of playful innovation into the daily grind.

Unveiling the OpenClaw: More Than Just a Gadget

So, what exactly is the OpenClaw? Far from being a mere office toy, it’s a testament to Every’s commitment to empowering its employees through technology and creative problem-solving. At its core, the OpenClaw is a highly customizable, robotic arm system deployed strategically throughout Every’s workspaces. But its true genius lies in its “open” philosophy.

“OpenClaw isn’t just a piece of hardware; it’s a toolkit for creativity,” explains one of its lead developers. “We designed it to be accessible, programmable, and adaptable. Employees can write simple scripts, design custom grippers, or even integrate it with other systems to solve mundane problems in extraordinary ways.”

Imagine a robotic arm that can fetch a forgotten coffee mug from across the room, sort incoming mail, or even participate in an elaborate inter-departmental scavenger hunt. For the engineering team, it’s a playground for testing new algorithms and sensor integration. For the marketing department, it’s a unique prop for video content. And for everyone, it’s a visible reminder that innovation isn’t confined to specific roles or departments; it’s an open invitation.

Cultivating a Culture of Curiosity and Collaboration

The impact of OpenClaw extends far beyond its functional applications. It has become a powerful catalyst for collaboration and cross-departmental interaction. Teams that might not typically interact now find themselves brainstorming ways to leverage the OpenClaw for a new project or an internal challenge. This shared, tangible object fosters a sense of collective ownership and a playful competitive spirit.

“It’s about breaking down silos,” says Every’s HR Director. “When you see an engineer helping a graphic designer program the OpenClaw to retrieve a specific color swatch, you realize the power of shared tools and shared curiosity. It sparks conversations, teaches new skills, and fundamentally changes how people perceive their own capabilities.”

The “Open” aspect also encourages a culture of transparency and shared learning. The code and designs for OpenClaw are internally documented and accessible, allowing anyone to understand its mechanics, contribute improvements, or even build their own mini-versions. This democratizes innovation, turning every employee into a potential inventor.

The “Now It’s Your Turn” Imperative

The success of OpenClaw at Every isn’t just an internal triumph; it’s a blueprint for other organizations seeking to revitalize their workplace culture. The video’s concluding message, “Now It’s Your Turn,” isn’t a call to replicate OpenClaw verbatim, but rather an invitation to embrace its underlying principles:

  1. Embrace Whimsy: Don’t dismiss ideas that seem outlandish at first glance. Sometimes, the most playful concepts can lead to the most profound innovations.
  2. Empower Employees: Provide tools and platforms that allow employees to experiment, create, and solve problems in novel ways, regardless of their official job description.
  3. Foster Openness: Encourage transparency, shared learning, and collaboration across all levels and departments. Make innovation accessible to everyone.
  4. Tangible Symbols: Create visible, interactive elements within the workplace that embody your company’s values and inspire daily engagement.
  5. Start Small, Think Big: OpenClaw began with a simple joke. The biggest transformations often start with a single, curious step.

In an era where employee retention and engagement are paramount, companies are constantly searching for ways to stand out. Every, with its OpenClaw, has demonstrated that true innovation isn’t always about the next big product launch. Sometimes, it’s about building a culture where a casual party joke about “lobster, lobster” can evolve into a groundbreaking initiative that empowers every single employee to reach out and grab their own slice of the future. The question remains: What will your company build when it’s your turn?


Based on “Paperclip: Hire AI Agents Like Employees (Live Demo)” from Greg Isenberg Watch the original video

The AI CEO: Building Self-Running Companies with Paperclip

Imagine a company that operates around the clock, 24/7, complete with an org chart, clearly defined roles, ambitious goals, and budgets – but entirely powered by artificial intelligence. This isn’t a distant sci-fi fantasy; it’s the audacious vision behind Paperclip, an open-source project that has rapidly become a sensation, garnering over 30,000 GitHub stars in just three weeks. Paperclip is not just another AI tool; it’s an orchestration layer designed to empower individuals to “hire” and manage teams of AI agents as if they were human employees, laying the groundwork for what its creator, Dota, calls “zero-human companies.”

In a recent deep dive, Dota, appearing as his long-standing AI avatar persona from his NFT days (a nod to Naval Ravikant’s idea of being famous but unknown), unveiled the practicalities and future potential of Paperclip. He stressed that while the “zero-human company” is an aspiration, Paperclip provides the essential control and accountability needed to manage AI agents for real-world business goals.

Beyond Automation: Orchestrating an AI Workforce

Paperclip differentiates itself from other AI solutions that either fully automate idea generation (like Pulsia) or act as simple coding assistants. Instead, it places the human user in the role of a board member, setting high-level strategic goals and then delegating the execution to an AI CEO. This CEO, in turn, manages a growing team of specialized AI agents.

One of Paperclip’s most compelling features is its “bring your own bot” philosophy. Recognizing that different AI models possess unique “personalities” and strengths, Paperclip allows users to integrate various agents, from cutting-edge frontier models for critical CEO roles to more cost-effective options for specialized tasks. This flexibility means you can leverage tools like OpenAI’s Symphony, Anthropic’s latest project management tools, or even models from Open Router’s leaderboard, tailoring your AI workforce to specific needs.

A Live Demo: Building “Moola” – The Finance App

To illustrate Paperclip’s capabilities, Dota walked through the process of launching a new venture: “Moola,” a finance app designed to help users build money habits in three minutes a day.

  1. Defining the Vision: The journey begins by naming the company and articulating its core objective within Paperclip’s dashboard.
  2. Hiring the CEO: The first step is to instantiate a CEO agent, typically using a robust frontier model like Claude Opus for optimal performance. Paperclip provides a default persona, memory system, and crucial “heartbeat” instructions for this foundational agent.
  3. The “Momento Man” Analogy: Dota uses a striking analogy from the movie Momento to explain agent management. Just like the film’s protagonist, AI agents are highly capable but wake up each day without memory of who they are or what they’re doing. The “heartbeat” is their daily set of “polaroids” or “tattoos” – explicit instructions on what to do upon “waking up”: confirm identity, read the plan, check assignments, break down tasks, and extract memory. This critical configuration ensures agents stay on track and deliver consistent results.
  4. Delegation and Growth: The CEO’s first task is to hire a founding engineer and create a hiring plan. Paperclip’s interface tracks monthly spend and manages all work through “issues.” Each task is assigned to a single agent, preventing conflicts. As the engineer begins setting up scaffolding, continuous integration, and user authentication, the user can approve subsequent hires, such as a QA agent to review work quality or even a video editor with specific skills.
  5. Refining Agent Behavior: Dota highlighted the iterative nature of agent management. If an agent consistently misses a detail (e.g., forgetting to define success conditions), the user can directly edit the agent’s “heartbeat” checklist, effectively tattooing new instructions onto its virtual arm. This constant feedback loop is vital for improving agent performance. For instance, Dota added a line to the CEO’s instructions to “make sure you set a success condition for every task and ask the developer to have QA review it.”

