English
Based on โThe One Rule That Separates Real AI Agents from Demos | Yutori, Abhishek Dasโ from EO Watch the original video
Beyond the Demo: The Uncompromising Rule for Real AI Agents
In the burgeoning landscape of artificial intelligence, the promise of autonomous agents capable of navigating the web on our behalf is a compelling vision. Yet, for many, this promise often crumbles upon the first interaction. Youโve likely encountered them: the โhundred different agent productsโ claiming to โdo anything on the web,โ only to falter on the simplest tasks. This pervasive unreliability is a significant pain point for Abhishek Das, co-founder and co-CEO of Yutori. Das is on a mission to redefine what an AI agent truly means, guided by a singular, uncompromising rule: โIf itโs not good enough to work on the first try, itโs not good enough.โ
The Pitfalls of Imperfect Automation
The current state of AI agents, while exciting in concept, often falls short in execution. Das explains the fundamental mathematical challenge inherent in complex workflows: โIf we think of a 10-step, 20-step or 50-step workflow, even if the accuracy at each step is like 90%, the 10% error rate compounds very quickly.โ This means a seemingly robust 90% accuracy per step can quickly lead to an overall success rate that is โquite lowโ for any multi-stage task. The technology, he asserts, is simply โnot there yet to do long horizon workflows.โ
What truly irks Das, however, is not just the technical limitation, but the industryโs growing โtolerance for non-determinism and low reliability in shipping products.โ He pushes back vehemently against this โnormalization of slop,โ especially for agentic products. For Das, a product that requires multiple attempts to succeed isnโt a product; itโs a frustrating prototype. He firmly believes that โif itโs not good enough to work on the first try, itโs not good enough.โ This โfirst-try imperativeโ is the bedrock of Yutoriโs philosophy.
A Vision Forged in Curiosity and Code
Abhishek Dasโs journey to co-founding Yutori is rooted in a deep-seated curiosity and a rebellious spirit. Hailing from a family of doctors, his aversion to blood quickly steered him away from medicine. Instead, he found his calling in engineering, captivated by โthe scientific process and methodโ โ the whole life cycle of coming up with hypotheses, designing experiments to validate or invalidate them, drawing conclusions, and then formulating the next set of hypotheses. โI think that is a very neat sort of process and method. I found that really inspiring,โ he recounts.
His time at IIT Rurki was transformative. Initially enrolled in electrical engineering, he soon realized his true passion lay elsewhere. โThat was sort of the first major sort of rebellious streak in me,โ he recalls, describing how he decided to stop paying as much attention to electrical engineering and instead spent his time learning programming and building software. The vibrant programming club and culture at IIT Rurki, particularly a group called SDS Labs, became his crucible. There, a small group of โ10-15 codersโ were constantly โtinkering and hack building a ton of applications for the internet for the rest of the campus.โ
Watching users interact with his creations โ โbuilding something from scratch and putting it out there and seeing how users interact with I think that was like a dopamine hit that kept like sort of fueling this.โ This early exposure to user-centric development, fueled by late nights and shared obsession with like-minded peers, ignited a lifelong ambition to build something of his own. He had considered it at the end of his undergrad and PhD, but โfor various reasons didnโt end up doing it. So, it was just a matter of time.โ He wanted to โpush on that vision that I care about as opposed to working on somebody elseโs vision.โ
That vision is nothing less than a complete reimagining of our digital interactions. โOver the last two or three decades, web browsers by and large have stayed the same,โ Das observes. โThere is an opportunity now to reimagine what that experience looks and feels like.โ Yutori envisions a future where we interact with โAI assistants that take actions and complete tasks on users behalf on the web,โ often proactively in the background. Das believes โdigital agents will become a realityโ sooner than physical ones, allowing us to interact with the web at a โslightly higher level of abstraction.โ
This isnโt about AI replacing humans, but about โhumans and agents working together to overall improve productivity.โ By delegating โall the mundane stuff to AI assistants, AI agents on our behalf,โ we can โfocus on tasks and stuff thatโs more meaningful, thatโs more interesting to us.โ Crucially, itโs also about accessibility. โMy parents for example no longer have to learn every new website and how to operate it,โ he illustrates. โIf they can just tell an assistant that this is what I want to do on this particular website and it does it for them reliably, then thatโs awesome.โ
Engineering Trust: Yutoriโs Blueprint for Reliability
The path to reliable agents, however, is fraught with challenges. The very nature of agents โ making a โsequence of decisionsโ โ means that even small errors can derail an entire workflow. Yutoriโs solution isnโt just about preventing errors, but about gracefully recovering from them. โBeing able to recognize when it makes mistakes and backtrack from that to then go down a different branch is really, really important.โ
This commitment translates into rigorous โevals and guardrails.โ Every single production query that a user runs goes through a โfairly comprehensive set of evalsโ that lets Yutori โquickly identify where these agents are doing well versus not, which domains need more work and so on.โ The webโs dynamic nature means agents will never be trained on every single website out there. โIt is very natural to expect models to also make mistakes,โ Das admits. The key, he reiterates, is the agentโs ability to โrecognize and then backtrack and correct itself to do the right thingโ when faced with an unfamiliar scenario.
Dasโs disdain for โthe normalization of slop and non-determinism and poor reliabilityโ is a core tenet of Yutoriโs development philosophy. They are not just building features; they are engineering trust.
The Craft of โFeeling Seenโ: Beyond Features
In a world where generative AI makes it โvery easy to come up with first prototypesโ using coding LLMs, Das argues that โthe true differentiator is in taste and craft โ in how intuitive and well-designed the product is.โ This goes beyond simply fulfilling user requests.
Yutori adopts an 80/20 approach, prioritizing features based on user feedback but also relying heavily on intuition to identify unspoken needs. โVery often there are ways to build product that users may not be asking for,โ he says, โbut if you built itโฆ then users feel seen and they feel like oh this is someone who is listening to us.โ He offers the example of two-factor authentication auto-fill on mobile phones. โIt is hard to imagine like a user asking for that feature. But it saves a few seconds multiple times a day for people all across the world.โ These โtiny paper cuts,โ he explains, are where thoughtful design truly shines, making users โfeel seen.โ
To cultivate this โtasteโ and โcraft,โ Yutori rigorously โdogfoodsโ its own product. Weekly sessions are dedicated to testing new features internally, with only a fraction making it to external users. This constant internal refinement builds the โmuscleโ for discerning what โawesome or magical feels like.โ Itโs โa lot of reps to build that muscle,โ he notes.
Transparency: The Foundation of Trust
Das also champions the concept of โproof of workโ in AI. Referencing his involvement in the highly cited Grad-CAM project โ which helped visualize what deep learning models โlook atโ to make predictions โ he emphasizes the importance for models to convey โnot just the final prediction or the final answer, but also the proof of work: what are the steps that went into coming up with this final prediction or the final answer.โ
Yutori integrates this philosophy directly into its product. When an agent (or โscoutโ) monitors the web and generates a report, users can โinspect the work that went in in behind the scenesโ โ seeing which websites were visited and what information the agent actually looked at to pull out a piece of information. This transparency, he stresses, is โvery, very important for trust building for users to be able to trust that yes this is a reliable product.โ
For Das, trust is earned through meticulous attention to detail, both visible and invisible. โIf we put attention to detail into parts of the product that users can see, then the user is more likely to trust the parts of the product that they cannot see.โ This philosophy underpins Yutoriโs commitment to โbuild something meaningful, to build something right, to bring a vision of the future to life and building like delightful and reliable product experiences. It doesnโt just appear out of nowhere.โ It takes time, hard work, and an unwavering dedication to getting it right every single time.
In a world saturated with AI demos that promise much and deliver little, Abhishek Das and Yutori are charting a different course. Their โfirst-try imperativeโ isnโt just a product development rule; itโs a foundational principle for an AI-powered future built on trust, reliability, and genuinely intuitive experiences. As AI agents increasingly weave into the fabric of our digital lives, Yutori aims to ensure they are not just smart, but consistently dependable โ transforming mundane digital chores into seamless, invisible assistance.
Based on โThe internal AI tool thatโs transforming how Stripe designs products | Owen Williamsโ from How I AI Watch the original video
Stripeโs AI Secret: How an Internal Tool is Redefining Product Design
In the fast-paced world of tech, where product design can often feel like a bottleneck, companies are constantly searching for ways to accelerate innovation while maintaining a high bar for quality. Stripe, a company renowned for its meticulous design and user experience, is no exception. However, instead of relying solely on off-the-shelf tools, Stripe has cultivated a powerful internal AI-driven prototyping platform called โProtodashโ and โProtodash Studio,โ spearheading a quiet revolution in how its products are conceptualized, built, and refined.
Owen Williams, a design manager at Stripe, found himself at the forefront of this transformation. As he observed designers grappling with generic AI tools, he recognized a critical gap: these tools, while powerful, often produced what he affectionately termed โTailwind indigo slopโโa bland, unbranded aesthetic that failed to capture Stripeโs distinctive visual identity. This โuncanny valleyโ effect broke immersion in design reviews, making it difficult to assess whether a prototype truly represented a Stripe product. His dream was simple yet ambitious: โI want something thatโs like vzero but fast,โ a tool deeply integrated with Stripeโs unique design system.
The Pain of Prototyping in a Data-Rich World
The challenge of creating high-fidelity prototypes is particularly acute for data-heavy products like Stripeโs dashboards. Imagine the sheer effort involved in prototyping a data dashboard in traditional tools like Figma: accounting for all interactions, filters, multiple states (zero data, a bunch of data, error states), internationalization (how does it look in Dutch?), or different business models (startup vs. enterprise). Owen highlights this pain point vividly: โIt is nearly impossible to do that in Figma.โ Designers would spend countless hours manually duplicating components, faking data with โ10 xxxโ placeholders, and creating static screenshots for every possible scenario.
This manual, labor-intensive process not only slowed down the design cycle but also limited the scope of exploration. Designers were often forced to present isolated JPEGs or simplified flows, making it difficult for stakeholders to grasp the full user journey or the nuances of dynamic data. The absence of interactive prototypes meant feedback often focused on static visuals rather than functional experience, leading to superficial discussions.
Protodash: Crafting Prototypes with Code and AI
Owen Williams, with his engineering background, approached this problem pragmatically. He envisioned a system that could lower the barrier to entry for designers while ensuring prototypes adhered to Stripeโs exacting quality standards. The solution was Protodash, an internal prototyping tool that leverages a bundle of Cursor rules, React components, and Stripeโs proprietary design system, โSale.โ
The first iteration of Protodash was a basic setup: a router, Sale components, and an integration with Stripeโs internal MCP (Managed Component Platform) server. The core innovation, however, lay in the โbundled rulesโ fed to LLMs like Cursor (or Claude Code, as mentioned in the transcript). These rules taught the AI how to interpret design requests, prioritize Stripeโs design system, and avoid common pitfalls like generating generic Tailwind styles. For instance, if a user pasted a Figma link, the AI was instructed to check the Sale MCP server before writing any code. This ensured that the generated prototype was built from Stripeโs established building blocks, not generic elements.
The result was transformative. โIt gets you there like 90% of the way most of the time,โ Owen explains. Designers, who are โhigh craft,โ could then refine the remaining 10%. The impact on design reviews was immediate: โItโs so convincing that Iโm like, is this the real product or am I looking at like something fake?โ This level of realism, achieved rapidly, meant stakeholders could interact with prototypes that felt indistinguishable from the live product.
The Magic of Devboxes and Protodash Studio
Initially, Protodash ran locally, requiring designers to โnpm run dev.โ However, Stripeโs sophisticated โdevboxโ infrastructure elevated the experience further. Devboxes are cloud-based development environments that can be spun up in minutes, pre-configured with all necessary tools and dependencies. Now, designers could simply navigate to a URL, and a devbox instance of Protodash would be ready to go, eliminating the need for local setup or managing a large monorepo. This โno npm requiredโ magic was a game-changer, especially for designers who might not have the same powerful machines as engineers.
Building on this, Owen developed โProtodash Studio,โ an extra layer that transformed Protodash into a browser-based โvibe coding platform.โ This studio environment allows users to interact with the LLM directly within their browser, without needing to open a separate coding editor. Users can manage their prototypes, view a โvibe prototyping feedโ for inspiration, and even remix othersโ prototypes. The most compelling feature is the embedded LLM, where a user can simply type a requestโlike โI want to add a new variant of my prototype where there chat a line chatโโand the AI will build it entirely in the browser.
Owen admits much of Protodash Studio was built through an iterative process of โyelling at Claude Code for 18 months,โ leveraging his engineering background to guide the AIโs architecture. The internal nature of the tool meant it didnโt need to be โproduction gradeโ in the traditional sense, allowing for rapid experimentation and feature addition.
Transforming the Design Review: Demos, Not Memos
One of the most significant impacts of Protodash has been on the design review process. Owen passionately advocates for โdemos not memos,โ a philosophy championed by other Stripe design leaders. He laments the โdrowning in presentationsโ of the past five years, where reviews involved static JPEGs and Figma screenshots. With Protodash, designers can share a URL, allowing everyone in the review to interact directly with a clickable, dynamic prototype. โCan I just say being in a design review where I can click things is my favorite,โ Owen enthuses, expressing his goal to โnever want to see a slideshow again.โ
Protodash Studio further enhances this by integrating a robust design review mode. Stakeholders can comment directly on the prototype at specific points, eliminating the need for fragmented Google Docs or lengthy email threads. These comments are then summarized by the AI, and crucially, they can be directly fed back to the AI for fixes. Owen describes the delight of being able to tell the AI, โThe filter pattern isnโt right here. Please add three more options,โ and then, after the AI implements the changes, responding to feedback providers with โI fixed Katie Dโs feedback for you. Hereโs the receipts thread.โ This drastically reduces the โbusy workโ associated with post-review follow-ups.
The PM Empowerment Revolution: A Surprising Twist
Perhaps the most unexpected and impactful outcome of Protodash is its adoption by Product Managers (PMs). Owen initially felt โa little nervousโ seeing PMs design, wondering, โwhatโs going to happen?โ However, this apprehension quickly turned to excitement. PMs can now articulate their ideas with unprecedented clarity by building high-quality, interactive prototypes that look and feel like Stripe products.
This empowerment has several profound benefits:
- Unblocking PMs: When a designer isnโt immediately available, PMs can explore ideas and build initial concepts themselves, preventing roadblocks.
- Better Communication: PMs can โmanifest the thing that they want to,โ leading to more concrete and productive conversations with their dedicated designers. Instead of abstract discussions, they can point to a functional prototype.
- Earlier User Research: With realistic prototypes, PMs can conduct user research much earlier in the product development cycle, gathering valuable feedback from actual users without the need for elaborate mockups or fully coded features.
- Stronger Advocacy: PMs can use high-fidelity prototypes to advocate more effectively for their projects, demonstrating the potential of an idea rather than just describing it.
The tool changes the nature of conversations, shifting them from arguments about staffing to discussions about the work itself. As Owen notes, it helps PMs โtalk the same languageโ as designers and engineers, especially for complex technical features like webhooks with multiple states.
Building a Bespoke Culture Tool
Owen strongly believes in the value of building internal tools that are precisely matched to a companyโs culture and cadence. While external tools like Vzero are excellent, they canโt be as deeply integrated or customized to Stripeโs specific needs, design system, and review processes. Protodash isnโt just a prototyping tool; itโs a cultural enabler.
The platform is designed for contribution, with Owen actively encouraging designers to submit pull requests to evolve its features. This fosters a sense of ownership and allows the tool to adapt organically to the changing needs of the design and product teams. The ability to quickly integrate a Google Doc Product Definition (PD) and have the AI build a prototype from it, or to enable โcrazy eightsโ mode for generative, out-of-the-box design exploration, highlights its flexibility. It even allows for temporary deviations from the core design system by granting the LLM access to other frameworks like Tailwind, enabling generative exploration beyond established components.