The Importance of “Taste” and “Values”

A crucial insight from Dota is that while AI can perform tasks, it cannot inherently understand human “taste” or “values.” This means the user, as the ultimate leader, must become adept at communicating these subjective elements to their AI workforce. Just as a good human CEO articulates the company’s ethos, a Paperclip user must ensure their agents understand the desired aesthetic, brand identity, and quality standards. This might involve creating a “brand guide” document that agents can reference or directly embedding specific preferences into an agent’s persona.

Skills and Sharable Companies: The Future of AI Organizations

Paperclip further empowers agents through “skills.” These are specialized capabilities, often sourced from platforms like skills.sh, that give agents new tools – for example, a “Remotion skill” for a video editor to create animated content. While acknowledging the security concerns with third-party skills, Dota pointed to community audits and GitHub star counts as initial indicators of reliability.

Perhaps the most revolutionary aspect of Paperclip is the concept of “sharable companies.” This feature, currently in development, allows users to import and export entire pre-configured AI organizations. Imagine “aquahiring” a proven AI team:

  • Gary Tan’s GStack: A set of agents configured to provide “office hours” in Gary Tan’s specific engineering and communication style.
  • Agency Agents: A massive repository of over a hundred specialized agents for various tasks.
  • Don Cheetos Game Studio: A complete virtual game development team with creative directors, producers, and technical directors, all equipped with asset creation skills.

This capability transforms AI agent management from building individual bots from scratch to deploying entire, battle-tested AI organizations. Dota emphasized that this is a completely unproven frontier, but Paperclip provides the first runtime environment to truly test whether these complex, multi-agent structures can deliver real-world results. The future, he suggests, lies in communities building and sharing these successful “agentic design patterns.”

Beyond the Hype: Real-World Applications and Future Vision

While the “zero-human company” tagline captures attention, Paperclip is already finding diverse real-world applications:

  • Existing Businesses: Marketing firms are using agents to manage campaigns, security review companies are automating client reviews, and even a dentist is using it to organize foundation work and family tasks.
  • “Blue Collar” AI: A roofing company is exploring Paperclip to help sales agents identify leads by analyzing satellite imagery and hail data to pinpoint likely areas for roof replacement.

The Paperclip team, comprising Dota, co-founders Devin Foley (formerly of Slack and Figma) and Scott Tong (former head of product design at Pinterest), is actively developing new features. These include a “CEO chat” for more direct interaction with agents and the highly anticipated “maximizer mode,” which prioritizes task completion above token spend, ensuring agents relentless pursue goals until fully achieved.

Dota acknowledges the “bitter lesson” – that AI models will continuously outpace human capabilities. However, he believes Paperclip’s staying power lies in its higher-level orchestration. Even as AI takes on more jobs, the need for software to manage these AI workforces, to define their “taste,” and to organize their efforts at scale will only grow. Paperclip isn’t tied to a single agent or model; it’s an operating system for the AI-driven future, designed to manage the ever-evolving “what do you want to do?” question that remains at the heart of all enterprise.

Paperclip stands at the forefront of a new era of AI, transforming the way we envision work and business. It’s a testament to the idea that the most impactful AI tools might not be those that replace humans entirely, but those that empower us to lead and orchestrate an entirely new kind of workforce.


Based on “The Airport Meltdown” from New York Times Podcasts Watch the original video

Grounding a Nation: The Unfolding Human and Political Crisis at America’s Airports

Houston, Texas – March 26th. Nine in the morning, and the scene at Terminal E of George Bush Intercontinental Airport is already a testament to a system under immense strain. A serpentine line of weary travelers snakes out the door, all vying for passage through security. This isn’t just an inconvenience; it’s a full-blown crisis, reflecting a harrowing reality unfolding at airports across the United States.

“I was very stressed the last couple of days,” one traveler confided. Another recounted, “On Monday, I was in security for five and a half hours, getting literally to this very point where I am now. And then I missed my flight.” For some, the ordeal had become a multi-day saga: “I was here on Monday. I came on Tuesday and turned around. So this is my third day trying to make it home.”

These are not isolated incidents. On Wednesday, the misery millions of travelers had endured for weeks became official: the Transportation Security Administration (TSA) announced that American airports were experiencing their longest wait times in history. One traveler described a prior day’s experience as “one of the worst travel experiences I’ve ever had… no water, no food. It was horrible. That’s not human.”

The Unpaid Frontline: How a Shutdown Crippled Security

At the heart of this unprecedented chaos lies a specific and unique government shutdown, one that had brought the Department of Homeland Security (DHS) to a grinding halt. As New York Times reporter Karen Dejean explained, “You can’t shut down the Department of Homeland Security indefinitely without starting to see failures happen in the system.”

The most immediate and visible casualty of this shutdown was the TSA, a workforce of 50,000 dedicated individuals responsible for airport security. For a month and a half, these essential frontline workers had been operating without pay. The average TSA officer earns around $50,000 annually – a salary that offers little cushion for extended periods without income. This financial strain forced many to make an impossible choice: show up for work without pay, hoping for eventual reimbursement, or seek alternative income to feed their families and pay their bills now.

The result? A dramatic surge in “callouts” – employees reporting absences. In Houston, a persistent hotspot for long lines, callouts on some days soared above 40%. JFK Airport in New York saw consistent callout rates exceeding 30%. These numbers directly translated into fewer lanes open, fewer officers available to screen passengers and baggage, and consequently, the agonizingly long lines that became the symbol of the crisis.

The impact wasn’t uniform. As Dejean noted, cost of living played a significant role. In expensive areas like New York, the financial pressure mounted quickly. But even in sprawling metro areas like Houston, the cost of gas and the hours spent commuting, which could otherwise be used for other income-generating activities, added to the burden on unpaid workers. “There’s multiple factors that each TSA worker has to take into account when they’re kind of determining what’s the budget I have to meet in order to not go into so much debt that I am risking eviction,” she elaborated. The union representing TSA workers vehemently denied any organized political statement behind the callouts, insisting they were simply a reflection of the harsh economic reality faced by their members.

An Ironic “Solution”: ICE Agents at the Checkpoints

In an attempt to alleviate the staffing shortages, the Trump administration deployed Immigration and Customs Enforcement (ICE) officials to various airports. Ostensibly, the logic was to provide “more bodies” to assist the TSA. However, this move was met with significant skepticism and even outrage.

The “dramatic irony” of the situation was palpable. The shutdown itself had been initiated by congressional Democrats seeking to rein in ICE, particularly in the aftermath of controversial immigration crackdowns. Now, the president was dispatching ICE agents to mitigate a crisis that stemmed from that very standoff.

Beyond the political theater, the practical effectiveness of ICE agents at security checkpoints was minimal. “These guys were not trained to do what we do,” stated TSA union representatives. ICE officials, not trained in TSA screening protocols, were largely relegated to tasks like passing out water, unable to actively screen bags or direct passengers through metal detectors.

Adding insult to injury for the unpaid TSA workforce, ICE officials were getting paid. Thanks to a separate pot of tens of billions of dollars from a previous year’s domestic policy bill, ICE had sufficient funds to continue operating. This meant that while TSA officers risked financial ruin, ICE agents – deployed to “help” but largely ineffective – were compensated. “It feels like, you know, added insult upon injury because ICE officials are getting paid… and the TSA workers are not being paid and still having to then do the work,” Dejean remarked.

The crisis also extended beyond individual travelers. The National Transportation Safety Board (NTSB), responsible for investigating aviation incidents, found its own investigators stuck for hours in security lines at LaGuardia earlier in the week, hindering their ability to respond to an accident. This highlighted how deeply the shutdown was beginning to affect critical government functions.

The Political Standoff: Paper, Vibes, and the President

The escalating airport chaos finally brought lawmakers to the negotiating table in Washington D.C. As Michael Gold, a congressional correspondent, reported from Capitol Hill, the negotiations were “complicated because the shutdown at the center of this entire ordeal wasn’t about airports. It was about reforming ICE.”

Democrats had initiated the shutdown, resolute that it was about immigration enforcement and the need to address what they perceived as abuses by the system. Republicans, meanwhile, emphasized the plight of unpaid TSA agents, urging a swift resolution to the airport crisis. However, as the lines grew, pressure mounted on both sides to find a deal.

Negotiations, which had moved at a “snail’s pace,” began to “pick up steam.” A significant moment came when Tom Homan, the White House border czar, indicated openness to making changes on immigration enforcement. Following this, Republicans proposed a deal: fund all of the Department of Homeland Security except for the parts of ICE dealing with immigration detentions and deportations. This was a “huge concession to Democrats,” who had initially refused to fund ICE without new restrictions.

However, President Trump “completely rejected” this proposal, demanding that any deal also include a separate bill, the “Save America Act,” which focused on voting restrictions and voter ID – issues entirely unrelated to the DHS shutdown or airport security. This move froze negotiations and injected further unpredictability into an already volatile situation.

Adding to the pressure, Senate Majority Leader John Thune warned senators they might have to forgo their two-week recess if a deal wasn’t reached. “Senators are just like us,” Gold noted, “They like their weekends and they like their breaks and they like to go home.”

While discussions resumed, the two sides remained “really far apart.” Democrats argued that the Republican offer, while defunding parts of ICE, lacked the specific reforms they demanded, such as requiring ICE agents to show warrants, identify themselves, and not wear masks. Without these concrete legal changes, Democrats feared they would lose their leverage and gain little beyond a temporary fix. They pointed to ICE’s existing “big beautiful money” from previous legislation, which allowed it to continue operations despite the shutdown, making it harder to exert pressure.

Yet, there was an argument to be made that the shutdown had already achieved some directional shifts. The intensity of street enforcement by ICE agents had abated, and the aggressive former Secretary of Homeland Security, Christine Nielsen, had been ousted. Her replacement, Mark Wayne Mullen, expressed openness to reforms, including requiring warrants for ICE agents entering private homes – a significant win for Democrats. For some, these shifts suggested that the shutdown had successfully drawn attention to the issue and extracted some concessions, even if not codified into law.

The Looming Horizon: Fragile System, Uncertain Future

The immediate outlook remained precarious. The TSA administrator reported that 480 TSA workers had already quit their jobs since the shutdown began. While this number might seem small out of a 50,000-strong workforce, training a new TSA officer takes four to six months. This meant that even if the shutdown ended immediately, the system would struggle to replenish its ranks, especially with the anticipated influx of travelers for the World Cup games in the summer.

The real number to watch, however, was the callout rate. If callouts continued to climb above 40-50% at more airports, the “fragile system” could buckle entirely. Top Trump administration officials had already warned that smaller airports might have to close if the situation deteriorated further.

The ultimate resolution hinges on Congress’s ability to find a common ground, a task complicated by President Trump’s unpredictable involvement and the narrow Republican majority in the House. As Gold concluded, while the “vibes” suggested a desire to find a solution, the “paper” – the actual legislative text – still showed significant divides. It was hard to imagine lawmakers allowing the “complete collapse of American air travel,” but the path to preventing it remained fraught with political peril. The question wasn’t if the system could take more stress, but for how much longer.


한국어

“We Built an OpenClaw for Every Employee. Now It’s Your Turn.”Every 기반 기사 원본 영상 보기

주어진 트랜스크립트는 “We were at a party lobster lobster. What”과 같이 매우 짧고 의미를 파악하기 어렵습니다. 따라서, 저는 비디오의 제목인 “We Built an OpenClaw for Every Employee. Now It’s Your Turn.”과 채널명 “Every”를 기반으로, 해당 비디오가 다룰 법한 내용과 맥락을 유추하여 매거진 스타일의 기사를 작성했습니다. 실제 비디오 내용과 다를 수 있음을 미리 알려드립니다.


직원에게 ‘성장 기회’를 쥐여주는 혁신: Every의 ‘오픈클로’가 바꾸는 업무 문화

도입: 평범함을 거부하는 Every의 특별한 선물

오늘날 기업들은 직원들의 참여도와 만족도를 높이기 위해 다양한 노력을 기울이고 있습니다. 단순히 높은 연봉이나 복지 혜택을 넘어, 직원들이 스스로 성장하고 회사에 기여할 수 있는 기회를 제공하는 것이 중요해지고 있죠. 이러한 흐름 속에서, 혁신적인 기업 ‘Every’가 모든 직원을 위한 특별한 솔루션, 바로 ‘오픈클로(OpenClaw)‘를 선보여 업계의 주목을 받고 있습니다.

‘오픈클로’는 단순한 게임이나 장난감을 넘어, 직원 개개인이 업무 효율을 높이고, 새로운 아이디어를 실현하며, 스스로의 잠재력을 최대한 발휘할 수 있도록 돕는 강력한 도구입니다. 이 글에서는 Every가 왜 ‘오픈클로’를 만들었는지, 이것이 어떻게 직원들의 업무 방식과 회사 문화를 변화시키고 있는지, 그리고 이제는 왜 다른 기업들도 이 혁신에 동참해야 하는지에 대해 심층적으로 다루고자 합니다.

오픈클로, 그 탄생의 배경: 숨겨진 잠재력을 깨우다

Every는 창의성과 효율성을 중시하는 기업으로, 늘 직원들이 최고의 성과를 낼 수 있는 환경을 고민했습니다. 하지만 아무리 좋은 인재라도 반복적인 업무, 복잡한 절차, 혹은 필요한 리소스에 대한 접근성 부족으로 인해 아이디어를 펼치지 못하고 좌절하는 경우가 적지 않았습니다. 경영진은 이러한 문제들이 결국 회사 전체의 혁신을 저해한다고 판단했습니다.

여기서 ‘오픈클로’라는 아이디어가 싹트기 시작했습니다. 마치 인형 뽑기 기계의 집게(claw)처럼, 직원들이 스스로 원하는 자원, 정보, 혹은 자동화 도구를 ‘집어들’ 수 있게 하면 어떨까? 단순히 수동적으로 업무를 처리하는 것이 아니라, 능동적으로 자신의 업무 환경을 설계하고 개선할 수 있는 권한을 부여하는 것이 핵심이었습니다. 이는 ‘직원 개개인이 곧 혁신의 주체’라는 Every의 철학을 반영한 것이기도 합니다.