During a live demo, Owen illustrated this power by prompting Protodash to build a โBlack Friday Cyber Monday dashboard for a pet store showing live sales on a chart at the top of the page, a ticker with the latest sales and top performing products, with realistic real-time data.โ Within moments, the AI began assembling the page, calling on Stripeโs Sale MCP server to select appropriate components and structure the layout. This rapid, on-the-fly generation of complex, data-driven interfaces underscores the toolโs transformative potential.
Owenโs journey with Protodash at Stripe is a compelling case study in how internal tooling, powered by thoughtful AI integration, can fundamentally reshape product design and development. By addressing specific pain points, fostering collaboration, and empowering diverse teams, Stripe is not just building better products; itโs building a better way to build.
Based on โWhat Drives Political Violence in Americaโ from New York Times Podcasts Watch the original video
Beyond the Fringe: How Mainstream America Embraces Political Violence
In recent times, a disturbing drumbeat of political violence has echoed across the United States. From the murder of a state representative in Minnesota to an arson at a governorโs residence, and multiple reported assassination attempts against former President Donald Trump, these acts signal a dangerous escalation. The question that persistently arises is: how did we arrive at this point, and how much worse could it possibly get?
To dissect this unsettling trend, the New York Times spoke with Professor Robert Pape of the University of Chicago, a leading expert on political violence who has advised every White House from 2001 to 2024. His insights paint a sobering picture of an America grappling with a new and profound threat: โviolent populism.โ
The Age of Violent Populism: A Threat from Within
Professor Pape asserts that the most critical fact about political violence in America today is not the rise of foreign adversaries or terrorist groups, but the alarming acceptance of violence by tens of millions of Americans across the political spectrum. This widespread normalization, he argues, poses a greater risk to American democracy than any external threat.
What exactly is โviolent populismโ? Itโs a phenomenon where political violence is no longer confined to a radical fringe or a few militia groups, but finds significant social approval among a broad segment of the populace. This acceptance, Pape explains, has a two-fold dangerous effect:
- Empowering Volatile Individuals: For individuals already predisposed to anger or impulsivity due to their psychosocial reasons, the perceived social approval for violent acts can push them โover the edge.โ They see a prospect of validation for actions that would otherwise be condemned.
- Silencing the Alarms: When millions embrace political violence, it makes it incredibly difficult for law enforcement agencies like the FBI to receive crucial tips. People are more likely to โlook the other wayโ or discount what they see, eroding the informal intelligence network vital for preventing attacks.
This era, Pape clarifies, is distinct from a civil war, as there are no organized armies clashing. Yet, itโs far beyond โnasty politics as usual.โ It represents a spiraling cycle of violence fueled by broad social acceptance.
Alarming Statistics: A Nation Normalizing Assassination
Papeโs research, conducted through 20 nationally representative surveys over the last five years by the University of Chicago Project on Security and Threats (CPOS), reveals shocking figures on the social acceptance of political violence.
For much of the Biden administration, surveys found that 10% of the body politic supported political violence to restore Donald Trump to the presidency. Further focus groups revealed an even more chilling detail: 55% of those who endorsed the โuse of forceโ believed it meant assassination. This implies that at one point, roughly half of that 10%โor 20 million American adultsโfound assassination acceptable for political ends.
Disturbingly, this percentage has since doubled. The latest surveys indicate that between 14% and 21% of Americans now support some form of political violence. The acceptance isnโt limited to one side; when asked about the perceived acceptability of violence against figures like Charlie Kirk or the attempted assassination of Nancy Pelosi, 10% for each event found it acceptable. This translates to 20 million American adults on each side of the political divide.
In essence, roughly one in five Americans now supports some form of political violence, and this number is increasing on both sides of the spectrum.
Echoes of History, With a Dangerous Twist
While the current climate feels unprecedented, Pape points to other periods of โviolent populismโ in American history. The 1920s saw the resurgence of the second Ku Klux Klan, growing from hundreds of thousands to 4-6 million members who actively supported political violence, not just in the South but across the nation. The 1960s also witnessed a string of political assassinations and collective violence, like the 1968 Chicago convention riots, with opinion polls from the era showing similar levels of public support for such acts.
However, a critical difference sets the current era apart: โIt is pretty much the first time itโs happened on both sides at the same time.โ The 1920s violence was predominantly from the right, and the 1960s largely from the left. Today, the escalation is bipartisan, making it uniquely perilous.
Quantifying this escalation, Papeโs study of threats to members of Congress from 2001 to 2024, based on Department of Justice prosecutions, reveals a fivefold increase in threats since the first year of the Trump presidency in 2017. This surge affected both Democratic and Republican targets, indicating that Trumpโs rise acted as a โlightning rod on both sides.โ The dramatic spike has remained consistently high for an eight-year period, spanning both the Trump and Biden presidencies, suggesting that the underlying drivers are deeper than any single individual.
The Tectonic Shifts: Demographics and Wealth
At its core, this rise in political violence is not simply the impact of a president or social media, but the profound effect of social change. Pape identifies two major, interconnected social changes reshaping the United States:
-
The Demographic Tipping Point: For the first time in its 250-year history, America is transitioning from a white-majority democracy to a white-minority democracy. In 1960, non-Hispanic whites constituted about 88% of Americans; today, that figure is 57%, projected to be 49% in about 20 years. Immigration is a primary driver of this shift, making it the โnumber one lightning rod issue in America today.โ
- On the Right: The established group fears losing political power and declining economic prospects, leading to a desire to halt or reverse this change.
- On the Left: The emerging groups welcome and often seek to accelerate this change, as they stand to benefit. This fundamental upheaval directly fuels support for political violence, as shown in Papeโs surveys.
-
The Concentration of Wealth: Beginning in the mid-1980s, an accelerating shift of wealth to the top 1% has occurred, regardless of which party holds power. The bottom 90% of Americans are effectively subsidizing this enormous gain for the wealthiest few. Both Republican and Democratic constituents, comprising the bottom 90%, perceive their economic prospects being โclobbered even more.โ This unresolved economic disparity, ignored by both parties, further fuels the bipartisan nature of rising political violence.
These shifts create a sense of profound powerlessness and a fear of โpolitical exclusionโ or โlockoutโ for many. As groups feel their grievances cannot be addressed through conventional politics, and their very future seems existentially threatened, the appeal of more drastic measures grows.
The New Face of the Attacker: From Fringe to Mainstream
A persistent misunderstanding, Pape argues, is the profile of those who commit political violence. We often imagine them as the โlosers of societyโโmarginalized, unemployed individuals with little to lose. However, Papeโs analysis of the January 6th attackers, compared to right-wing political violence perpetrators from previous decades, reveals a starkly different picture.
The โold profileโ typically consisted of unemployed individuals, often members of militia groupsโthe โfringe.โ But among the hundreds arrested after January 6th, the profile was dramatically different:
- Unemployment rates were only at the national average (around 7%).
- Attackers included โbusiness owners, doctors, lawyers, CEOsโโinsurrectionists in business suits living in nice suburbs.
- Only about 10% were members of militia groups, a significant drop from previous findings.
This new profile indicates that the โhalvesโโeducated, middle-class, and even upper-middle-class individualsโare increasingly supporting political violence. Their motivation stems not from having nothing, but from the fear of losing what they have, of seeing their stature and influence decline rapidly. โWe like our monsters to be villains that are far away from us,โ Pape notes, explaining why this shift in attacker profile is so difficult for society to acknowledge.
From Rhetoric to Reality: The Role of Leaders and Social Media
While structural issues are the โfire,โ rhetoric acts as an accelerant. Pape highlights how top-down rhetoric from political leaders can push already predisposed individuals towards violence. He cites examples like Donald Trump famously suggesting โpunching people out,โ Beto OโRourke advocating to โthrow that punch first,โ and Gavin Newsom speaking of โpunching a bully back in the mouth.โ When authority figures normalize or even glorify aggressive action, others are more likely to follow suit.
Social media, while not the root cause, certainly โadds gasoline to a fire.โ It provides an immediate manifestation of the millions who support political violence, creating echo chambers where such acts are cheered on and lionized. Donald Trump Jr.โs social media post mocking the attack on Paul Pelosi, which generated โenormous social media support, lots of laughter,โ is a prime example of how top-down rhetoric can interact with bottom-up amplification, fostering a sense of social approval for would-be attackers. Yet, Pape cautions against overstating social mediaโs role, reminding us that previous eras of violent populism occurred long before the internet.
A Choice for a Better Future
Given these deep-seated drivers, what can be done? Pape outlines a two-stage approach:
1. Longer-Term Structural Solutions (Post-Midterms):
- Addressing Demographic Change: While not aiming to stop the demographic shift, Pape suggests โslowing it down so our political institutions can catch up.โ This, he clarifies, would mean adopting policies akin to the Obama-era immigration approach.
- Reversing Wealth Inequality: Pape argues for reducing the wealth shift to the top 1%, rather than merely stopping or continuing it. He notes that both parties currently cater to their donors, failing to address this core economic disparity.
2. Immediate Action (Before the Midterms):
- Leadersโ Joint Condemnation: In the short term, the most impactful action is for political leaders to jointly condemn political violence. Pape proposes a โradicalโ idea: Hakeem Jeffries (Democratic House leader) reaching out to Donald Trump for a joint video statement condemning violence from all sides.
- Evidence of Impact: Pape points to the effectiveness of such rhetoric in the past. After the first Trump assassination attempt in summer 2024, President Biden delivered a speech condemning violence, and surveys showed a 20% drop in support for political violence. A similar effect was observed after a second attempt, when Biden was joined by Kamala Harris. While not the sole cause, this rhetoric โcertainly didnโt make anything worse.โ
The overarching message, Pape emphasizes, is that despite all structural determinants and pressures, โthe future of political violence is a choice.โ While tens of millions support violence, a significant counterweight exists: 75% of the U.S. population โabhors political violence.โ This silent majority, Pape insists, must exert its agency. They need to actively voice their opposition to violence, sending letters and making calls to their representatives, demanding joint condemnation from leaders.
โLeaders listen to their constituents,โ Pape concludes. If their inboxes are filled only with demands to โhit the other side harder,โ thatโs what theyโll do. But if the 75% who oppose violence step up and make their voices heard, a different future is possible. It is a choice that โwe the peopleโ have to make, for the sake of the country and its future.
ํ๊ตญ์ด
โThe One Rule That Separates Real AI Agents from Demos | Yutori, Abhishek Dasโ โ EO ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
โ์ฒซ ์๋์ ์๋ํ์ง ์์ผ๋ฉด, ์ถฉ๋ถ์น ์๋คโ โ ์ ๋ขฐํ ์ ์๋ AI ์์ด์ ํธ์ ์๋๋ฅผ ์ด๋ค
ํ์ฌ AI ์์ด์ ํธ์ ํ๊ณ: โ์ฒซ ์๋์ ์๋ํ์ง ์์ผ๋ฉด ์ถฉ๋ถ์น ์๋คโ
์ค๋๋ ์ธ๊ณต์ง๋ฅ(AI) ๊ธฐ์ ์ ๋๋ถ์๊ฒ ๋ฐ์ ํ๊ณ ์์ง๋ง, ์น ๊ธฐ๋ฐ AI ์์ด์ ํธ(agent) ์ ํ๋ค์ ์ฌ์ ํ ์ ๋ขฐ์ฑ์ด๋ผ๋ ๊ทผ๋ณธ์ ์ธ ๋ฌธ์ ์ ์ง๋ฉดํด ์์ต๋๋ค. ์ ํ ๋ฆฌ(Yutori)์ ๊ณต๋ ์ฐฝ์ ์์ด์ ๊ณต๋ CEO์ธ ์๋น์ ๋ค์ค(Abhishek Das)๋ ์ด๋ฌํ ํ์ค์ ๋ํด ๊ฐํ ๋นํ์ ๋ชฉ์๋ฆฌ๋ฅผ ๋ ๋๋ค. ๊ทธ๋ ์๋ง์ ์์ด์ ํธ ์ ํ๋ค์ด โ์น์์ ๋ฌด์์ด๋ ํ ์ ์๋คโ๊ณ ์ฃผ์ฅํ์ง๋ง, ๋ง์ ์ฌ์ฉํด๋ณด๋ฉด ์ ๋๋ก ์๋ํ์ง ์๋ ๊ฒฝ์ฐ๊ฐ ํ๋ฐ์ด๋ผ๊ณ ์ง์ ํฉ๋๋ค.
๋ฌธ์ ์ ํต์ฌ์ ๋ณต์กํ ์์ ํ๋ฆ(workflow)์์ ๋ฐ์ํ๋ ์ค๋ฅ์ ๋์ (compounding error)์ ์์ต๋๋ค. ๋ค์ค์ ๋ฐ๋ฅด๋ฉด, 10๋จ๊ณ, 20๋จ๊ณ, ํน์ 50๋จ๊ณ๋ก ์ด๋ฃจ์ด์ง ์์ ํ๋ฆ์์ ๊ฐ ๋จ๊ณ์ ์ ํ๋๊ฐ 90%์ ๋ฌํ๋๋ผ๋, 10%์ ์ค๋ฅ์จ์ ๋งค์ฐ ๋น ๋ฅด๊ฒ ๋์ ๋์ด ์ ์ฒด ์์ ์ ์ฑ๊ณต๋ฅ ์ ํ์ ํ ๋ฎ์ถฅ๋๋ค. ์ด๋ โ์ฅ๊ธฐ์ ์ธ ์์ ํ๋ฆ(long horizon workflows)โ์ ์ํํ๊ธฐ ์ํ ๊ธฐ์ ์ด ์์ง ์ถฉ๋ถํ ๋ฐ์ ํ์ง ๋ชปํ์์ ์๋ฏธํฉ๋๋ค.
๋ค์ค CEO๋ ํนํ โ์ ํ ์ถ์ ๊ณผ์ ์์ ๋น๊ฒฐ์ ๋ก ์ (non-determinism)์ด๊ณ ๋ฎ์ ์ ๋ขฐ์ฑ(low reliability)์ ๋ํ ๊ด์ฉ์ด ์ผ๋ฐํ๋๊ณ ์๋ ๋ฏํ๋คโ๋ ์ ์ ๋ถ๋ง์ ํํฉ๋๋ค. โ์ฒซ ์๋์ ์ ๋๋ก ์๋ํ ๋งํผ ์ถฉ๋ถํ ์ข์ง ์๋ค๋ฉด, ๊ทธ๊ฒ์ ์ถฉ๋ถํ ์ข์ ์ ํ์ด ์๋๋คโ๋ผ๊ณ ๊ทธ๋ ๋จํธํ๊ฒ ๋งํฉ๋๋ค. ๊ทธ๋ ์ด๋ฌํ ๋ฌด์ฑ ์ํ ์ ํ ์ถ์ ๊ดํ์ด ์์ด์ ํธ ์ ํ ๋ถ์ผ์์๋ ๊ฒฐ์ฝ ์ฉ๋ฉ๋ ์ ์๋ค๊ณ ๊ฐ์กฐํ๋ฉฐ, ์ ๋ขฐ์ฑ์ด์ผ๋ง๋ก ์ง์ ํ ์ฐจ๋ณ์ ์ด๋ผ๊ณ ์ญ์คํฉ๋๋ค.