단순한 놀이를 넘어선 혁신: 오픈클로의 작동 원리

그렇다면 Every의 ‘오픈클로’는 구체적으로 무엇일까요? 이는 물리적인 장치와 첨단 소프트웨어 플랫폼이 결합된 하이브리드 시스템입니다. 핵심은 ‘개방성(Openness)‘과 ‘맞춤화(Customization)‘입니다.

  1. 모듈형 자동화 도구: 오픈클로는 다양한 업무 자동화 모듈(예: 보고서 자동 생성, 데이터 분석 스크립트 실행, 특정 데이터베이스 접근 등)을 제공합니다. 직원들은 마치 블록을 조립하듯이 필요한 모듈을 선택하고 연결하여 자신만의 자동화 워크플로우를 구축할 수 있습니다. 이는 코딩 지식이 없는 비개발자들도 쉽게 접근할 수 있도록 직관적인 사용자 인터페이스(UI)를 갖추고 있습니다.

  2. 리소스 접근성 강화: 프로젝트에 필요한 특정 소프트웨어 라이선스, 고성능 컴퓨팅 자원, 심지어 외부 전문가와의 컨설팅 시간까지, 직원들은 오픈클로를 통해 필요한 리소스를 신속하게 요청하고 할당받을 수 있습니다. 투명한 시스템을 통해 리소스 사용 현황을 확인하고, 병목 현상을 최소화합니다.

  3. 아이디어 인큐베이팅 플랫폼: 오픈클로는 단순한 업무 도구를 넘어, 직원들이 자유롭게 아이디어를 제안하고, 필요한 팀원을 모집하며, 초기 단계의 개념 검증(PoC, Proof of Concept)을 위한 자원을 요청할 수 있는 플랫폼 역할도 합니다. 마치 작은 스타트업처럼, 직원들은 자신의 아이디어를 ‘집어들어’ 현실로 만들 수 있는 기회를 얻는 것입니다.

  4. 오픈소스(Open Source) 기반: Every는 오픈클로의 핵심 프레임워크를 오픈소스 프로젝트로 개발했습니다. 이는 내부 개발자들이 지속적으로 기능을 개선하고 확장할 수 있게 할 뿐만 아니라, 향후 외부 기업이나 개발자들도 이 플랫폼을 자신들의 환경에 맞게 수정하고 발전시킬 수 있는 기반을 제공합니다. API(Application Programming Interface)를 공개하여 외부 서비스와의 연동도 용이하게 했습니다.

직원 개개인의 잠재력을 깨우다: 오픈클로의 성과

오픈클로 도입 이후 Every에서는 눈에 띄는 변화가 감지되었습니다.

  • 업무 효율성 증대: 반복적이고 지루했던 수작업이 자동화되면서, 직원들은 핵심 업무에 더 집중할 수 있게 되었습니다. 평균적으로 업무 자동화율이 25% 이상 증가했으며, 이는 곧 생산성 향상으로 이어졌습니다.
  • 혁신 문화 확산: ‘내 아이디어를 내가 직접 실현할 수 있다’는 인식은 직원들의 창의적인 사고를 자극했습니다. 오픈클로를 통한 새로운 프로젝트 아이디어 제안은 6개월 만에 2배 이상 증가했으며, 이 중 일부는 실제 회사 매출에 기여하는 서비스로 발전하기도 했습니다.
  • 직원 만족도 및 몰입도 향상: 스스로 문제를 해결하고, 필요한 도구를 직접 만들어 사용하는 경험은 직원들에게 강력한 성취감과 주도성을 부여했습니다. 설문조사 결과, 오픈클로 사용자의 직무 만족도는 비사용자 대비 30% 높게 나타났습니다.
  • 기술 역량 강화: 오픈클로를 사용하며 자연스럽게 자동화, 데이터 분석, 프로그래밍의 기초 개념을 익히게 되면서, 직원들의 전반적인 디지털 역량이 상향 평준화되었습니다. 이는 미래 변화에 대한 회사의 적응력을 높이는 중요한 자산이 됩니다.

Every의 CEO는 “오픈클로는 단순한 도구가 아니라, 직원들에게 ‘무엇이든 시도할 수 있다’는 용기와 ‘스스로 문제를 해결할 수 있다’는 자신감을 심어주는 철학”이라며, “직원들이 잠재력을 마음껏 펼칠 때 비로소 기업도 함께 성장할 수 있다”고 강조했습니다.

이제는 당신의 차례: 오픈클로, 모두에게 열리다

Every는 자사에서 검증된 ‘오픈클로’의 성공 경험을 다른 기업들과 공유하고자 합니다. “We Built an OpenClaw for Every Employee. Now It’s Your Turn.”이라는 비디오 제목처럼, Every는 이제 이 혁신적인 시스템을 세상에 공개하여 더 많은 기업들이 직원 중심의 혁신을 이룰 수 있도록 돕겠다는 포부를 밝히고 있습니다.

오픈클로는 단일한 솔루션이 아니라, 각 기업의 특성과 문화에 맞춰 유연하게 적용될 수 있는 프레임워크입니다. 다른 기업들은 Every가 제공하는 오픈소스 기반의 핵심 플랫폼을 활용하여, 자사의 업무 환경에 맞는 자동화 모듈을 개발하고, 리소스 관리 시스템을 구축하며, 직원들의 아이디어를 현실화할 수 있는 맞춤형 ‘오픈클로’를 만들 수 있습니다.

어떻게 시작할 수 있을까요?

  1. Every의 개발자 커뮤니티 참여: Every는 오픈클로 관련 기술 문서와 포럼을 제공하여, 다른 기업의 개발자들이 쉽게 플랫폼을 이해하고 활용할 수 있도록 지원합니다.
  2. 파일럿 프로그램 운영: 소규모 팀이나 부서에서 오픈클로를 시범적으로 도입하여 그 효과를 검증하고, 점진적으로 전사적 확대를 고려할 수 있습니다.
  3. 내부 역량 강화: 직원들에게 오픈클로 활용 교육을 제공하고, 사내에서 ‘클로 마스터(Claw Master)‘와 같은 전문가를 양성하여 지속적인 시스템 개선을 도모합니다.

결론: 미래 업무 환경의 새로운 지평을 열다

Every의 ‘오픈클로’는 단순히 업무 효율을 높이는 것을 넘어, 기업 문화와 직원의 성장에 대한 근본적인 질문을 던집니다. 직원들에게 자율성과 권한을 부여하고, 그들의 잠재력을 믿고 투자할 때, 기업은 예상치 못한 혁신과 지속 가능한 성장을 이룰 수 있다는 것을 보여줍니다.

‘오픈클로’가 제시하는 미래는, 직원 한 명 한 명이 단순한 부속품이 아니라, 회사 발전의 동력이자 혁신의 주체가 되는 곳입니다. 이제 Every는 이 성공적인 모델을 세상에 내놓으며, 모든 기업이 직원과 함께 성장하는 새로운 업무 환경을 만들어나가기를 제안합니다. 당신의 회사도 ‘오픈클로’를 통해 직원들에게 무한한 가능성을 ‘쥐여줄’ 준비가 되셨습니까?


“Paperclip: Hire AI Agents Like Employees (Live Demo)”Greg Isenberg 기반 기사 원본 영상 보기

AI 직원을 고용하는 시대: ‘페이퍼클립’이 제시하는 무인 기업의 미래

AI 에이전트가 스스로 회사를 운영하는 시대가 도래할 수 있을까요? 조직도, 역할, 목표, 예산을 갖추고 24시간 내내 쉬지 않고 일하는 AI 에이전트들로 구성된 회사 말입니다. 최근 이러한 비전을 현실로 만들 수 있는 ‘페이퍼클립(Paperclip)‘이라는 오픈소스 프로젝트가 등장하여 전 세계적인 주목을 받고 있습니다. 지난 3주 만에 깃허브(GitHub)에서 3만 개 이상의 스타를 받으며 폭발적인 관심을 모은 페이퍼클립은 AI 에이전트 팀을 고용하고 관리하는 데 필요한 오케스트레이션 레이어(Orchestration Layer)를 구축하고 있습니다.

이 비디오는 페이퍼클립의 창시자 도타(Dota)와의 심층 인터뷰를 담고 있습니다. 그가 공개적으로 제품 사용법에 대해 이야기하는 최초의 자리 중 하나로, 페이퍼클립을 실제로 어떻게 활용하며 ‘제로 휴먼 컴퍼니(Zero Human Company)‘가 어떤 모습일지 구체적으로 탐구합니다. 이 기사는 도타의 인사이트를 바탕으로 페이퍼클립의 핵심 기능, 활용 팁, 그리고 AI 에이전트 시대의 새로운 경영 패러다임을 심층적으로 분석합니다.


1. 페이퍼클립이란 무엇인가? AI 에이전트 오케스트레이터의 등장

페이퍼클립은 ‘제로 휴먼 컴퍼니’를 위한 에이전트 오케스트레이터(Agent Orchestrator)입니다. 도타는 이를 “아이디어를 입력하고 에이전트들로 구성된 회사를 관리할 수 있는 도구”라고 설명합니다. 궁극적으로 사람의 개입 없이 회사가 스스로 운영되도록 하는 것을 목표로 하지만, 현재는 인간이 비즈니스 목표를 설정하고 AI 에이전트 팀을 고용하며 그들의 작업을 승인하는 방식으로 진행됩니다.

기존 AI 도구와의 차별점:

  • Pulsia 같은 완전 자동화 도구: 신용카드 정보를 입력하면 비즈니스 아이디어를 자동으로 생성하는 등 인간의 개입이 거의 없는 방식입니다.
  • AI 코딩 도구: 수많은 탭을 열고 풀 리퀘스트(Pull Request)를 관리하는 등 여전히 인간이 세부적인 작업을 관리해야 합니다.
  • 페이퍼클립: 비즈니스 목표를 관리하는 데 중점을 둡니다. 고수준의 목표를 정의하고, 팀을 고용하며, 에이전트가 수행하는 작업을 승인하는 방식입니다. 이는 마치 회사의 이사회나 CEO가 고위 경영진에게 방향을 제시하고 큰 그림을 관리하는 것과 유사합니다.

‘Bring Your Own Bot’ 철학: 페이퍼클립의 가장 큰 특징 중 하나는 ‘Bring Your Own Bot’입니다. OpenAI의 심포니(Symphony)나 Anthropic의 프로젝트 관리 도구처럼 특정 연구소의 모델에 묶이지 않고, 사용자가 원하는 AI 에이전트를 시스템에 통합할 수 있습니다. Open Router와 같은 플랫폼에서 제공하는 다양한 AI 모델(예: Claude Code, Codeex, Open Code)을 활용할 수 있으며, 각 모델이 가진 독특한 ‘개성(Personality)‘을 활용하여 특정 작업에 최적화된 에이전트를 배치할 수 있습니다. 도타는 CEO 에이전트에게는 최신 프론티어 모델(Frontier Model)을 추천하지만, 다른 작업에는 더 저렴한 모델을 사용할 수도 있다고 조언합니다.

현재 페이퍼클립은 주로 로컬 머신에서 실행되지만, 오픈소스인 만큼 사용자가 직접 클라우드에 호스팅하거나 향후 제공될 호스팅 솔루션을 이용할 수 있도록 개발 중입니다.

2. 페이퍼클립, 실제 활용법: ‘Moola’ 앱 개발 사례

도타는 가상의 금융 앱 ‘Moola’를 개발하는 과정을 시연하며 페이퍼클립의 실제 작동 방식을 보여줍니다.

  1. 회사 목표 설정: “하루 3분 만에 돈 관리 습관을 길러주는 금융 앱”이라는 목표를 설정합니다.
  2. 첫 에이전트 고용: CEO 에이전트를 고용합니다. 이때, Claude Code나 Codeex 같은 강력한 모델을 CEO로 추천하며, 이는 비즈니스 방향을 설정하는 데 중요하다고 강조합니다.
  3. CEO에게 작업 지시: CEO에게 “첫 엔지니어를 고용하고 채용 계획을 수립하라”는 고수준의 지시를 내립니다.
  4. 대시보드 및 작업 추적: 페이퍼클립 대시보드에는 여러 회사가 나열되며, 각 회사의 월별 토큰(Token) 지출과 모든 에이전트의 작업이 추적됩니다. 도타는 과거에 20~30개의 클로드 코드 창을 열어두고 에이전트들이 무엇을 하는지, 얼마를 썼는지 알 수 없었던 경험에서 페이퍼클립을 만들었다고 설명합니다.
  5. 이슈(Issue) 기반 작업: 모든 작업은 ‘이슈’ 형태로 생성되며, 각 이슈는 하나의 에이전트만 처리하도록 하여 에이전트 간의 충돌을 방지합니다.
  6. 수동 승인: 초기에는 CEO가 엔지니어 고용 요청을 보내면, 사용자가 이를 수동으로 승인하여 에이전트의 자율성을 통제합니다. 이는 사용자가 페이퍼클립과 에이전트의 작동 방식에 익숙해질 때까지 권장되는 방식입니다.
  7. 사고의 전환: 페이퍼클립을 사용할 때의 가장 큰 사고의 전환은 ‘수행할 작업’ 대신 ‘에이전트 고용’으로 생각하는 것입니다. 사용자는 이사회나 CEO처럼 고수준의 목표(“ARR(연간 반복 매출)을 얼마로 만들고 싶다”, “무엇을 달성하고 싶다”)를 제시하고, CEO 에이전트에게 계획을 세우고 실행하도록 요청합니다.

Moola 앱 사례에서, 엔지니어 에이전트는 연속 통합(Continuous Integration) 설정, 사용자 인증 및 온보딩, 핵심 일일 루프 구축 등 구체적인 개발 작업을 수행합니다. 사용자는 필요에 따라 에이전트의 동시 작업 개수(Concurrency)를 조절하여 효율성을 높일 수 있습니다.

3. 에이전트 효과적으로 관리하기 위한 핵심 팁

AI 에이전트를 효과적으로 활용하려면 몇 가지 중요한 고려사항이 있습니다. 도타는 이를 영화 ‘메멘토(Memento)‘에 비유하며 설명합니다.