์ ํ ๋ฆฌ(Ytori)์ ๋น์ : ์น๊ณผ์ ์ํธ์์ฉ์ ์ฌ์ ์ํ๋ค
์ ํ ๋ฆฌ๋ ๋ฐ๋ก ์ด๋ฌํ ๋ฌธ์ ์์์์ ์ถ๋ฐํ์ต๋๋ค. ์ ํ ๋ฆฌ๋ ์ฌ์ฉ์ ๋์ ์น์์ ํ๋์ ์ทจํ๊ณ ์์ ์ ์๋ฃํ๋ ์ ๋ขฐํ ์ ์๋ ์์ด์ ํธ๋ฅผ ๊ตฌ์ถํ์ฌ, ์ฌ์ฉ์๊ฐ ์์ ์๊ฒ ๊ฐ์ฅ ์๋ฏธ ์๋ ์ผ์ ์ง์คํ ์ ์๋๋ก ๋๋ ๊ฒ์ ๋ชฉํ๋ก ํฉ๋๋ค. ์ธ ๋ช ์ ๊ณต๋ ์ฐฝ์ ์ ๋ชจ๋ AI ์ฐ๊ตฌ์ ์ถ์ ์ธ ์ ํ ๋ฆฌ๋ ๋จ์ํ ์์ด์ ํธ ํ์ฌ๋ฅผ ๋์ด, ์น๊ณผ์ ์ํธ์์ฉ ๋ฐฉ์์ ๊ทผ๋ณธ์ ์ผ๋ก ์ฌ์ ์ํ๋ ค๋ ํฐ ๋น์ ์ ํ๊ณ ์์ต๋๋ค.
๋ค์ค CEO๋ ์ง๋ 20~30๋ ๊ฐ ์น ๋ธ๋ผ์ฐ์ ์ ํํ๊ฐ ํฌ๊ฒ ๋ณํ์ง ์์๋ค๋ ์ ์ ์ง์ ํ๋ฉฐ, ์ด์ ๋ ์น ๊ฒฝํ์ ์ฌ๊ณ (reimagine)ํ ๊ธฐํ๊ฐ ์๋ค๊ณ ๋งํฉ๋๋ค. ๋ฏธ๋์๋ AI ๋น์์ ๋ํํ๋ฉฐ ์น์์ ์์ ์ ์ฒ๋ฆฌํ๊ณ , ๋ง์ ์์ด์ ํธ๊ฐ ์ฌ์ฉ์ ๋ชฐ๋ ๋ฐฑ๊ทธ๋ผ์ด๋์์ ๋ฅ๋์ ์ผ๋ก(proactive manner) ์๋ํ ๊ฒ์ด๋ผ๋ ์์ธก์ ๋๋ค. ๊ทธ๋ ๋ฌผ๋ฆฌ์ ์์ด์ ํธ(physical agents)๋ณด๋ค ๋์งํธ ์์ด์ ํธ(digital agents)๊ฐ ๋จผ์ ํ์คํ๋ ๊ฒ์ด๋ผ๊ณ ๋ด๋ค๋ด ๋๋ค.
๋ฏธ๋์ ์น ์ํธ์์ฉ์ ํจ์ฌ ๋์ ์ถ์ํ ์์ค(higher level of abstraction)์์ ์ด๋ฃจ์ด์ง ๊ฒ์ ๋๋ค. ์ฌ์ฉ์๊ฐ ์ง์ ๋์งํธ ํ๋๋ ์ผ(digital chores)์ ์๋์ผ๋ก ์ฒ๋ฆฌํ๋ ๋์ , AI ์์ด์ ํธ์๊ฒ ์ผ์์ ์ด๊ณ ๋ฐ๋ณต์ ์ธ ์์ ์ ์์ํจ์ผ๋ก์จ ๋์ฑ ์๋ฏธ ์๊ณ ํฅ๋ฏธ๋ก์ด ์ผ์ ์ง์คํ ์ ์๊ฒ ๋ฉ๋๋ค. ์ด๋ ์ธ๊ฐ๊ณผ ์์ด์ ํธ๊ฐ ํ๋ ฅํ์ฌ ์ ๋ฐ์ ์ธ ์์ฐ์ฑ์ ํฅ์์ํค๋ ๋ชจ๋ธ์ด๋ฉฐ, ์์ด์ ํธ๊ฐ ์ธ๊ฐ์ ๋์ฒดํ๋ ๊ฒ์ด ์๋๋ผ ์ธ๊ฐ์ ์ญ๋์ ๊ฐํํ๋ ๋ฐฉํฅ์ผ๋ก ๋์๊ฐ ๊ฒ์ ๋๋ค. ๋ํ, ์น ์ ๊ทผ์ฑ(accessibility)์ ๋์ด๋ ๋ฐ๋ ๊ธฐ์ฌํฉ๋๋ค. ์๋ฅผ ๋ค์ด, ๋ค์ค CEO๋ ์์ ์ ๋ถ๋ชจ๋์ด ๋งค๋ฒ ์๋ก์ด ์น์ฌ์ดํธ ์ฌ์ฉ๋ฒ์ ์ตํ ํ์ ์์ด, AI ๋น์์๊ฒ ์ํ๋ ๋ฐ๋ฅผ ๋งํ๋ฉด ์์ด์ ํธ๊ฐ ์ ๋ขฐ์ฑ ์๊ฒ ์์ ์ ์ฒ๋ฆฌํด ์ค ์ ์๋ค๋ฉด โ์ ๋ง ๋ฉ์ง ์ผโ์ด ๋ ๊ฒ์ด๋ผ๊ณ ์ค๋ช ํฉ๋๋ค.
์ฐฝ์ ์์ ์ฌ์ : ์ํ์์ AI๋ก, ๊ทธ๋ฆฌ๊ณ ์ง์ฐฉ์ ํ
์๋น์ ๋ค์ค CEO์ ๊ฐ์ธ์ ์ธ ์ฌ์ ์ ๊ทธ์ ํ๊ณ ํ ์ ํ ์ฒ ํ์ ์ดํดํ๋ ๋ฐ ์ค์ํ ๋ฐฐ๊ฒฝ์ด ๋ฉ๋๋ค. ๊ทธ๋ ์์ฌ์ ์๋ฃ์ธ ๊ฐ์กฑ์์ ํ์ด๋ฌ์ง๋ง, ์์ด๋ฌ๋ํ๊ฒ๋ ํผ๋ฅผ ๋ฌด์์ํ์ฌ ์ํ์ ๊ธธ์ ํฌ๊ธฐํ๊ณ ๊ณตํ์ ์ ํํ์ต๋๋ค. ๊ทธ๋ ๊ฐ์ค์ ์ธ์ฐ๊ณ , ์ด๋ฅผ ๊ฒ์ฆํ๊ฑฐ๋ ๋ฐ์ฆํ๊ธฐ ์ํ ์คํ์ ์ค๊ณํ๊ณ , ๊ฒฐ๋ก ์ ๋์ถํ์ฌ ๋ค์ ๊ฐ์ค์ ์ธ์ฐ๋ โ๊ณผํ์ ๊ณผ์ (scientific process)โ์ ๊น์ด ๋งค๋ฃ๋์์ต๋๋ค.
์ธ๋ ๊ณต๊ณผ๋ํ๊ต ๋ฃจ๋ฅดํค(IIT Rurki)์์ ํ๋ถ ๊ณผ์ ์ ๋ฐ๋ ์ฒซ ํด, ๋ค์ค๋ ์ข์ ์ฑ์ ์ ๋ฐ์์ง๋ง, ์ ๊ธฐ ๊ณตํ(ํนํ ์ ๋ ฅ ์์คํ ๋ถ์ผ)์ด ์์ ์ ํฅ๋ฏธ์ ๋ง์ง ์๋๋ค๋ ๊ฒ์ ๋น ๋ฅด๊ฒ ๊นจ๋ฌ์์ต๋๋ค. ๊ทธ๋ ์ด๋ฅผ โ์ธ์์ ์ค์ํ ๊ฐ๋ฆผ๊ธธ(significant fork)โ์ด์ โ์ฒซ ๋ฒ์งธ ๋ฐํญ์ ์ธ ๊ธฐ์ง(rebellious streak)โ์ด์๋ค๊ณ ํ์ํฉ๋๋ค. ๊ทธ๋ ์ ๊ธฐ ๊ณตํ์ ๋ํ ๊ด์ฌ์ ์ค์ด๊ณ ๋์ ํ๋ก๊ทธ๋๋ฐ๊ณผ ์ํํธ์จ์ด ๊ฐ๋ฐ์ ๋ง์ ์๊ฐ์ ํ ์ ํ๊ธฐ ์์ํ์ต๋๋ค.
๊ทธ์๊ฒ ํฐ ๋์์ด ๋ ๊ฒ์ IIT ๋ฃจ๋ฅดํค์ 1314๋
์ ๋ถํฐ ๊ฐ๋ ฅํ ํ๋ก๊ทธ๋๋ฐ ํด๋ฝ ๋ฌธํ๊ฐ ์์๋ค๋ ์ ์
๋๋ค. ํนํ โSDS Labsโ๋ผ๋ ๊ทธ๋ฃน์ ๋งค๋
1015๋ช
์ ์ฝ๋๋ค์ด ๋ชจ์ฌ ์บ ํผ์ค ๋ด ์ธํฐ๋ท์ ์ํ ์๋ง์ ์ ํ๋ฆฌ์ผ์ด์
์ ๋ง๋ค๊ณ ์คํํ์ต๋๋ค. ๋ค์ค CEO๋ โ์ฌ์ฉ์๋ค์ด ์ง์ ๋ง๋ ๊ฒ์ ์ฌ์ฉํ๊ณ , ์ํธ์์ฉํ๋ ๋ชจ์ต์ ๋ณด๋ ๊ฒ์ด ๋ํ๋ฏผ(dopamine)์ด ํฐ์ง๋ ๋ฏํ ์ฆ๊ฑฐ์์ด์๋คโ๊ณ ๋งํฉ๋๋ค. ๊ทธ๋ ๋๋ฃ๋ค๊ณผ ํจ๊ป ๋ฐค์ ์์
์ ํ๋ฉฐ ๊ธฐ๋ฅ ์ถ๊ฐ์ ๊ฐ์ ์ ๋ชฐ๋ํ๊ณ , ์ด๋ฌํ โ์ง์ฐฉ(obsession)โ์ด ๊ฐ๋ ฅํ ๋๊ธฐ ๋ถ์ฌ๊ฐ ๋์์ต๋๋ค.
๋ค์ค๋ ์ค๋ซ๋์ ์์ ๋ง์ ์ฌ์ ์ ์์ํ๊ณ ์ถ๋ค๋ ์ด๋ง์ ๊ฐ์ง๊ณ ์์์ต๋๋ค. ํ๋ถ ์กธ์ ํ์ ๋ฐ์ฌ ํ์ ์ทจ๋ ํ์๋ ์ฌ๋ฌ ๋ฒ ์ฐฝ์ ์ ๊ณ ๋ คํ์ง๋ง, ๊ฒฐ๊ตญ ์คํ์ ์ฎ๊ธฐ์ง ๋ชปํ์ต๋๋ค. ๊ทธ๋ฌ๋ ๊ทธ๋ โ์ธ์์๋ ํด๊ฒฐํด์ผ ํ ํฅ๋ฏธ๋ก์ด ๋ฌธ์ ๋ค์ด ๋ง๊ณ , ๋ค๋ฅธ ์ฌ๋์ ๋น์ ์ด ์๋ ๋ด๊ฐ ์ค์ํ๊ฒ ์๊ฐํ๋ ๋น์ ์ ์ถ์งํ๊ณ ์ถ์๋คโ๊ณ ๋งํ๋ฉฐ, ์ ํ ๋ฆฌ ์ฐฝ์ ์ ์๊ฐ ๋ฌธ์ ์๋ค๊ณ ์ค๋ช ํฉ๋๋ค.
์ ํ ์ฒ ํ: ์ ๋ขฐ, ์ง๊ด, ๊ทธ๋ฆฌ๊ณ โ๋ณด์ด์ง ์๋โ ๋ ธ๋ ฅ
์ ํ ๋ฆฌ๊ฐ ์ถ๊ตฌํ๋ ์ ํ ์ฒ ํ์ ์ ๋ขฐ์ฑ๊ณผ ์ฌ์ฉ์ ๊ฒฝํ์ ์ต์ฐ์ ์ผ๋ก ํฉ๋๋ค.
์ ๋ขฐ์ฑ ์ต์ฐ์ : ์ค๋ฅ ์ธ์๊ณผ ์ญ์ถ์ (Backtracking)
๋ค์ค CEO๋ ์์ด์ ํธ์ ์ ๋ขฐ์ฑ์ ํ๋ณดํ๊ธฐ ์ํด โ์ค๋ฅ ์ธ์โ๊ณผ โ์ญ์ถ์ (backtracking)โ ๋ฅ๋ ฅ์ด ๋งค์ฐ ์ค์ํ๋ค๊ณ ๊ฐ์กฐํฉ๋๋ค. ์์ด์ ํธ๊ฐ ์ค์๋ฅผ ์ ์ง๋ ์ ๋, ์ด๋ฅผ ์ค์ค๋ก ์ธ์งํ๊ณ ์ด์ ๋จ๊ณ๋ก ๋์๊ฐ ๋ค๋ฅธ ๊ฒฝ๋ก๋ฅผ ํ์ํ์ฌ ์ฌ๋ฐ๋ฅธ ์์ ์ ์ํํ๋ ๋ฅ๋ ฅ์ด ํ์์ ์ด๋ผ๋ ๊ฒ์ ๋๋ค.
์ ํ ๋ฆฌ๋ ์ด๋ฅผ ์ํด โํ๊ฐ(evals)โ์ โ์์ ์ฅ์น(guardrails)โ ๊ตฌ์ถ์ ๋ง์ ๋ ธ๋ ฅ์ ๊ธฐ์ธ์ ๋๋ค. ์ฌ์ฉ์๊ฐ ์คํํ๋ ๋ชจ๋ ํ๋ก๋์ ์ฟผ๋ฆฌ(production query)๋ ํฌ๊ด์ ์ธ ํ๊ฐ ์์คํ ์ ๊ฑฐ์ณ ์์ด์ ํธ๊ฐ ์ด๋ค ๋ถ๋ถ์์ ์ ์๋ํ๊ณ ์ด๋ค ๋๋ฉ์ธ์์ ๊ฐ์ ์ด ํ์ํ์ง ๋น ๋ฅด๊ฒ ์๋ณํฉ๋๋ค.
์น ์์ด์ ํธ์ ํน์ฑ์ ์ธ์์ ๋ชจ๋ ์น์ฌ์ดํธ๋ฅผ ํ์ต์ํค๋ ๊ฒ์ ๋ถ๊ฐ๋ฅํฉ๋๋ค. ์๋ก์ด ์น์ฌ์ดํธ๋ ๋์์์ด ์๊ฒจ๋๊ณ , ๊ธฐ์กด ์น์ฌ์ดํธ์ ์๋ ๋ฐฉ๋ํ๊ธฐ ๋๋ฌธ์ ๋๋ค. ๋ฐ๋ผ์ ์ ํ ๋ฆฌ๋ ์ ํ๋ ์น์ฌ์ดํธ ์ธํธ์์ ๋ชจ๋ธ์ ํ๋ จํ๊ณ ๊ฐ์ ํด๋๊ฐ๋๋ค. ์ธ๊ฐ๋ ์๋ก์ด ์น์ฌ์ดํธ์์ ์ค์๋ฅผ ์ ์ง๋ฅด๋ฏ, ๋ชจ๋ธ ์ญ์ ์ค์๋ฅผ ํ๋ ๊ฒ์ ์์ฐ์ค๋ฌ์ด ์ผ์ ๋๋ค. ๊ทธ๋ฌ๋ ์ค์ํ ๊ฒ์ ์ค์๋ฅผ ํ์ ๋, ์์ด์ ํธ๊ฐ ์ด๋ฅผ ์ธ์ํ๊ณ ์ค์ค๋ก ์์ ํ์ฌ ์ฌ๋ฐ๋ฅธ ๋ฐฉํฅ์ผ๋ก ๋์๊ฐ ์ ์๋ ๋ฅ๋ ฅ์ ๋๋ค. ์ด๊ฒ์ด ์ ํ ๋ฆฌ๊ฐ ๋ชจ๋ธ์ ํ๋ จํ๊ณ ๊ตฌ์ถํ๋ ํต์ฌ์ ์ธ ์์์ ๋๋ค.
์ฌ์ฉ์ ์ค์ฌ, ๊ทธ๋ฌ๋ ์ง๊ด์ ๋ํ๋ค
์ ํ ๋ฆฌ๋ ์ ํ ๊ฐ๋ฐ์ 80/20 ์ ๊ทผ ๋ฐฉ์์ ์ทจํฉ๋๋ค. ์ฆ, ์ฌ์ฉ์๊ฐ ์์ฒญํ๋ ๊ธฐ๋ฅ ์ธ์๋ ๊ฐ๋ฐํ์ โ์ง๊ด(intuition)โ์ ํตํด ์ ์ฌ์ ์ผ๋ก ์ฌ์ฉ์์๊ฒ ํฐ ๊ฐ์น๋ฅผ ์ ๊ณตํ ์ ์๋ ๊ธฐ๋ฅ์ ๋ฐ๊ตดํฉ๋๋ค. ๋ค์ค CEO๋ โ์ฌ์ฉ์๋ค์ด ๋ช ์์ ์ผ๋ก ์์ฒญํ์ง ์์ ์๋ ์์ง๋ง, ์ผ๋จ ๊ตฌํ๋๋ฉด ์ฌ์ฉ์๋ค์ โ์ฐ๋ฆฌ๋ฅผ ๋ณด๊ณ ์๊ตฌ๋โ๋ผ๊ณ ๋๋ผ๊ฒ ๋ ๊ฒโ์ด๋ผ๊ณ ์ค๋ช ํฉ๋๋ค.
๋ํ์ ์ธ ์์๋ก iOS๋ ์๋๋ก์ด๋์ 2๋จ๊ณ ์ธ์ฆ(two-factor authentication) SMS ์๋ ์์ฑ ๊ธฐ๋ฅ์ ๊ผฝ์ต๋๋ค. ์ฌ์ฉ์๊ฐ ์ด ๊ธฐ๋ฅ์ ์ง์ ์์ฒญํ๊ธฐ๋ ์ด๋ ต์ง๋ง, ์ด ์์ ๊ธฐ๋ฅ์ ์ ์ธ๊ณ ์๋ง์ ์ฌ๋๋ค์ ์ผ์์์ ํ๋ฃจ์๋ ์ฌ๋ฌ ๋ฒ ๋ช ์ด์ฉ ์๊ฐ์ ์ ์ฝํด ์ค๋๋ค. ๊ทธ๋ ์ด๋ฌํ โ์์ ๋ถํธํจ(tiny paper cuts)โ์ ์ค์ฌ์ฃผ๋ ์ธ์ฌํ ๋ฐฐ๋ ค๊ฐ ์ฌ์ฉ์์๊ฒ ํฐ ๋ง์กฑ๊ฐ์ ์ ์ฌํ๋ค๊ณ ๋งํฉ๋๋ค. ๋ฐ๋ผ์ ์ ํ ๋ฆฌ์ ์ ํ ๊ฐ๋ฐ์ ์ฌ์ฉ์์ ํผ๋๋ฐฑ๊ณผ ๊ฐ๋ฐํ์ ์ง๊ด์ด ์กฐํ๋ฅผ ์ด๋ฃจ๋ ๊ณผ์ ์ ๋๋ค.
โ๋งโ๊ณผ โ์ฅ์ธ์ ์ โ์ผ๋ก ์ฐจ๋ณํํ๋ค (Taste and Craft)
์ฝ๋ฉ LLM(coding LLM)์ ํ์ฉํด ์ด๊ธฐ ํ๋กํ ํ์ ์ ๋น ๋ฅด๊ฒ ๋ง๋๋ ๊ฒ์ด ์ฌ์์ง ์๋์, ์ ํ ๋ฆฌ๋ โ๋ง(taste)โ๊ณผ โ์ฅ์ธ์ ์ (craft)โ์ ์ง์ ํ ์ฐจ๋ณ์ ์ผ๋ก ์ผ์ต๋๋ค. ์ ํ์ด ์ผ๋ง๋ ์ง๊ด์ ์ด๊ณ ์ ์ค๊ณ๋์๋์ง๊ฐ ์ค์ํ๋ค๋ ๊ฒ์ ๋๋ค.
์ด๋ฅผ ์ํด ์ ํ ๋ฆฌ ํ์ด ์ค์ํ๊ฒ ์ฌ๊ธฐ๋ ๊ดํ ์ค ํ๋๋ โ์์ฌ ์ ํ ์ฌ์ฉ(dogfooding)โ์ ๋๋ค. ๋งค์ฃผ ํ๋ ์๊ฐ์ฉ ์๋ก์ด ๊ธฐ๋ฅ์ ์ง์ ์ฌ์ฉํด๋ณด๋ ์๊ฐ์ ๊ฐ์ง๋ฉฐ, ํญ์ ์์ญ ๊ฐ์ง์ ๋ด๋ถ ์คํ์ ์งํํฉ๋๋ค. ์ด ์ค ์ค์ ๋ก ์ธ๋ถ ์ฌ์ฉ์์๊ฒ ์ ๊ณต๋๋ ํ๋ก๋์ ๋ฒ์ ์ผ๋ก ์ถ์๋๋ ๊ฒ์ ๊ทนํ ์ผ๋ถ์ ๋๋ค. ์ด๋ฌํ ์ง์์ ์ธ ์์ฌ ์ ํ ์ฌ์ฉ์ โ๋ฌด์์ด ์ข๊ณ ๋์์งโ, โ์ด๋ค ๊ฒฝํ์ด ๋๋๊ณ ๋ง๋ฒ ๊ฐ์์งโ์ ๋ํ ํ์ โ๋งโ์ ๋ค๋ฌ๋ ๊ณผ์ ์ ๋๋ค. ๋ค์ค CEO๋ ๋ค๋ฅธ ๋ชจ๋ ๊ธฐ์ ๊ณผ ๋ง์ฐฌ๊ฐ์ง๋ก, ์ด๋ฌํ โ๊ทผ์กโ์ ํค์ฐ๊ธฐ ์ํด์๋ ์๋ง์ ๋ฐ๋ณต(a lot of reps)์ด ํ์ํ๋ค๊ณ ๊ฐ์กฐํฉ๋๋ค.
ํฌ๋ช ์ฑ์ผ๋ก ์๋ ์ ๋ขฐ: โ์์ ์ฆ๋ช โ์ ์ค์์ฑ
AI ๋ชจ๋ธ์ ๊ฒฝ์ฐ, ์ต์ข ์์ธก์ด๋ ๋ต๋ณ๋ฟ๋ง ์๋๋ผ โ์์ ์ฆ๋ช (proof of work)โ, ์ฆ ์ต์ข ๊ฒฐ๊ณผ๊ฐ ๋์ถ๋๊ธฐ๊น์ง์ ๊ณผ์ ์ ๋ณด์ฌ์ฃผ๋ ๊ฒ์ด ์ค์ํฉ๋๋ค. ๋ค์ค CEO๋ ๊ณผ๊ฑฐ ์์ ์ ์ฐ๊ตฌ ๊ฒฝํ์์ ๋์จ โ๊ทธ๋ ๋์บ (Grad-CAM)โ ํ๋ก์ ํธ๋ฅผ ์๋ก ๋ค์ด ์ค๋ช ํฉ๋๋ค. ๊ทธ๋ ๋์บ ์ ๋ฅ๋ฌ๋ ๋ชจ๋ธ์ด ์ด๋ฏธ์ง ๋ถ๋ฅ ์์ธก์ ํ ๋, ์ด๋ฏธ์ง์ ์ด๋ ๋ถ๋ถ์ ๋ณด๊ณ ๊ทธ๋ฌํ ํ๋จ์ ๋ด๋ ธ๋์ง ์๊ฐ์ ์ผ๋ก ๋ณด์ฌ์ฃผ๋ ๊ธฐ์ ์ ๋๋ค. ๋น์ 25์ธ์๋ ๋ค์ค๊ฐ ๊ณต๋ ์ ์๋ก ์ฐธ์ฌํ ์ด ๋ ผ๋ฌธ์ ํด์ ๊ฐ๋ฅ์ฑ(interpretability) ๋ถ์ผ์์ 2๋ง~3๋ง ํ์ ์ธ์ฉ์ ๊ธฐ๋กํ๋ฉฐ ํฐ ๋ฐํฅ์ ์ผ์ผ์ผฐ์ต๋๋ค.
๊ทธ๋ ๊ทธ๋ ๋์บ ์ฒ๋ผ AI ๋ชจ๋ธ์ด โ์โ ๊ทธ๋ฌํ ๊ฒฐ๋ก ์ ๋๋ฌํ๋์ง ๋ณด์ฌ์ฃผ๋ ๊ฒ์ด ์ฌ์ฉ์ ์ ๋ขฐ๋ฅผ ๊ตฌ์ถํ๋ ๋ฐ ํต์ฌ์ด๋ผ๊ณ ๋งํฉ๋๋ค. ์ ํ ๋ฆฌ์ ์ ํ์์๋ ์ด๋ฌํ ์ฒ ํ์ด ๋ฐ์๋์ด ์์ต๋๋ค. ์ฌ์ฉ์๊ฐ ์น์ ๋ชจ๋ํฐ๋งํ๋๋ก ์ค์ ํ โ์ ์ฐฐ ์์ด์ ํธ(scouts and agents)โ๊ฐ ๋ณด๊ณ ์๋ฅผ ์์ฑํ ๋, ์ฌ์ฉ์ ์ธํฐํ์ด์ค(UI) ๋ด์ ๋ฒํผ์ ํตํด ์์ด์ ํธ๊ฐ ์ด๋ค ์น์ฌ์ดํธ๋ฅผ ๋ฐฉ๋ฌธํ๊ณ , ์ด๋ค ์ ๋ณด๋ฅผ ๋ฐํ์ผ๋ก ๋ณด๊ณ ์๋ฅผ ์์ฑํ๋์ง ์ง์ ํ์ธํ ์ ์์ต๋๋ค.
๋ค์ค CEO๋ โ์ฌ์ฉ์๋ค์ด ๋ณผ ์ ์๋ ์ ํ์ ๋ถ๋ถ์ ์ธ์ฌํ ์ฃผ์๋ฅผ ๊ธฐ์ธ์ด๋ฉด, ์ฌ์ฉ์๋ ๋ณด์ด์ง ์๋ ์ ํ์ ๋ถ๋ถ๋ ๋ ์ ๋ขฐํ ๊ฐ๋ฅ์ฑ์ด ๋๋คโ๊ณ ๋งํฉ๋๋ค. ์ธ์์ ๋ชจ๋ ์๋ํ ๊ฒ๋ค์ด ๊ฐ์ธ์ด๋ ๊ทธ๋ฃน์ ์์ฒญ๋ ๋ ธ๋ ฅ๊ณผ ์ธ์ฌํ ์ฃผ์๋ฅผ ํตํด ๋ง๋ค์ด์ง๋ฏ, ์ ํ ๋ฆฌ ์ญ์ ๋ชจ๋ ์ ํ ๊ฐ๋ฐ์ ์ด๋ฌํ ์ฒ ํ์ผ๋ก ์ ๊ทผํฉ๋๋ค. ์๋ฏธ ์๋ ๊ฒ์ ๋ง๋ค๊ณ , ์ฌ๋ฐ๋ฅด๊ฒ ๋ง๋ค๊ณ , ๋ฏธ๋์ ๋น์ ์ ํ์ค๋ก ๋ง๋ค๋ฉฐ, ์ฆ๊ฒ๊ณ ์ ๋ขฐํ ์ ์๋ ์ ํ ๊ฒฝํ์ ๊ตฌ์ถํ๋ ๋ฐ๋ ์๊ฐ๊ณผ ๋ ธ๋ ฅ์ด ํ์ํ๋ฉฐ, ๊ฒฐ์ฝ ์ฐ์ฐํ ๋ํ๋๋ ๊ฒ์ด ์๋๋ผ๋ ๊ฒ์ด ๊ทธ์ ๋ณ์น ์๋ ์ ๋ ์ ๋๋ค.
โThe internal AI tool thatโs transforming how Stripe designs products | Owen Williamsโ โ How I AI ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
์คํธ๋ผ์ดํ์ ๋์์ธ ํ์ : AI ๊ธฐ๋ฐ ๋ด๋ถ ํด โํ๋กํ ๋์โ๊ฐ ์ ํ ๊ฐ๋ฐ ๋ฌธํ๋ฅผ ๋ฐ๊พธ๋ ๋ฐฉ๋ฒ
์คํธ๋ผ์ดํ(Stripe)๋ ์ฌ์ฉ์ ๊ฒฝํ(UX)๊ณผ ๋์์ธ ํ์ง์ ๋ํ ๋์ ๊ธฐ์ค์ผ๋ก ์ ์๋ ค์ ธ ์์ต๋๋ค. ํ์ง๋ง ์ด๋ฌํ ๋ช ์ฑ์๋ ๋ถ๊ตฌํ๊ณ , ์ ํ ๋์์ธ ๊ณผ์ ์์ ํ๋กํ ํ์ดํ์ ์ฌ์ ํ ๋ง์ ์ด๋ ค์์ ์๊ณ ์์์ต๋๋ค. ํนํ ๋ฐ์ดํฐ๊ฐ ๋ณต์กํ๊ฒ ์ฝํ ๋์๋ณด๋ ๊ฐ์ ์ ํ์ ๊ธฐ์กด ํด๋ก๋ ๊ฑฐ์ ๊ตฌํ ๋ถ๊ฐ๋ฅ์ ๊ฐ๊น์ ์ต๋๋ค. ์คํธ๋ผ์ดํ์ ๋์์ธ ๋งค๋์ ์ธ ์ค์ฌ ์๋ฆฌ์์ค(Owen Williams)๋ ์ด๋ฌํ ๋ฌธ์ ๋ฅผ ํด๊ฒฐํ๊ธฐ ์ํด AI ๊ธฐ๋ฐ์ ๋ด๋ถ ํ๋กํ ํ์ดํ ํด โํ๋กํ ๋์(Protodash)โ๋ฅผ ๊ฐ๋ฐํ์ต๋๋ค. ์ด ํด์ ๋จ์ํ ๋์์ธ ํ๋ก์ธ์ค๋ฅผ ํจ์จํํ๋ ๊ฒ์ ๋์ด, ์คํธ๋ผ์ดํ์ ์ ํ ๊ฐ๋ฐ ๋ฌธํ๋ฅผ ๊ทผ๋ณธ์ ์ผ๋ก ๋ณํ์ํค๊ณ ์์ต๋๋ค.