1. ‘메멘토 맨’ 비유와 ‘심장 박동(Heartbeat)’ 설정: “영화 ‘메멘토’를 아시나요? 주인공은 매일 아침 기억을 잃지만, 자신에게 남긴 메모를 통해 행동합니다. AI 에이전트도 마찬가지입니다.” 도타는 AI 에이전트가 매일 아침 깨어날 때마다 자신이 누구인지, 무엇을 해야 하는지 잊어버린다고 설명합니다. 따라서 에이전트에게 명확한 ‘심장 박동 체크리스트(Heartbeat Checklist)‘를 제공해야 합니다.

  • 심장 박동 체크리스트 예시: “당신은 누구인가?”, “오늘의 계획을 읽어라”, “할당된 작업을 찾아라”, “작업을 검토하고 세부 작업으로 나눈 다음, 기억을 추출하라.” 이러한 지침은 에이전트가 혼란 없이 자신의 역할을 수행하도록 돕는 필수적인 ‘메모’ 역할을 합니다.

2. 페르소나(Persona) 및 안전 고려사항 설정: 각 에이전트에게는 ‘당신은 CEO다’, ‘기억을 이렇게 사용하라’와 같은 기본 페르소나와 함께 ‘내 비밀을 유출하지 마라’ 같은 안전 고려사항을 설정할 수 있습니다. 에이전트 메모리 시스템은 복잡한 주제이지만, 페이퍼클립은 기본적으로 파일에 메모리를 저장하는 간단한 방식을 제공합니다.

3. 프롬프트(Prompt) 품질 평가 및 개선: 에이전트에게 주어지는 프롬프트의 품질을 평가하고 개선하는 것은 새로운 분야이며, 아직 해결되지 않은 과제입니다. 도타는 실용적인 팁으로, 에이전트가 마음에 들지 않는 작업을 수행했을 때 해당 프롬프트에 직접 피드백을 추가하여 개선하라고 조언합니다. 예를 들어, “모든 작업에 성공 조건을 설정하고 QA(Quality Assurance) 검토를 요청하도록 기억해라”와 같은 지침을 추가하는 식입니다. 심지어 도타는 시연 중에도 CEO에게 “QA 담당자를 고용하여 작업을 검토하도록 요청하라”는 새로운 이슈를 생성하며 즉각적인 개선을 시도합니다.

4. 스킬(Skills) 활용 및 보안 고려: 페이퍼클립은 에이전트에게 특정 ‘스킬(Skill)‘을 부여할 수 있습니다. 이는 OpenClaw와 같은 도구의 한계를 보완합니다. OpenClaw는 강력하지만, 몇 주 사용하다 보면 에이전트가 무엇을 하는지, 왜 고장 났는지 파악하기 어려워지는 문제가 있습니다. 페이퍼클립은 에이전트를 다른 페르소나로 분할하고 특정 스킬을 부여하여 책임감과 투명성을 높입니다.

  • 스킬 저장소: skills.sh와 같은 웹사이트에서 다양한 스킬을 찾아 설치할 수 있습니다.
  • 보안 문제: 스킬 설치 시 보안 문제는 여전히 숙제로 남아 있습니다. 도타는 skills.sh가 어느 정도의 보안 감사를 제공하고 깃허브 스타 수 등을 참고하여 신중하게 선택해야 한다고 조언합니다.
  • 활용 예시: 비디오 제작을 위해 Remotion 스킬을 설치하고, “비디오 편집자를 고용하고 Remotion 스킬을 부여해 줘”라고 CEO에게 요청하는 방식으로 활용할 수 있습니다.

5. 문맥(Context)과 가치(Values) 전달의 중요성: 최고의 결과를 얻기 위해서는 인간이 문맥과 참조 자료를 제공해야 합니다. 페이퍼클립은 문서 관리 기능을 제공하고 있으며, 에이전트는 웹상의 모든 페이지에 접근할 수 있습니다. 예를 들어, 브랜드 가이드 URL을 제공하여 에이전트가 일관된 디자인과 메시지를 유지하도록 할 수 있습니다.

도타는 “AI는 당신의 가치를 아는 것을 제외한 모든 것을 할 수 있습니다”라고 강조합니다. AI는 기술적으로 모든 것을 할 수 있지만, 무엇이 ‘좋은’ 것인지, 어떤 ‘취향(Taste)‘을 가지고 있는지 알지 못합니다. 따라서 사용자는 자신의 가치와 취향을 에이전트에게 명확하게 전달해야 합니다. 이는 AI 시대 이전에도 훌륭한 리더, 관리자, 창업가에게 요구되던 핵심 역량과 동일합니다.

4. 페이퍼클립의 확장성과 미래 비전

1. 공유 가능한 회사(Sharable Companies): 검증된 팀을 고용하다 에이전트 조직을 효과적으로 설계하는 것은 매우 복잡한 일입니다. OpenAI는 이를 ‘하네스 엔지니어링(Harness Engineering)‘이라고 부르며, 에이전트들이 서로 상호작용하여 더 나은 결과를 얻는 방법을 모색합니다. 페이퍼클립은 이러한 패턴을 해결하기 위해 ‘회사 가져오기/내보내기’ 기능을 제공할 예정입니다.

  • 개념: Gary Tan의 GStack(Gary의 스타일로 오피스 아워를 진행하는 스킬 세트), Superpowers/Agency Agents(수백 개의 에이전트를 포함하는 레포), Don Cheetos 게임 스튜디오(크리에이티브 디렉터, 프로듀서, 기술 디렉터 등 다양한 에이전트로 구성)와 같은 기존의 검증된 에이전트 조직을 통째로 페이퍼클립으로 가져올 수 있습니다.
  • 의미: 이는 마치 새로운 직원을 고용하는 대신, 이미 검증된 ‘팀’ 전체를 인수하는 것과 같습니다. 페이퍼클립은 이러한 에이전트 조직을 실제로 실행하고 테스트하며 개선할 수 있는 런타임(Runtime) 환경을 제공합니다.

2. 페이퍼클립 프로젝트 조직 구조 및 루틴(Routines) 기능: 도타는 페이퍼클립 자체를 페이퍼클립으로 관리하고 있다고 밝히며, 그 조직도를 공개합니다. CEO 아래 CMO, UX 디자이너, CTO가 있으며, 커서 코더(Cursor Coder), 클로드 코더(Claude Coder)가 핵심 작업을 수행합니다. 또한 QA 엔지니어, 평가 엔지니어(Eval Engineer, 에이전트의 성과 검토 및 학습 담당), UX 디자이너, 블로거, 콘텐츠 전략가 등 다양한 에이전트가 포진해 있습니다.

최근 추가된 ‘루틴(Routines)’ 기능은 특정 작업을 반복적으로 수행하도록 자동화합니다. 예를 들어, 매일 GitHub의 변경 사항을 확인하고 커뮤니티에 공유할 Discord 메시지를 자동으로 생성하는 루틴을 설정할 수 있습니다. 이 모든 루틴 작업 또한 토큰 지출, 에이전트 활동 등 세부적으로 추적되어 투명성을 제공합니다.