๋ฌธ์ ์ ๊ธฐ: โํ ์ผ์๋ ์ธ๋๊ณ ์ฌ๋กญโ๊ณผ ํผ๊ทธ๋ง์ ํ๊ณ
์ค์ฌ์ ๋์์ธ ๋ฆฌ๋ทฐ์ ์ฐธ์ฌํ๋ฉฐ ๋์์ด๋๋ค์ด ๊ฒช๋ ๊ณ ์ถฉ์ ์ง์ ๋ชฉ๊ฒฉํ์ต๋๋ค. ๋์์ด๋๋ค์ Vzero์ ๊ฐ์ ์ธ๋ถ AI ํด์ ํ์ฉํด ํ๋กํ ํ์ ์ ๋ง๋ค๋ ค ํ์ง๋ง, ๊ทธ ๊ฒฐ๊ณผ๋ฌผ์ ์คํธ๋ผ์ดํ์ ์๊ฒฉํ ๋์์ธ ์์คํ ๊ณผ๋ ๋๋จ์ด์ง โํ ์ผ์๋ ์ธ๋๊ณ ์ฌ๋กญ(Tailwind indigo slop)โ์ ๋ถ๊ณผํ์ต๋๋ค. ์ด๋ ์ธ๋ถ AI ํด์ด ์คํธ๋ผ์ดํ์ ๊ณ ์ ํ ๋์์ธ ์์คํ ์ ์ดํดํ์ง ๋ชปํด ๋ฐ์ํ๋, ์ด๋๊ฐ ์ด์ํ๊ณ ์ผ๋ฐ์ ์ธ ๋๋์ ๊ฒฐ๊ณผ๋ฌผ์ ์๋ฏธํฉ๋๋ค. ๋ด๋น๊ฒ์ด์ , ํฐํธ ๋ฑ์ด ๋ฏธ๋ฌํ๊ฒ ์ด๊ธ๋ ๋ชฐ์ ์ ๋ฐฉํดํ๋ ๊ฒฝ์ฐ๊ฐ ๋ง์์ต๋๋ค.
๋ ํฐ ๋ฌธ์ ๋ ๋ฐ์ดํฐ๊ฐ ํต์ฌ์ธ ์คํธ๋ผ์ดํ์ ์ ํ ํน์ฑ์ด์์ต๋๋ค. ์๋ง์ ์ํธ์์ฉ, ํํฐ, ๋ค์ํ ๋ฐ์ดํฐ ์ํ(๋ฐ์ดํฐ๊ฐ ์ ํ ์๋ โ์ ๋ก ์ํโ๋ถํฐ ๋ฐฉ๋ํ ๋ฐ์ดํฐ, ํน์ ๋น์ฆ๋์ค ๋ชจ๋ธ์ ๋ฐ๋ฅธ ๋ฐ์ดํฐ ๋ณํ, ๊ตญ์ ํ์ ๋ฐ๋ฅธ ํ ์คํธ ๊ธธ์ด ๋ณํ ๋ฑ)๋ฅผ ๊ฐ์ง ๋์๋ณด๋๋ฅผ ํผ๊ทธ๋ง(Figma) ๊ฐ์ ๊ธฐ์กด ๋์์ธ ํด๋ก ํ๋กํ ํ์ดํํ๋ ๊ฒ์ โ๊ฑฐ์ ๋ถ๊ฐ๋ฅ์ ๊ฐ๊น์ดโ ์ผ์ด์์ต๋๋ค. ๊ณผ๊ฑฐ์๋ โ๋ก๋ ์ ์จ(Lorem Ipsum)โ ํ ์คํธ๋ ๊ฐ์ง ๊ธ์ก์ ์๋์ผ๋ก ์ ๋ ฅํด์ผ ํ๊ณ , ๋ค๋จ๊ณ ํ๋ก์ฐ๋ ์๋ฌ ์ํ๋ฅผ ์๊ฐํํ๋ ๊ฒ์ ์์ฒญ๋ ๋ ธ๋ ฅ์ ์๊ตฌํ์ต๋๋ค. ์ด๋ฌํ ํ๊ณ๋ ๋์์ด๋๋ค์ด ์ค์ ์ฌ์ฉ ์๋๋ฆฌ์ค๋ฅผ ์ถฉ๋ถํ ํ์ํ๊ณ , ์ ํ์ ํ์ค์ ์ธ ๋ฌธ์ ๋ฅผ ์กฐ๊ธฐ์ ๋ฐ๊ฒฌํ๋ ๊ฒ์ ์ด๋ ต๊ฒ ๋ง๋ค์์ต๋๋ค.
๋ํ, ๋๋ถ๋ถ์ ๋์์ธ ๋ฆฌ๋ทฐ๊ฐ ์ ์ ์ธ JPEG ์ด๋ฏธ์ง๋ ์ฌ๋ผ์ด๋์ผ ํํ๋ก ์งํ๋๋ฉด์, ์ ํ์ ์ค์ ๋์์ ํ์ ํ๊ณ ์ฌ์ธต์ ์ธ ํผ๋๋ฐฑ์ ์ฃผ๊ณ ๋ฐ๋ ๋ฐ ํ๊ณ๊ฐ ์์์ต๋๋ค. ์ค์ฌ์ ์ด๋ฌํ ๋นํจ์จ์ฑ์ ๊ทน๋ณตํ๊ณ ์ถ์์ต๋๋ค.
ํด๊ฒฐ์ฑ : ์ค์ฌ์ โํ๋กํ ๋์โ ํ์
์์ง๋์ด๋ง ๋ฐฐ๊ฒฝ์ ๊ฐ์ง ์ค์ฌ ์๋ฆฌ์์ค๋ ์ด๋ฌํ ๋ฌธ์ ์ ๊น์ด ๊ณต๊ฐํ์ต๋๋ค. ๊ทธ๋ ๊ธฐ์ ์ ์์์ ๊ฐ์ถ ๋์์ด๋๋ฅผ ์ ํธํ๋ฉฐ, โํฐ๋ฏธ๋์ ๋ค๋ฃฐ ์ค ์๋โ ๋์์ด๋๊ฐ ๊ฐ๋ ฅํ ์ํผํ์๋ฅผ ๊ฐ์ง ์ ์๋ค๊ณ ๋ฏฟ์์ต๋๋ค. ํ์ง๋ง ๋์์ ๋์์ด๋๋ค์ด ์น ๊ฐ๋ฐ์ ๋ณต์กํ ๊ธฐ์ ์คํ(npm, Vit, React ๋ฑ)์ ์ข์ ํ๋ ๊ฒ์ ๋ณด๋ฉฐ, ๊ธฐ์ ์ฅ๋ฒฝ์ ๋ฎ์ถ๋ ๊ฒ์ด ์ค์ํ๋ค๊ณ ์๊ฐํ์ต๋๋ค. AI์ ๋ฑ์ฅ์ ์ด๋ฌํ ๊ทธ์ ์ฒ ํ์ ์ ํ์ ์ด ๋์์ต๋๋ค. ์ด์ ๋์์ด๋๋ค์ ๋ณต์กํ ๋ช ๋ น์ด๋ฅผ ๋ชฐ๋ผ๋ AI์๊ฒ ์ง์ ์ง๋ฌธํ๋ฉฐ ํ์ํ ์์ ์ ์ํํ ์ ์๊ฒ ๋ ๊ฒ์ ๋๋ค.
์ค์ฌ์ ์ด๋ฌํ ๋ฐฐ๊ฒฝ์ ๋ฐํ์ผ๋ก ๋ด๋ถ ํ๋กํ ํ์ดํ ํด์ธ โํ๋กํ ๋์(Protodash)โ๋ฅผ ๊ฐ๋ฐํ๊ธฐ ์์ํ์ต๋๋ค. ์ด๋ ์คํธ๋ผ์ดํ์ ๋์์ธ ์์คํ ์ธ โ์ธ์ผ(Sale)โ ์ปดํฌ๋ํธ์ MCP(Message Control Plane) ์๋ฒ ํตํฉ, ๊ทธ๋ฆฌ๊ณ ์ปค์(Cursor)์ ๊ฐ์ AI ์ฝ๋ฉ ํด์ ํนํ๋ ๊ท์น ๋ฒ๋ค๋ก ๊ตฌ์ฑ๋ ๋ฆฌ์กํธ(React) ๊ธฐ๋ฐ์ ์น ์ฑ์ด์์ต๋๋ค. ์ด ๊ท์น ๋ฒ๋ค์ AI์๊ฒ ํ๋ก์ ํธ ์ฌ์ฉ๋ฒ, Figma ๋งํฌ ์ฒ๋ฆฌ ๋ฐฉ๋ฒ, ๋์์ธ ์์คํ ์ปดํฌ๋ํธ ํ์ฉ๋ฒ, ํํ ์ค์ ๋ฐฉ์ง๋ฒ ๋ฑ์ ๊ฐ๋ฅด์ณ, ์คํธ๋ผ์ดํ์ ๋์ ๋์์ธ ํ์ง ๊ธฐ์ค์ ์ถฉ์กฑํ๋ ํ๋กํ ํ์ ์ 90% ์ ๋ ์๋์ผ๋ก ์์ฑํ ์ ์๋๋ก ํ์ต๋๋ค.
๋ฐ๋ธ๋ฐ์ค(Devbox) ์ธํ๋ผ๋ฅผ ํตํ ์ ๊ทผ์ฑ ํฅ์
์ด๊ธฐ์๋ ํ๋กํ ๋์๋ฅผ ๋ก์ปฌ์์ ์คํํ์ง๋ง, ์คํธ๋ผ์ดํ์ โ๋ฐ๋ธ๋ฐ์ค(Devbox)โ ์ธํ๋ผ๋ฅผ ํ์ฉํ๋ฉด์ ์ ๊ทผ์ฑ์ด ํ์ ์ ์ผ๋ก ๊ฐ์ ๋์์ต๋๋ค. ๋ฐ๋ธ๋ฐ์ค๋ ํด๋ผ์ฐ๋ ๊ธฐ๋ฐ์ ๊ฐ๋ฐ ํ๊ฒฝ์ผ๋ก, ๋์์ด๋๋ ๋ณต์กํ ๋ก์ปฌ ์ค์ ์ด๋ npm run dev์ ๊ฐ์ ๋ช
๋ น์ด๋ฅผ ์ง์ ์
๋ ฅํ ํ์ ์์ด ์น URL(์: go/protodash)๋ง์ผ๋ก ๋ช ๋ถ ์์ ํ๋กํ ํ์
ํ๊ฒฝ์ ์์ฑํ๊ณ ์คํํ ์ ์๊ฒ ๋์์ต๋๋ค. ์ด๋ ๋ก์ปฌ ํ๊ฒฝ์ ๋๋ฆฐ ์๋๋ ๋๊ท๋ชจ ๋ชจ๋
ธ๋ ํฌ(monorepo) ๊ด๋ฆฌ์ ์ด๋ ค์์ ํด์ํ์ผ๋ฉฐ, ์์ฑ๋ URL์ ํตํด ๋์์ธ ๋ฆฌ๋ทฐ์์ ํ๋ฉด ๊ณต์ ์์ด๋ ์ค์ ๋์ํ๋ ํ๋กํ ํ์
์ ๊ณต์ ํ๊ณ ํด๋ฆญํ๋ฉฐ ํผ๋๋ฐฑ์ ๋ฐ์ ์ ์๊ฒ ํ์ต๋๋ค. ์ค์ฌ์ ์ด๋ฅผ โ๋ฐ๋ชจ๊ฐ ๊ณง ๋ฉ๋ชจ(demos not memos)โ๋ผ๋ ์คํธ๋ผ์ดํ์ ๋์์ธ ๋ฆฌ๋์ญ ์ฒ ํ์ ์คํํ๋ ํต์ฌ์ ์ธ ์์๋ก ๊ผฝ์์ต๋๋ค.
๋ฐ์ดํฐ ์ค์ฌ ๋์์ธ์ ํ์
ํ๋กํ ๋์๋ ํผ๊ทธ๋ง๋ก๋ ์๋๋ ๋ด์ง ๋ชปํ๋ ๋ค์ํ ๋ฐ์ดํฐ ์๋๋ฆฌ์ค๋ฅผ ๋น ๋ฅด๊ฒ ์๋ฎฌ๋ ์ด์ ํ ์ ์๊ฒ ํ์ต๋๋ค. ์๋ฅผ ๋ค์ด, ๋ฐ์ดํฐ๊ฐ ์ ํ ์๋ โ์ ๋ก ์ํโ๋ถํฐ, ๋๊ธฐ์ ๊ณผ ์คํํธ์ ์ ๋น์ฆ๋์ค ๋ชจ๋ธ์ ๋ฐ๋ฅธ ๋ฐ์ดํฐ ๋ณํ, ๋ณต์กํ ๊ตญ์ ํ(internationalization)์ ๋ฐ๋ฅธ ํ ์คํธ ๊ธธ์ด ๋ณํ, ์ฌ์ง์ด โ์ง์ ๋ถํ ๋ฐ์ดํฐ(messy data)โ ์๋๋ฆฌ์ค๊น์ง ์ฆ๊ฐ์ ์ผ๋ก ํ์ธํ ์ ์์ต๋๋ค. ์ด๋ ๋์์ด๋๊ฐ ์ค์ ์ฌ์ฉ์๊ฐ ์ ํ์ ์ด๋ป๊ฒ ๊ฒฝํํ ์ง ๋ ๋นจ๋ฆฌ ์ดํดํ๊ณ , ๋ค์ํ ๋น์ฆ๋์ค ์ผ์ด์ค๋ฅผ ๊ณ ๋ คํ ๊ฒฌ๊ณ ํ ๋์์ธ์ ๋ง๋ค ์ ์๋๋ก ๋์ต๋๋ค.
์ค์ฌ์ โ๋ก๋ ์ ์จ(Lorem Ipsum) ํ ์คํธ๋ฅผ ์๋์ผ๋ก ๋ถ์ฌ ๋ฃ๊ณ ๊ฐ์ง ๊ธ์ก์ ์ ๋ ฅํ๋ ๊ณผ๊ฑฐ์๋ ๋น๊ตํ ์ ์๋ ํ์ โ์ด๋ผ๊ณ ๊ฐ์กฐํฉ๋๋ค. ์ด์ ๋์์ด๋๋ ๋ค๋จ๊ณ ํ๋ก์ฐ, ์๋ฌ ์ํ, ๋ค์ํ ์ํธ์์ฉ ๋ฑ ๋ณต์กํ ์ฌ์ฉ์ ์ฌ์ ์ ํจ์ฌ ๋ ์ฝ๊ณ ํ์ค์ ์ผ๋ก ํ๋กํ ํ์ดํํ ์ ์๊ฒ ๋์์ผ๋ฉฐ, ์ด๋ ์ ํ์ ๋ชจ๋ ์์๊ฐ ์๋์ ์ผ๋ก ์ค๊ณ๋์๋์ง ๊ฒํ ํ๋ ๋ฐ ํฐ ๋์์ด ๋ฉ๋๋ค.
์งํํ๋ โํ๋กํ ๋์ ์คํ๋์คโ: ๋ธ๋ผ์ฐ์ ๊ธฐ๋ฐ์ ํ์ ํ๋ซํผ
์ค์ฌ์ ๊ถ๊ทน์ ์ธ ๋ชฉํ๋ Vzero์ ๊ฐ์ด ๋ธ๋ผ์ฐ์ ์์ ์ง์ ์คํธ๋ผ์ดํ ๋ง์ถคํ ํ๋กํ ํ์ ์ ๊ตฌ์ถํ ์ ์๋ ํ๊ฒฝ์ ๋ง๋๋ ๊ฒ์ด์์ต๋๋ค. ์ด๋ฅผ ์ํด ๊ทธ๋ โํ๋กํ ๋์ ์คํ๋์ค(Protodash Studio)โ๋ผ๋ ์ถ๊ฐ ๋ ์ด์ด๋ฅผ ๊ฐ๋ฐํ์ต๋๋ค. ์ด๋ ๋ฐ๋ธ๋ฐ์ค ์์์ ๋์ํ๋ฉฐ, ๋์์ด๋๊ฐ ๋ณ๋์ ์ฑ ์ค์น ์์ด ์น ๋ธ๋ผ์ฐ์ ๋ง์ผ๋ก ํ๋กํ ํ์ ์ ๋ง๋ค๊ณ ๊ด๋ฆฌํ ์ ์๊ฒ ํฉ๋๋ค.
์ฃผ์ ๊ธฐ๋ฅ ๋ฐ ํ์ ๋ฐฉ์
- ๋ธ๋ผ์ฐ์ ๋ด ์๋ฒ ๋๋ LLM(Large Language Model) ์๋ํฐ: ๋์์ด๋๋ ์ฝ๋๋ฅผ ์ง์ ๋ค๋ฃจ์ง ์๊ณ ๋ ์์ฐ์ด๋ก ํ๋กํ ํ์ ์ ์์ ํ๊ณ ๋ค์ํ ๋ณํ(variant)์ ๋ง๋ค ์ ์์ต๋๋ค. ์๋ฅผ ๋ค์ด, โ๋ผ์ธ ์ฐจํธ(line chart)๊ฐ ์๋ ์๋ก์ด ํ๋กํ ํ์ ๋ณํ์ ์ถ๊ฐํด์คโ๋ผ๊ณ ์์ฒญํ๋ฉด AI๊ฐ ์๋์ผ๋ก ์ฝ๋๋ฅผ ์์ฑํ๊ณ ์ ์ฉํฉ๋๋ค.
- ํ๋กํ ํ์ ํผ๋ ๋ฐ ๋ฆฌ๋ฏน์ค: ๋ค๋ฅธ ๋์์ด๋๊ฐ ๋ง๋ ํ๋กํ ํ์ ์ ์ฝ๊ฒ ์ฐพ์๋ณด๊ณ , ๊ทธ๋ค์ ์ ๊ทผ ๋ฐฉ์์์ ์๊ฐ์ ์ป๊ฑฐ๋ ์ฝ๋๋ฅผ โ๋ฆฌ๋ฏน์ค(remix)โํ์ฌ ์์ ์ ์์ ์ ํ์ฉํ ์ ์์ต๋๋ค. ์ด๋ ์ง์ ๊ณต์ ์ ํ์ ์ ์ด์งํฉ๋๋ค.
- ๋ค์ํ ๋์์ธ ์์คํ ์ง์: ์คํธ๋ผ์ดํ์ โ์ธ์ผโ ๋์์ธ ์์คํ ์ธ์๋ ํ ์ผ์๋(Tailwind)์ ๊ฐ์ ์ธ๋ถ ํ๋ ์์ํฌ๋ฅผ LLM์ ํ์ฉํ์ฌ, ๋์์ธ ์์คํ ์ ์ ์ฝ ์์ด ์์ ํ ์๋ก์ด ๊ฐ๋ ์ ํ์ํ ์ ์๋ ์ ์ฐ์ฑ์ ์ ๊ณตํฉ๋๋ค.
- AI ํผ๋๋ฐฑ ๋ฐ ์๋ ์์ : โ์ฃผ์ AI(annotate AI)โ ๊ธฐ๋ฅ์ ํตํด ๋์์ด๋๋ ํ๋กํ ํ์ ์บ๋ฒ์ค์์ ํน์ ์์๋ฅผ ์ง์ ์ ํํ๊ณ , ํดํ ์ถ๊ฐ, ํจ๋ฉ ์กฐ์ ๋ฑ ๊ตฌ์ฒด์ ์ธ ํผ๋๋ฐฑ์ AI์ ์ ๋ฌํ ์ ์์ต๋๋ค. AI๋ ์ด ํผ๋๋ฐฑ์ ๋ฐํ์ผ๋ก ์ฝ๋๋ฅผ ์์ ํ๊ณ , ์ฌ์ง์ด ์ค์ค๋ก ์ค๋ฅ๋ฅผ ๊ฐ์งํ๊ณ ์์ ํ๋ โ์ ํ ํ ์คํธ(self-test)โ ๊ธฐ๋ฅ๊น์ง ๊ฐ์ถ๊ณ ์์ต๋๋ค.
- ๋์์ธ ๋ฆฌ๋ทฐ ๋ชจ๋: ์คํธ๋ผ์ดํ์ ๊ณ ์ ํ ๋์์ธ ๋ฆฌ๋ทฐ ๋ฌธํ๋ฅผ ๋ฐ์ํ์ฌ, ์ฌ๋ฌ ์ฐธ์ฌ์๊ฐ ํ๋กํ ํ์ ์ ์ง์ ๋๊ธ์ ๋ฌ๊ณ ํผ๋๋ฐฑ์ ๊ณต์ ํ ์ ์์ต๋๋ค. AI๋ ์ด ํผ๋๋ฐฑ์ ์๋์ผ๋ก ์์ฝํ์ฌ ์์ ์์ ๋๊ธฐ์ด์ ์ถ๊ฐํ๋ฉฐ, ๋์์ด๋๋ ๋ฆฌ๋ทฐ ํ์ ์์ ์ ๋ํญ ์ค์ด๊ณ , โ์์ ์๋ฃโ ์ฆ๋ช ์๊น์ง AI์๊ฒ ๋ฐ์ ์ ์์ด ์ ๋ฌด ๋ถ๋ด์ ํฌ๊ฒ ์ค์์ต๋๋ค.
๋ด๋ถ ํด ๊ตฌ์ถ์ ๊ฐ์น์ ๋ฌธํ์ ๋ณํ
์ค์ฌ์ ํ๋กํ ๋์ ์คํ๋์ค๋ฅผ ๊ฐ๋ฐํ๋ฉฐ โ๋ด๋ถ ํด์ ์์ฐ ๋ฑ๊ธ(production grade)์ผ ํ์๊ฐ ์๋คโ๋ ์ฒ ํ์ ๊ฐ์ง๊ณ ์์์ต๋๋ค. ์ด๋ ์คํ๊ณผ ๋น ๋ฅธ ๋ฐ๋ณต์ ๊ฐ๋ฅํ๊ฒ ํ๊ณ , ๋ฒ๊ทธ๊ฐ ๋ฐ์ํด๋ ์ฉ์ธ๋๋ ์ ์ฐํ ๊ฐ๋ฐ ํ๊ฒฝ์ ์กฐ์ฑํ์ต๋๋ค.
๋์์ด๋์ ๊ธฐ์ฌ์ ๋ฌธํ ์งํ
ํ๋กํ ๋์๋ ๋จ์ํ ํจ์จ์ฑ์ ๋์ด๋ ๊ฒ์ ๋์ด, ์กฐ์ง์ ๋ฌธํ์ ๋ณํ๋ฅผ ์ด๋์ด๋์ต๋๋ค. ์ค์ฌ์ โ๋๊ตฌ๋ ๊ธฐ์ฌํ ์ ์๋คโ๋ ์์น ์๋, ๋์์ด๋๋ค์ด ์ง์ ํ ๋ฆฌํ์คํธ(pull request)๋ฅผ ๋ณด๋ด ํด์ ๊ฐ์ ํ๋๋ก ์ฅ๋ คํฉ๋๋ค. ์ด๋ ๋์์ธ ๋ฆฌ๋ทฐ ๋ฐฉ์์ด๋ ์์ ํ๋ฆ์ ๋ํ ๋ถ๋ง์ ํด์ํ๊ณ , ๋ฌธํ๋ฅผ ๋ฅ๋์ ์ผ๋ก ์งํ์ํค๋ ๊ฐ๋ ฅํ ๋๊ธฐ๊ฐ ๋ฉ๋๋ค. ๋์์ด๋๋ค์ ์ด์ โ๋์์ธ ๋ฆฌ๋ทฐ๊ฐ ์ ๋๋ก ์๋ํ์ง ์์. ๋ฌด์์ด ํ์ํ ๊น?โ๋ผ๋ ์ง๋ฌธ์ ๋ํด ์ง์ ์๋ฃจ์ ์ ๊ตฌํํ ์ ์๊ฒ ๋ ๊ฒ์ ๋๋ค.
PM(Product Manager)์ ์ญ๋ ๊ฐํ
๊ฐ์ฅ ํฅ๋ฏธ๋ก์ด ๋ณํ ์ค ํ๋๋ PM(Product Manager)์ ์ญํ ๋ณํ์ ๋๋ค. ์ค์ฌ์ ์ฒ์์๋ PM์ด ๋์์ธํ๋ ๊ฒ์ ๋ํด โ์ฝ๊ฐ ๋ถ์๊ฐโ์ ๋๊ผ์ง๋ง, ์ด์ ๋ ์ด๋ฅผ ๊ธ์ ์ ์ผ๋ก ํ๊ฐํฉ๋๋ค. PM์ ๋์์ด๋์ ๋์ ์์ด๋ ์คํธ๋ผ์ดํ ๊ธฐ์ค์ ๋ง๋ ๊ณ ํ์ง ํ๋กํ ํ์ ์ ๋ง๋ค ์ ์๊ฒ ๋์๊ณ , ์ด๋ PM ์ค์ค๋ก๊ฐ ์์ด๋์ด๋ฅผ ํ์ํ๊ณ ๊ตฌ์ฒดํํ๋ ๋ฐ ํฐ ๋์์ด ๋ฉ๋๋ค.
- ์ํต ๊ฐ์ : PM์ ์์ ์ ์์ด๋์ด๋ฅผ ์๊ฐ์ ์ผ๋ก ๊ตฌํํจ์ผ๋ก์จ ๋์์ด๋์ ๋ ํจ๊ณผ์ ์ผ๋ก ์ํตํ ์ ์๊ฒ ๋ฉ๋๋ค.
- UXR(User Experience Research) ์กฐ๊ธฐ ์งํ: ์ ์ ์ธ ๊ทธ๋ฆผ์ด ์๋ ์ค์ ๋์ํ๋ ํ๋กํ ํ์ ์ผ๋ก ์ฌ์ฉ์ ํ ์คํธ๋ฅผ ์กฐ๊ธฐ์ ์งํํ์ฌ, ์ ํ ์์ด๋์ด๋ฅผ ๊ฒ์ฆํ๊ณ ๋ถํ์ํ ๊ฐ๋ฐ ๋ฆฌ์คํฌ๋ฅผ ์ค์ผ ์ ์์ต๋๋ค.
- ํ๋ก์ ํธ ์นํธ: ๊ณ ํ์ง ํ๋กํ ํ์ ์ PM์ด ์์ ์ ํ๋ก์ ํธ์ ํ์ํ ๋ฆฌ์์ค(์: ๋์์ด๋ ์ธ๋ ฅ)๋ฅผ ๋ ์ค๋๋ ฅ ์๊ฒ ์นํธํ๊ณ ํ๋ณดํ๋ ๋ฐ ๊ธฐ์ฌํฉ๋๋ค.
- ๋ณต์กํ ๊ฐ๋ ์๊ฐํ: ๊ฐ๋ฐ์ ๊ฒฝํ(Developer Experience) ๋ถ์ผ์ ์นํ (webhooks)์ฒ๋ผ ๋ณต์กํ ๊ธฐ์ ์ ๊ฐ๋ ๋ ๋ค์ํ ์ํ์ ํ๋ก์ฐ๋ฅผ ํตํด ์๊ฐ์ ์ผ๋ก ๋ช ํํ๊ฒ ์ค๋ช ํ ์ ์๊ฒ ๋์์ต๋๋ค.
๊ฒฐ๋ก : ์คํธ๋ผ์ดํ์ ๋ฏธ๋ ๋์์ธ, AI์ ๋ด๋ถ ํด์ด ์ด๋๋ ํ์
์คํธ๋ผ์ดํ์ โํ๋กํ ๋์โ ์ฌ๋ก๋ AI ์๋์ ๋ด๋ถ ํด์ด ๋จ์ํ ํจ์จ์ฑ ์ฆ๋๋ฅผ ๋์ด, ์กฐ์ง์ ํ์ ๋ฐฉ์, ๋ฌธํ, ๊ทธ๋ฆฌ๊ณ ๊ถ๊ทน์ ์ผ๋ก ์ ํ ๊ฐ๋ฐ์ ๋ณธ์ง์ ์ด๋ป๊ฒ ๋ณํ์ํฌ ์ ์๋์ง ๋ณด์ฌ์ฃผ๋ ๊ฐ๋ ฅํ ์ฆ๊ฑฐ์ ๋๋ค. ์ธ๋ถ ์๋ฃจ์ ์ด ์๋, ํ์ ๊ณ ์ ํ ๋ฌธํ์ ์์ ๋ฐฉ์์ ์ ํํ ๋ง์ถฐ์ง ํด์ ๊ตฌ์ถํจ์ผ๋ก์จ ์คํธ๋ผ์ดํ๋ ๋ ๋์ ํ์ง๊ณผ ํจ์จ์ฑ์ ๋ฌ์ฑํ๊ณ ์์ต๋๋ค.
์ค์ฌ ์๋ฆฌ์์ค๋ ํ๋กํ ๋์๋ฅผ ํตํด ๋์์ด๋์ PM์ด ๊ธฐ์ ์ ์ฅ๋ฒฝ ์์ด ์ฐฝ์์ ์ธ ์์ด๋์ด๋ฅผ ์์ ๋กญ๊ฒ ํ์ํ๊ณ , ์ ํ์ ๋ ํ์ค์ ์ผ๋ก ์๋ฎฌ๋ ์ด์ ํ๋ฉฐ, ํ์ ์ ์ง์ ๋์ด๋ ์๋ก์ด ์๋๋ฅผ ์ด์์ต๋๋ค. ์ด๋ ์คํธ๋ผ์ดํ๊ฐ ์์ผ๋ก๋ ํ์ ์ ์ธ ์ ํ์ ์ง์์ ์ผ๋ก ๋ง๋ค์ด๋ผ ์ ์๋ ๊ฐ๋ ฅํ ๊ธฐ๋ฐ์ด ๋ ๊ฒ์ ๋๋ค.
์ด ์ํผ์๋๋ ์ ๋ฆฌ๊ณ (Celigo)์ ์ปค์(Cursor)์ ํ์์ผ๋ก ์ ์๋์์ต๋๋ค.
โWhat Drives Political Violence in Americaโ โ New York Times Podcasts ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
๋ฏธ๊ตญ์ ๋คํ๋๋ โํญ๋ ฅ์ ํฌํฐ๋ฆฌ์ฆโ: ์จ๊ฒจ์ง ๋ ๊ฐ์ง ์์ธ๊ณผ ํด๊ฒฐ์ฑ
์ต๊ทผ ๋ฏธ๊ตญ์ ์ ์น์ ์ธ๋ฌผ๋ค์ ํฅํ ํญ๋ ฅ ํ์์ ๋ฌผ๊ฒฐ์ ํฉ์ธ์ฌ ์์ต๋๋ค. ๋ฏธ๋ค์ํ์์๋ ์ฃผ ์์ ๋ถ๋ถ๊ฐ โ์ ์น์ ๋๊ธฐ๊ฐ ์๋ ๊ณต๊ฒฉโ์ผ๋ก ์ดํด๋นํ๊ณ , ํ์ค๋ฒ ์ด๋์ ์ฃผ์ง์ฌ ๊ด์ ์๋ ๋ฐฉํ๊ฐ ๋ฐ์ํ์ผ๋ฉฐ, ๋๋๋ ํธ๋ผํ ์ ๋ํต๋ น์ ๋ํ ์ธ ์ฐจ๋ก์ ์์ด ์๋๊น์ง ์ด์ด์ก์ต๋๋ค. ์ด์ฒ๋ผ ์ถฉ๊ฒฉ์ ์ธ ์ฌ๊ฑด๋ค์ด ์ฐ์ด์ด ํฐ์ง๋ฉด์, โ์ด์ฉ๋ค ์ฌ๊ธฐ๊น์ง ์๋๊ฐ, ๊ทธ๋ฆฌ๊ณ ์ํฉ์ ์ผ๋ง๋ ๋ ๋๋น ์ง ์ ์๋๊ฐ?โ๋ผ๋ ์ง๋ฌธ์ด ๋ฏธ๊ตญ ์ฌํ ์ ๋ฐ์ ๊ฑธ์ณ ์ ๊ธฐ๋๊ณ ์์ต๋๋ค.