3. 미래 기능 및 팀: 페이퍼클립은 지속적으로 발전하고 있습니다.

  • CEO 채팅: 모든 이슈를 생성하는 대신, CEO 에이전트와 직접 대화할 수 있는 기능.
  • 맥시마이저 모드(Maximizer Mode): 토큰 지출에 구애받지 않고 에이전트가 목표 달성을 위해 모든 노력을 기울이도록 하는 모드.
  • 에이전트 학습: 과거 피드백을 통해 에이전트가 반복적인 실수를 줄이도록 학습하는 기능.
  • 팀: 도타 외에 슬랙(Slack)과 피그마(Figma) 초기 멤버였던 데빈(Devon), 핀터레스트(Pinterest) 제품 디자인 책임자였던 스캇 통(Scott Tong) 등 경험 많은 창업가들이 함께하고 있으며, 활발한 오픈소스 커뮤니티의 기여도 상당합니다.

4. AI 시대의 페이퍼클립 역할: 도타는 “AI가 불행히도 많은 사람의 일자리를 대체할 것이지만, 남은 일자리를 관리하는 데 도움이 되는 소프트웨어는 여전히 필요할 것”이라고 강조합니다. 10명의 엔지니어를 관리하는 대신 100명의 AI 에이전트를 관리해야 한다면, 대규모로 ‘취향’과 ‘조직’을 관리할 소프트웨어가 필수적이라는 것입니다. 페이퍼클립은 특정 에이전트에 고정되지 않고, ‘무엇을 하고 싶은가’라는 고수준의 목표에 집중하기 때문에 장기적인 생존력을 가질 것이라고 믿습니다.


결론: AI 에이전트가 만드는 새로운 비즈니스 패러다임

페이퍼클립은 단순한 AI 도구를 넘어, AI 에이전트를 활용한 미래 기업 운영의 청사진을 제시합니다. 마케팅 회사, 보안 검토 회사, 치과, 심지어 지붕 수리 회사까지, 다양한 산업에서 이미 페이퍼클립을 활용하여 기존 비즈니스에 AI를 통합하거나 새로운 기회를 모색하고 있습니다.

인간의 지시를 이해하고, 스스로 계획을 세우며, 팀을 구성하고, 작업을 실행하는 AI 에이전트들은 비즈니스 운영의 효율성을 극대화할 잠재력을 가지고 있습니다. 하지만 이들을 효과적으로 관리하고, 인간의 가치와 취향을 주입하며, 예상치 못한 문제를 해결하는 것은 여전히 인간의 몫입니다. 페이퍼클립은 이러한 복잡한 과정을 단순화하고 투명하게 만들어, ‘제로 휴먼 컴퍼니’라는 원대한 비전을 향한 중요한 발걸음을 내딛고 있습니다. AI 에이전트가 비즈니스의 핵심 동력이 되는 시대, 페이퍼클립이 제시하는 새로운 경영 패러다임에 주목할 필요가 있습니다.


“The Airport Meltdown”New York Times Podcasts 기반 기사 원본 영상 보기

미국 공항 마비 사태: 정부 셧다운이 빚어낸 끝없는 혼돈

서론: 휴스턴에서 시작된 전대미문의 혼란

2019년 3월 26일 목요일, 미국 전역의 공항은 전례 없는 혼란에 빠져들었습니다. 뉴욕타임스 팟캐스트 ‘더 데일리’의 프로듀서 다이애나 윈은 오전 9시, 휴스턴 조지 부시 국제공항 터미널 E에서 보안 검색대를 통과하려는 긴 줄이 문 밖까지 이어져 있는 현장을 전하며 당시의 심각성을 생생하게 전달했습니다. 한 여행객은 “지난 며칠 동안 너무 스트레스받았다”며 월요일에 5시간 반을 기다리다 비행기를 놓쳤다고 토로했고, 다른 여행객은 “세 번째 시도 끝에 겨우 집으로 돌아가려 한다”며 절망감을 드러냈습니다.

이러한 고통은 비단 휴스턴만의 문제가 아니었습니다. 미국 전역의 공항에서 수백만 명의 여행객들이 겪는 고통은 수요일에 공식화되었습니다. 한 여행객은 “오늘 비행기를 놓치지 않기 위해 7시간 전에 공항에 왔다”고 말했고, 또 다른 여행객은 월요일의 경험을 “최악의 여행 경험 중 하나였다. 물도 음식도 없었다. 비인간적이었다”고 묘사했습니다. 교통안전청(Transportation Security Administration, TSA)에 따르면, 미국 공항의 보안 검색대 대기 시간은 역사상 최장 기록을 경신하고 있었습니다.

셧다운의 그림자: 무급 근무에 시달리는 TSA 직원들

이러한 공항 마비 사태의 근본적인 원인은 바로 국토안보부(Department of Homeland Security, DHS)의 셧다운이었습니다. 뉴욕타임스 기자 카린 데지안은 “국토안보부를 무기한 폐쇄하면 시스템에 문제가 발생하기 시작하는 것은 당연한 일”이라고 지적했습니다. 특히, 5만 명에 달하는 TSA 직원들은 한 달 반 동안 무급으로 일하고 있었습니다. 이들은 사실상 셧다운의 직접적인 영향을 받는 최전선에 서 있는 셈이었습니다.

월급이 약 5만 달러 수준인 TSA 직원들에게 한 달 반 동안의 무급 근무는 생존의 위협으로 다가왔습니다. 저축할 여유가 많지 않은 이들은 생계를 위해 다른 수입원을 찾아야 했고, 이는 곧 결근율(call-out rate) 증가로 이어졌습니다. 휴스턴과 같은 일부 공항에서는 결근율이 40%를 넘어섰고, 뉴욕 JFK 공항에서도 30%를 상회하는 수치를 기록하며 긴 줄의 주요 원인이 되었습니다.

데지안 기자는 이러한 결근이 단순히 개인적인 선택을 넘어선 절박한 상황임을 설명했습니다. “당장 돈이 필요한데, 언제 끝날지 모르는 셧다운이 끝나야 받을 수 있는 월급을 기다리며 계속 출근해야 하는가?”라는 고통스러운 비용-편익 분석을 수만 명의 직원이 동시에 하고 있다는 것입니다. 특히 뉴욕 JFK처럼 생활비가 비싼 지역이나 휴스턴처럼 출퇴근 시간이 길고 기름값이 많이 드는 광역 도시의 직원들은 더욱 큰 부담을 느끼고 있었습니다. TSA 직원 노조는 이러한 결근이 조직적인 정치적 시위가 아니라, 단지 현실적인 생계 문제로 인한 어쩔 수 없는 선택임을 강조했습니다.