์ด๋ฌํ ๋ฌผ์์ ๋ตํ๊ธฐ ์ํด, ๋ด์ํ์์ค ํ์บ์คํธ โ๋ ๋ฐ์ผ๋ฆฌ(The Daily)โ๋ ์์ญ ๋ ๊ฐ ์ ์น ํญ๋ ฅ ๋ถ์ผ์ ์ต๊ณ ์ ๋ฌธ๊ฐ๋ก ํ๋ํ๋ฉฐ 2001๋ ๋ถํฐ 2024๋ ๊น์ง ๋ชจ๋ ๋ฐฑ์ ๊ด์ ์๋ฌธํด์จ ์์นด๊ณ ๋ํ๊ต ๋ก๋ฒํธ ํ์ดํ(Robert Pape) ๊ต์๋ฅผ ์ธํฐ๋ทฐํ์ต๋๋ค. ํ์ดํ ๊ต์๋ ์ข์ฐ๋ฅผ ๋ง๋ก ํ๊ณ ํญ๋ ฅ์ด ์ฆ๊ฐํ๋ ์ด์ ์ ์ด๋ฅผ ๋ฉ์ถ๊ธฐ ์ํ ๋ฐฉ์์ ๋ํด ์ฌ์ธต์ ์ธ ๋ถ์์ ๋ด๋์์ต๋๋ค.
โํญ๋ ฅ์ ํฌํฐ๋ฆฌ์ฆ(Violent Populism)โ์ ์๋
ํ์ดํ ๊ต์๋ ํ์ฌ ๋ฏธ๊ตญ์ด โํญ๋ ฅ์ ํฌํฐ๋ฆฌ์ฆ(Violent Populism)โ์ ์๋์ ์ง์ ํ๋ค๊ณ ์ง๋จํฉ๋๋ค. ์ด๋ ๊ทธ ์ด๋ค ํ ๋ฌ ์ง๋จ์ด๋ ์ธ๊ตญ ๊ตญ๊ฐ๋ณด๋ค๋ ๋ฏธ๊ตญ ๋ฏผ์ฃผ์ฃผ์์ ๋ ํฐ ์ํ์ด ๋๋ค๋ ์ฃผ์ฅ์ ๋๋ค. ๊ทธ์ ๋ฐ๋ฅด๋ฉด, ์ค๋๋ ๋ฏธ๊ตญ ์ ์น ํญ๋ ฅ์ ๊ฐ์ฅ ์ค์ํ ์ฌ์ค์ **โ์์ฒ๋ง ๋ช ์ ๋ฏธ๊ตญ์ธ์ด ์์ชฝ ์ง์ ๋ชจ๋์์ ์ ์น์ ํญ๋ ฅ์ ์ฉ๋ฉํ ๋งํ๋ค๊ณ ๋ณธ๋คโ**๋ ์ ์ ๋๋ค.
์ด๋ ์์์ ๊ทน๋จ์ฃผ์์๋ ๋ฏผ๋ณ๋ ์ง๋จ์ ๊ตญํ๋ ํ์์ด ์๋๋๋ค. ์์ฒ๋ง ๋ช ์ ์๋ฏผ์ด ํญ๋ ฅ์ ์ฉ์ธํ๋ ์ํฉ์ ๊ณต๊ฒฉ์ ์ํ์ ๋ํ ๋ชจ๋ ๊ฒ์ ๋ณํ์ํต๋๋ค. ๊ฐ์ธ์ ์ธ ์ฌ๋ฆฌ ์ฌํ์ ์ด์ ๋ก ๋ถ์์ ํ ์ฌ๋๋ค์ด ์ฌํ์ ์น์ธ์ ๊ธฐ๋ํ๋ฉฐ ํญ๋ ฅ์ ๊ฒฝ๊ณ์ ์ ๋๊ฒ ๋ง๋ญ๋๋ค. ๋ํ, ์์ฒ๋ง ๋ช ์ด ์ ์น ํญ๋ ฅ์ ์ง์งํ๋ฉด FBI์ ์ ๋ณดํ๋ ๊ฒ์กฐ์ฐจ ๊ทน๋๋ก ์ด๋ ค์์ง๋๋ค. ํญ๋ ฅ์ ๋ํ ์ง์ง๊ฐ ๋ง์์๋ก ์ฌ๋๋ค์ ๋จ์ํ ๋ชป ๋ณธ ์ฒํ๊ฑฐ๋, ๋ชฉ๊ฒฉํ ๊ฒ์ ๋์๋กญ์ง ์๊ฒ ์ฌ๊ธฐ๊ธฐ ๋๋ฌธ์ ๋๋ค.
ํ์ดํ ๊ต์๋ ์ด๋ฌํ ํ์์ด ์ขํ์ ์ฐํ ์์ชฝ ๋ชจ๋์์ ๋ํ๋๋ฉฐ, ์ด์ ๋ ๋จ์ํ ์ฆ๊ฐํ๋ ๊ฒ์ ๋์ด **โ๋์ ํ์ผ๋ก ์ฆํญ๋๊ธฐ ์์ํ๋คโ**๊ณ ๊ฒฝ๊ณ ํฉ๋๋ค. ๊ทธ๋ ์ต๊ทผ 12๊ฐ์๊ฐ์ ์ฌ๋ก๋ก 2025๋ 6์ ๋ฏธ๋ค์ํ ๋ฏผ์ฃผ๋น ์ง๋์ ์์ด, ๋ช ๋ฌ ํ ์ฐฐ๋ฆฌ ์ปคํฌ(Charlie Kirk) ์์ด, ํ์ค๋ฒ ์ด๋์ ์ฃผ์ง์ฌ ๊ด์ ๋ฐฉํ, ๊ทธ๋ฆฌ๊ณ ์ฌ๋ฌ ์ฐจ๋ก์ ํธ๋ผํ ์ ๋ํต๋ น ์์ด ์๋๋ฅผ ์ธ๊ธํ๋ฉฐ, ์ ์น ํญ๋ ฅ์ด ์์ชฝ ์ง์์์ ๋ถ๋ช ํ ์ฆ๊ฐํ๊ณ ์์์ ๊ฐ์กฐํ์ต๋๋ค.
๋ฐ์ดํฐ๋ก ๋ณธ ์ ์น ํญ๋ ฅ์ ํ์ฐ
ํ์ดํ ๊ต์ ์ฐ๊ตฌํ์ ์ง๋ 5๋ ๊ฐ 20์ฐจ๋ก์ ๊ฑธ์ณ ์ ๊ตญ ๋ํ์ฑ ์๋ ๋ฏธ๊ตญ์ธ ์ค๋ฌธ์กฐ์ฌ๋ฅผ ์ค์ํ์ต๋๋ค. ์์นด๊ณ ๋ํ๊ต ์๋ณด ๋ฐ ์ํ ํ๋ก์ ํธ(CPOS, University of Chicago Project on Security and Threats)์์ 2021๋ ์ฌ๋ฆ๋ถํฐ ์์๋ ์ด ์กฐ์ฌ๋ ๋ฏธ๊ตญ์ธ์ ์ ์น ํญ๋ ฅ์ ๋ํ ํ๋๋ฅผ ์ฌ์ธต์ ์ผ๋ก ํ์ ํ์ต๋๋ค.
์กฐ์ฌ ๊ฒฐ๊ณผ๋ ์ถฉ๊ฒฉ์ ์ ๋๋ค.
- ๋ฐ์ด๋ ํ์ ๋ถ ์ด๊ธฐ ๋ช ๋ ๊ฐ, ๋ฏธ๊ตญ ์ ์น ์ฐธ์ฌ ์ธ๊ตฌ์ **10%**๊ฐ ๋๋๋ ํธ๋ผํ๋ฅผ ๋ํต๋ น์ผ๋ก ๋ณต๊ถ์ํค๊ธฐ ์ํ ์ ์น ํญ๋ ฅ์ ์ง์งํ์ต๋๋ค.
- ์ด ์ค 55%๋ โ๋ฌด๋ ฅ ์ฌ์ฉ(use of force)โ์ด โ์์ด(assassination)โ์ ์๋ฏธํ๋ค๊ณ ์๊ฐํ์ต๋๋ค. ์ฆ, ๋น์ ์ ์น ์ฐธ์ฌ ์ธ๊ตฌ์ ์ ๋ฐ์ธ 10%๊ฐ ์์ด์ ์ง์งํ ๊ฒ์ ๋๋ค.
- ํ์ฌ ๊ทธ ๋น์จ์ 14%์์ 21% ์ฌ์ด๋ก ๊ฑฐ์ ๋ ๋ฐฐ ์ฆ๊ฐํ์ต๋๋ค. ํธ๋ผํ ์ ๋ํต๋ น ๋น์ ์งํ๋ถํฐ ์ด ์์น๋ ๋ ๋ฐฐ๊ฐ ๋์๊ณ , ์ดํ์๋ ์ค์ด๋ค์ง ์๊ณ ์์ต๋๋ค.
- 2025๋ 9์, ์ฐฐ๋ฆฌ ์ปคํฌ ์์ด๊ณผ ๋ธ์ ํ ๋ก์ ์์ด ์๋์ ๋ํด โ์ฉ๋ฉํ ๋งํ๋คโ๊ณ ์๋ตํ ๋น์จ์ ๊ฐ๊ฐ 10%์์ต๋๋ค. ์ด๋ ์ฝ 2์ฒ๋ง ๋ช ์ ์ฑ์ธ ๋ฏธ๊ตญ์ธ์ด ๊ฐ ์ฌ๊ฑด์ ์ฉ์ธํ๋ค๋ ์๋ฏธ์ ๋๋ค.
์ข ํฉ์ ์ผ๋ก ๋ณผ ๋, ํ์ฌ ๋ฏธ๊ตญ์ธ 5๋ช ์ค 1๋ช ๊ผด๋ก ์ด๋ค ํํ์ ์ ์น ํญ๋ ฅ์ ์ง์งํ๊ณ ์์ผ๋ฉฐ, ์ด ์์น๋ ์คํํธ๋ผ์ ์์ชฝ์์ ๋ชจ๋ ์ฆ๊ฐํ๊ณ ์์ต๋๋ค.
๊ณผ๊ฑฐ์๋ ๋ค๋ฅธ ํ์ฌ์ ํญ๋ ฅ
ํ์ดํ ๊ต์๋ ๋ฏธ๊ตญ์ด ๊ณผ๊ฑฐ์๋ โํญ๋ ฅ์ ํฌํฐ๋ฆฌ์ฆโ ์๊ธฐ๋ฅผ ๊ฒช์๋ค๊ณ ์ค๋ช ํฉ๋๋ค.
- 1920๋ ๋: ์ 1์ฐจ ์ธ๊ณ๋์ ์ดํ ์ 2์ฐจ ์ฟ ํด๋ญ์ค ํด๋(Ku Klux Klan)์ด ์์ญ๋ง ๋ช ์์ 400๋ง~600๋ง ๋ช ์ผ๋ก ๊ธ์ฆํ๋ฉฐ ์๋ฐฑ๋ง ๋ช ์ ๋ฏธ๊ตญ์ธ์ด ์ ์น ํญ๋ ฅ์ ์ง์งํ์ต๋๋ค.
- 1960๋ ๋: ์ผ๋ จ์ ์ ์น์ ์์ด ์ฌ๊ฑด๊ณผ ํจ๊ป 1968๋ ์์นด๊ณ ์ ๋น๋ํ์ ๊ฐ์ ์ง๋จ ์ ์น ํญ๋ ฅ๊ณผ ํญ๋์ด ๋ฐ์ํ์ผ๋ฉฐ, ๋น์ ์ฌ๋ก ์กฐ์ฌ์์๋ ์์ฒ๋ง ๋ช ์ ํญ๋ ฅ ์ง์ง๊ฐ ๋ํ๋ฌ์ต๋๋ค.
๊ทธ๋ฌ๋ ํ์ฌ์ ์ํฉ์ ๊ณผ๊ฑฐ์ ์ค์ํ ์ฐจ์ด์ ์ด ์์ต๋๋ค. ๊ณผ๊ฑฐ์๋ ์ฃผ๋ก ํ์ชฝ ์ง์์์ ํญ๋ ฅ์ ํฌํฐ๋ฆฌ์ฆ์ด ๋ฐ์ํ์ง๋ง, ํ์ฌ๋ ์์ชฝ ์ง์์์ ๋์์ ๋ฐ์ํ๊ณ ์๋ค๋ ์ ์ ๋๋ค. 1960๋ ๋๋ ์ฃผ๋ก ์ขํ์์, 1920๋ ๋ ์ด๋ฐ์ ์ฃผ๋ก ์ฐํ์์ ํญ๋ ฅ์ด ๋ฐ์ํ์ต๋๋ค.
์์น๋ก ๋ด๋ ์ ์น ํญ๋ ฅ์ ์ฆ๊ฐ๋ ๋ช ํํฉ๋๋ค. ํ์ดํ ๊ต์์ ์ฐ๊ตฌํ์ด 2001๋ ๋ถํฐ 2024๋ ๊น์ง 25๋ ๊ฐ ์ฐ๋ฐฉ์์ฌ๊ตญ(FBI) ๋ฐ์ดํฐ๋ฅผ ํฌํจํ ๋ค์ํ ๋ฐ์ดํฐ๋ฒ ์ด์ค๋ฅผ ๋ถ์ํ ๊ฒฐ๊ณผ, ํธ๋ผํ ๋ํต๋ น ์ฌ์ ์ฒซํด์ธ 2017๋ ๋ถํฐ ์์๋ค์ ๋ํ ์ํ์ด 5๋ฐฐ ์ฆ๊ฐํ์ต๋๋ค. ์ด ์ฆ๊ฐ๋ ๋ฏผ์ฃผ๋น๊ณผ ๊ณตํ๋น ์์ ๋ชจ๋์๊ฒ ํด๋นํ๋ฉฐ, ํธ๋ผํ ์ ๋ํต๋ น์ ๋ถ์์ด ์์ชฝ ์ง์ ๋ชจ๋์๊ฒ โํผ๋ขฐ์นจโ ์ญํ ์ ํ์์ ๋ณด์ฌ์ค๋๋ค. ํธ๋ผํ ์์ ๋ฅผ ๊ณผ๋ํ๊ฒ ์ง์งํ๊ฑฐ๋(์: ๋ง์กฐ๋ฆฌ ํ ์ผ๋ฌ ๊ทธ๋ฆฐ), ์ถฉ๋ถํ ์ง์งํ์ง ์๋ ์์๋ค ๋ชจ๋๊ฐ ํ์ ์ด ๋๋ฉด์, ์ํ ์์ค์ ๊ธ๊ฒฉํ ์์นํ์ฌ ์ฝ 8๋ ๊ฐ ํธ๋ผํ์ ๋ฐ์ด๋ ํ์ ๋ถ ๋ด๋ด ๊ทธ ๋์ ์์ค์ ์ ์งํ์ต๋๋ค. ๋ํต๋ น์ด ๋ฐ๋์ด๋ ์ํ์ด ์ค์ด๋ค์ง ์์๋ค๋ ๊ฒ์ ๋จ์ํ ํน์ ์ธ๋ฌผ(ํธ๋ผํ) ์ด์์ ๋ฌธ์ ๊ฐ ์์์ ์์ฌํฉ๋๋ค.