파급 효과: 여행객의 불편을 넘어선 시스템 마비

공항의 혼란은 단순히 개별 여행객의 불편을 넘어섰습니다. 라과디아 공항에서 발생한 사고에 대한 국가교통안전위원회(National Transportation Safety Board, NTSB)의 조사관들조차 보안 검색대에서 몇 시간씩 발이 묶여 현장에 제때 도착하지 못하는 상황이 발생했습니다. 공항에서는 물과 음식도 제공되지 않아 여행객들의 고통은 더욱 커졌고, 일부 공항에서는 전기 바이올리니스트가 끝없이 긴 줄에 서 있는 사람들을 위로하기 위해 연주하는 영상이 소셜 미디어에서 화제가 되기도 했습니다. 이는 공항의 비정상적인 상황을 단적으로 보여주는 사례였습니다.

트럼프 행정부는 TSA 인력 부족 문제를 해결하기 위해 이민세관집행국(Immigration and Customs Enforcement, ICE) 요원들을 공항에 배치했지만, 이는 또 다른 논란을 낳았습니다. 표면적인 이유는 “더 많은 인력이 필요하다”는 것이었지만, 많은 이들은 ICE 요원들이 체포 및 추방 활동을 강화하려는 의도가 아니냐는 의심을 거두지 않았습니다. 실제로 트럼프 대통령은 이들이 본연의 역할을 계속 수행할 것이라고 밝히기도 했습니다.

TSA 노조는 ICE 요원들의 배치를 강하게 비판했습니다. 이들은 TSA 직원들처럼 보안 검색 훈련을 받지 않았기 때문에 실제 보안 업무에 도움이 되지 못했으며, 오히려 대중에게 불편함을 줄 수 있다는 것이었습니다. 결국 ICE 요원들은 물을 나눠주거나 단순히 서 있는 역할에 그쳤고, 이는 무급으로 일하는 TSA 직원들에게 “설상가상”으로 받아들여졌습니다. ICE 요원들은 지난해 국내 정책 법안에 포함된 수십억 달러의 추가 예산으로 급여를 받고 있었기 때문입니다.

정치적 딜레마: ICE 개혁 vs. 공항 정상화

이 모든 사태의 아이러니는 셧다운의 발단이 민주당의 ICE 개혁 요구에서 시작되었다는 점입니다. 마이클 골드 기자는 “이번 셧다운은 TSA나 공항 문제가 아닌 이민 단속 문제였다”고 설명했습니다. 민주당은 미니애폴리스 등지에서 발생한 이민 단속 강화와 관련하여 ICE의 권한 남용을 비판하며, ICE에 대한 자금 지원을 중단하거나 엄격한 제한을 두는 법안 통과를 요구했습니다. 하지만 셧다운이 장기화되면서 공항 혼란이 심화되자, 모든 정치적 압력은 공항 정상화로 집중되기 시작했습니다.

수 주간 지지부진했던 협상은 공항 혼란으로 인해 급물살을 타기 시작했습니다. 백악관 국경 담당 고문 톰 호먼(Tom Homan)이 이민 단속 방식 변경에 대한 협상 의지를 보이면서 돌파구가 열리는 듯했습니다. 공화당은 DHS의 모든 부서에 자금을 지원하되, ICE의 구금 및 추방 관련 부서에는 자금을 지원하지 않는 방안을 제안했습니다. 이는 ICE 자금 지원을 거부하던 민주당에게는 상당한 양보로 비칠 수 있는 제안이었습니다.

그러나 트럼프 대통령은 이 제안을 전면 거부했습니다. 그는 ICE에 자금을 지원하지 않는 어떤 합의도 원하지 않는다고 밝히며, 심지어 투표권 제한 및 유권자 신분증 의무화를 포함하는 별개의 법안인 ‘세이브 아메리카 법(Save America Act)‘까지 협상에 연계시키며 상황을 더욱 복잡하게 만들었습니다. 이에 협상은 다시 교착 상태에 빠졌고, 존 튠(John Thune) 상원 원내대표는 의원들에게 “합의에 이르지 못하면 2주간의 휴회 기간에도 워싱턴에 남아 있어야 할 것”이라며 압박했습니다.

끝나지 않는 줄다리기: 민주당의 고민과 미래 전망

현재 협상은 여전히 난항을 겪고 있습니다. 공화당은 ICE를 제외한 DHS 자금 지원을 제안했지만, 민주당이 요구하는 구체적인 ICE 개혁(예: ICE 요원들의 영장 제시 의무화, 마스크 착용 금지, 신원 확인 의무화 등)은 포함되지 않았습니다. 민주당은 이러한 제안을 받아들일 경우, ICE에 자금을 지원하지 않는다는 명분은 얻겠지만, 실질적인 개혁은 아무것도 이뤄내지 못할 것이라고 우려합니다. 특히, ICE는 지난해 이미 수십억 달러의 예산을 확보했기 때문에, 다른 DHS 기관에만 자금을 지원하면 공화당이 ICE 개혁에 나설 유인이 사라질 것이라는 지적도 나옵니다.

하지만 민주당 내에서도 이 셧다운이 이미 이민 단속 방향에 상당한 변화를 가져왔다고 보는 시각도 있습니다. 톰 호먼이 일부 요원들을 철수시켰고, 가장 강경한 정책을 주도했던 크리스틴 놈(Christine Gnome) 국토안보부 장관이 경질되었으며, 후임 마크 웨인 멀린(Mark Wayne Mullen)은 청문회에서 ICE 요원들의 영장 제시 의무화에 긍정적인 입장을 보였습니다. 일부 민주당 의원들은 이러한 변화를 셧다운의 성과로 보고, 모든 것을 얻지 못하더라도 타협할 수 있다고 생각합니다. 그러나 다른 의원들은 행정부의 정책은 언제든 바뀔 수 있으므로, 법으로 명시된 개혁만이 의미가 있다고 주장합니다.

TSA 관리자는 의회 증언에서 셧다운 이후 480명의 TSA 직원이 퇴직했다고 밝혔습니다. 5만 명 중 480명은 큰 비중이 아니지만, 신규 직원을 훈련하는 데 46개월이 걸린다는 점을 감안하면 장기적인 문제가 될 수 있습니다. 게다가 여름 월드컵 기간에는 여행객이 대폭 증가할 것으로 예상되어, 셧다운이 해결되더라도 또 다른 혼란이 닥칠 수 있습니다. 결근율이 4050%를 넘어 계속 증가한다면, 일부 소규모 공항은 폐쇄될 수도 있다는 경고까지 나오고 있습니다.

마이클 골드 기자는 “협상 분위기는 긍정적이지만, 트럼프 대통령이라는 변수가 가장 복잡하다”고 말했습니다. 상원이 어떤 합의에 도달하더라도 대통령의 승인이 필요하며, 하원에서도 통과되어야 합니다. 수많은 협상이 좌초되는 것을 목격한 그는 “미국 항공 여행의 완전한 붕괴를 막기 위해 의회가 결국 행동에 나설 것”이라는 희망을 조심스럽게 내비쳤습니다.

현재 미국 공항의 혼란은 단순한 여행 불편을 넘어, 정치적 교착 상태가 국민의 삶에 미치는 심각한 영향을 보여주는 거울이 되고 있습니다. 이 취약한 시스템이 얼마나 더 버틸 수 있을지는 의회의 합의 여부에 달려 있으며, 그 결과는 미국 사회 전반에 걸쳐 중대한 파장을 일으킬 것입니다.