์ ์น ํญ๋ ฅ์ ๊ทผ๋ณธ ์์ธ: ๋ ๊ฐ์ง ๊ฑฐ๋ํ ์ฌํ ๋ณํ
ํ์ดํ ๊ต์๋ ์ด๋ฌํ ํญ๋ ฅ ์ฆ๊ฐ์ ๊ทผ๋ณธ ์์ธ์ด ๋จ์ํ ๊ฐ์ธ์ด๋ ์์ ๋ฏธ๋์ด์ ์ํฅ์ด ์๋๋ผ๊ณ ๊ฐ์กฐํฉ๋๋ค. ๊ทธ๋ **โ์ฌํ ๋ณํ์ ์ํฅโ**์ด ๊ฐ์ฅ ์ค์ํ๋ค๊ณ ๋งํฉ๋๋ค. ๊ตญ๊ฐ์์ ์ฃผ์ ์ฌํ ๋ณํ๊ฐ ๋ฐ์ํ๋ฉด ์ ์น์ ๊ธ์งํ๋ฅผ ์ด๋ํ๊ณ , ์ด๋ ๋ค์ ์ ์น ํญ๋ ฅ์ ์ ๋ฐํ ์ ์๋ค๋ ๊ฒ์ ๋๋ค. ํ์ฌ ๋ฏธ๊ตญ์์ ์ผ์ด๋๊ณ ์๋ ๋ ๊ฐ์ง ์ฃผ์ ์ฌํ ๋ณํ๋ ๋ค์๊ณผ ๊ฐ์ต๋๋ค.
1. ์ธ๊ตฌํต๊ณํ์ ๋ณํ (Demographic Shift): ๋ฐฑ์ธ ๋ค์์์ ์์๋ก
๋ฏธ๊ตญ 250๋ ์ญ์ฌ์ ์ฒ์์ผ๋ก ๋ฐฑ์ธ ๋ค์ ๋ฏผ์ฃผ์ฃผ์์์ ๋ฐฑ์ธ ์์ ๋ฏผ์ฃผ์ฃผ์๋ก ์ ํํ๊ณ ์๋ค๋ ์ ์ ๋๋ค. 1960๋ ์๋ ๋ฏธ๊ตญ์ธ์ ์ฝ 88%๊ฐ ๋นํ์คํจ๋๊ณ ๋ฐฑ์ธ์ด์์ง๋ง, ์ค๋๋ ์๋ 57%์ ๋ถ๊ณผํ๋ฉฐ, ์ฝ 20๋ ํ์๋ 49%๊ฐ ๋ ๊ฒ์ผ๋ก ์์๋ฉ๋๋ค. ์ด๋ฏผ์ด ์ด๋ฌํ ๋ณํ์ ์ฃผ๋ ์์ธ ์ค ํ๋์ ๋๋ค.
ํ์ดํ ๊ต์๋ ์ด โํฐํ ํฌ์ธํธ(tipping point)โ ์ธ๋, ์ฆ ๋ฐฑ์ธ ๋ค์์์ ์์๋ก ์ ํ๋๋ 20๋ ๋์ ์ด๋ฏผ ๋ฌธ์ ๊ฐ ๋ฏธ๊ตญ์์ ๊ฐ์ฅ ํฐ โํผ๋ขฐ์นจโ์ด ๋๋ ๊ฒ์ ๋น์ฐํ๋ค๊ณ ๋งํฉ๋๋ค.
- ๊ธฐ์กด ์ง๋จ(๋ฐฑ์ธ ๋ค์): ์ด๋ฌํ ๋ณํ๋ฅผ ๋ฉ์ถ๊ฑฐ๋ ๋๋๋ฆฌ๊ธฐ๋ฅผ ์ํฉ๋๋ค. ๋ณํ๊ฐ ์งํ๋๋ฉด ์ ์น์ ๊ถ๋ ฅ์ ์๊ณ ๊ฒฝ์ ์ ์ ๋ง์ด ์ ํ๋ ๊ฒ์ด๊ธฐ ๋๋ฌธ์ ๋๋ค.
- ์๋ก์ด ์ง๋จ(์์ ์ง๋จ): ์ด๋ฌํ ๋ณํ๊ฐ ๊ณ์๋๊ธฐ๋ฅผ ๊ฐ์ ํ ์ํ๋ฉฐ, ์คํ๋ ค ๊ฐ์ํํ๊ธฐ๋ฅผ ๋ฐ๋๋๋ค. ๊ทธ๋ค์ ์ด ๋ณํ๋ก๋ถํฐ ํํ์ ๋ฐ์ ๊ฒ์ด๊ธฐ ๋๋ฌธ์ ๋๋ค.
์ด๋ฌํ ์ธ๊ตฌํต๊ณํ์ ๋ณํ๋ ์ ์น ํญ๋ ฅ ์ง์ง๋ฅผ ์ง์ ์ ์ผ๋ก ๋ถ์ถ๊ธฐ๋ ํต์ฌ ์์ ์ด ๋๊ณ ์์ผ๋ฉฐ, ์ด๋ ์ค๋ฌธ์กฐ์ฌ์์๋ ๋ช ํํ ๋๋ฌ๋ฉ๋๋ค.
2. ๋ถ์ ์ง์ค (Wealth Shift): ์์ 1%๋ก์ ๋ถ์ ์ด๋
์ธ๊ตฌํต๊ณํ์ ๋ณํ์ ๋๋ถ์ด, 1980๋ ๋ ์ค๋ฐ๋ถํฐ ์์๋ ์์ 1%๋ก์ ๋ถ์ ์ง์ค ํ์์ด ์์ต๋๋ค. ์ง๊ถ๋น์ด ์ด๋ ๋น์ด๋ ์๊ด์์ด, ํ์ 90%๋ก๋ถํฐ ์์ 1%๋ก ๋ถ๊ฐ ์ด๋ํ๊ณ ์์ต๋๋ค. ์ด๋ ํ์ 90%์ ์ํ๋ ๊ณตํ๋น ์ง์ง์๋ค๊ณผ ๋ฏผ์ฃผ๋น ์ง์ง์๋ค ๋ชจ๋๊ฐ ์์ ๋ค์ ๊ฒฝ์ ์ ์ ๋ง์ด ๋์ฑ ๋๋น ์ง ์ ์๋ค๋ ์ฐ๋ ค๋ฅผ ๊ฐ์ง๊ฒ ๋ง๋ญ๋๋ค.
์ด๋ฌํ ๋ถ์ ์ง์ค ํ์์ ์ด๋ ์ ๋น๋ ํด๊ฒฐํ์ง ๋ชปํ๊ณ ์๋ค๋ ์ ์ด ๋ฌธ์ ์ ๋๋ค. ์๋น ๋ชจ๋ ๋ณธ์ง์ ์ผ๋ก ๊ธฐ๋ถ์๋ค์๊ฒ ๋์ ์ฃผ๊ณ ์์ผ๋ฉฐ, ์ด ์ํฉ์ ๋ณํ์ง ์๊ณ ๊ณ์๋๊ณ ์์ต๋๋ค. ์ด๋ ์ ์น ํญ๋ ฅ์ ์ด๋นํ์ (bipartisan) ์ฆ๊ฐ์ ์ถ๊ฐ์ ์ธ ์ฐ๋ฃ๊ฐ ๋ฉ๋๋ค.
๊ฒฐ๋ก ์ ์ผ๋ก, ํ์ดํ ๊ต์๋ ์ด๋ฌํ ๋ ๊ฐ์ง ๊ฑฐ๋ํ ์ฌํ ๋ณํ๊ฐ ํน์ ์ฌ๋๋ค์ ๊ถ๋ ฅ๊ณผ ์ํฅ๋ ฅ์ ๋ณํ, ๊ทธ๋ฆฌ๊ณ ๋ฏธ๋์ ๋ํ ๋ถ์๊ฐ์ผ๋ก ์ด์ด์ง๋ฉฐ, **โ๋ง์ ์ฌ๋๋ค์ด ์์ ๋ค์ ๊ถ๋ ฅ์ ์๊ณ ์๋ค๊ณ ๋๋ผ๋ ๋๊ท๋ชจ์ ๋ณํโ**๊ฐ ์ ์น ํญ๋ ฅ์ ํต์ฌ ๋๋ ฅ์ด๋ผ๊ณ ์ง์ ํฉ๋๋ค.
โ์์ ๊ฒ ๋ง์โ ์ด๋ค์ ํญ๋ ฅ: ์ ์น์ ๋ฐฐ์ ์ ๋ํ ๋๋ ค์
์ฐ๋ฆฌ๋ ํํ ์ ์น ํญ๋ ฅ์ ์ ์ง๋ฅด๋ ์ฌ๋๋ค์ โ์์ ๊ฒ์ด ์๋โ ์ฌํ์ ์ฃผ๋ณ์ธ, ์ฆ ์ฌํ์ ์ฝ์๋ผ๊ณ ์๊ฐํ๋ ๊ฒฝํฅ์ด ์์ต๋๋ค. ๊ทธ๋ฌ๋ ํ์ดํ ๊ต์์ ์ฐ๊ตฌ๋ ์ด๋ฌํ ํต๋ ์ ๋ค์ง์ต๋๋ค. ๊ทธ์ ์ค๋ฌธ์กฐ์ฌ์ ๋ฐ๋ฅด๋ฉด, ๊ต์ก ์์ค์ด ๋๊ณ ์ค์ฐ์ธต ๋ฐ ์๋ฅ์ธต์ ์ํ๋ ์๋น์์ ์ฌ๋๋ค์ด ์ ์น ํญ๋ ฅ์ ์ง์งํ๊ณ ์์ต๋๋ค. ๊ทธ๋ค์ ํ์ฌ ๊ฐ์ง ๊ฒ์ด ๋ง์ผ๋ฉฐ, ๋ฏธ๋์ ๋ ๋ง์ ๊ฒ์ ์์๊น ๋ด ๋๋ ค์ํ๋ ์ฌ๋๋ค์ ๋๋ค.
๊ทธ๋ค์ด ํญ๋ ฅ์ ์์กดํ๋ ์ด์ ๋ ๋ฌด์์ผ๊น์? ํ์ดํ ๊ต์๋ ํต์ฌ์ ์ธ ๋๋ ค์์ผ๋ก **โ์ ์น์ ๋ฐฐ์ (Political Exclusion)โ**๋ฅผ ๊ผฝ์ต๋๋ค. ์์์๊ฐ ๋์ด ์ ์น ์์คํ ๋ด์์ ์์ ๋ค์ ๋ถ๋ง์ ํด๊ฒฐํ๊ฑฐ๋ ์ด์ต์ ๊ด์ฒ ํ ์ ์๊ฒ ๋ ๊ฒ์ด๋ผ๋ ๋๋ ค์์ ๋๋ค. ์ด๋ฌํ ๋๋ ค์์ ์์ชฝ ์ง์ ๋ชจ๋์์ ๋ฐ์ํฉ๋๋ค. ์๋ฅผ ๋ค์ด, ํธ๋ผํ ์ ๋ํต๋ น์ด ํฌํ ๊ธฐ๊ณ ์๋ฅ์ ๋ํด ์ด์ผ๊ธฐํ๋ ๊ฒ์ ๋ฏผ์ฃผ๋น ์ง์ง์๋ค์๊ฒ โ์ ์น์ ๋ด์(lockout)โ์ ๋ํ ๋๋ ค์์ ์ฌ์ด์ค๋๋ค. ์ ํ๊ธฐ์ธ 10~20๋ ๋์ ์์ชฝ ๋ชจ๋ ์ ์น์ ๋ด์๋ฅผ ๋๋ ค์ํ ์ ์์ผ๋ฉฐ, ์ด๋ ๋์ดํฌ ์ ์๊ฒ ๋๊ธฐ ์ ๋ง์ง๋ง ๊ธฐํ๋ผ๋ ์๊ฐ์ ๋์ฑ ๊ฒฉ๋ ฌํ๊ฒ ์ธ์ฐ๊ฒ ๋ง๋ญ๋๋ค.
๊ฒฐ๊ตญ, ์ด๋ ํ์ฌ์ ์ ์น ์์คํ ์ ๋ํ ๋นํ์ ๋๋ค. ์๋น ๋ชจ๋ ๋ง์ ์ฌ๋๋ค์๊ฒ ์ ์น์ ๋ณํ๊ฐ ์ค์ ๋ก ๊ฐ๋ฅํ๋ค๋ ๋๋์ ์ฃผ์ง ๋ชปํ๊ณ ์์ผ๋ฉฐ, ์ด๋ฌํ ๋๋ ค์์ ํน์ ์ฌ๋๋ค์๊ฒ ์ค์กด์ (existential)์ธ ๋ฌธ์ ๋ก ๋ค๊ฐ์ต๋๋ค. ๊ทธ๋ค์ ์์ ์ด ๋ฐฐ์ ๋ ๊ฒ์ด๋ฉฐ, ๊ณง ์์ ๋ค์ ๋ชฉ์๋ฆฌ๊ฐ ๋ค๋ฆด ๋ค๋ฅธ ๋ฐฉ๋ฒ์ด ์์ ๊ฒ์ด๋ผ๊ณ ๋๋๋๋ค. ์ด๋ ์ฌํ๋ณด์ฅ์ธ์จ๊ณผ ๊ฐ์ ์ ์ฑ ๋ ผ์์ ๋์ด์ , **โ์ง์ ํ ์ค์กด์ ์ ์น(existential politics)์ ๋ถ์โ**์ผ๋ก ์ด์ด์ง๊ณ ์์ต๋๋ค.
ํญ๋ ฅ์ผ๋ก ๊ฐ๋ ๊ธธ: ์ง๋์์ ์์ฌ(Rhetoric)์ ์ฌํ์ ์น์ธ
๊ทธ๋ ๋ค๋ฉด, ๋ฌด๋ ฅ๊ฐ์ ๋๋ผ๋ ์ฌ๋๋ค์ด ์ด๋ป๊ฒ ์ค์ ๋ก ํญ๋ ฅ์ ํ์ฌํ๋ ์ง์ ๊น์ง ๋์๊ฐ๊น์? ํ์ดํ ๊ต์๋ **โ์ง๋์๋ค์ ๋ฐ์ธโ**์ด ์ค์ํ๋ค๊ณ ์ง์ ํฉ๋๋ค. ๋๋๋ ํธ๋ผํ ์ ๋ํต๋ น์ด โ๊ทธ๋ฅผ ์ผ๊ตด์ ์ฃผ๋จน์ผ๋ก ๋๋ฆฌ๊ณ ์ถ๋คโ๊ณ ๋งํ๋ฉฐ ํญ๋ ฅ์ ์์ํ๋ ๋ฐ์ธ์ ํ ๊ฒ์ด ๋ํ์ ์ ๋๋ค. ๋ํ ๋ฒ ํ ์ค๋ฃจํฌ(Beto OโRourke)๋ ๊ฐ๋น ๋ด์ฌ(Gavin Newsom) ๊ฐ์ ๋ฏผ์ฃผ๋น ์ง๋์๋ค๋ โ๋จผ์ ์ฃผ๋จน์ ๋ ๋ ค์ผ ํ๋คโ๊ฑฐ๋ โ๊นกํจ์๊ฒ ๋ง์ ์ธ์ด๋คโ๋ ์์ ๊ฐ๊ฒฝํ ํํ์ ์ฌ์ฉํ๋ฉฐ ์ง์ง์๋ค์๊ฒ ์์ ๋ค์ด ์ธ์ธ ๊ฒ์์ ๋ณด์ฌ์ค๋๋ค.
์ด๋ฌํ ์ง๋์๋ค์ ๊ถ์ ์๋ ์ธ๋ฌผ์ด๋ฉฐ, ๊ทธ๋ค์ด ํญ๋ ฅ์ ์ฉ์ธํ ๋งํ๋ค๊ณ ๋งํ๋ฉด(ํน์ ์๋์น ์๊ฒ ๊ทธ๋ ๊ฒ ๋ค๋ฆฌ๊ฒ ํ๋ฉด), ๋ค๋ฅธ ์ฌ๋๋ค๋ ๊ทธ ๋ฐฉํฅ์ผ๋ก ์์ง์ผ ์ ์์ต๋๋ค. ํ์ดํ ๊ต์๋ ์ด๋ฅผ โ์์์ ์๋๋ก(top-down)โ ์ ๋ฌ๋๋ ์์ฌ๋ผ๊ณ ํํํฉ๋๋ค. ํน์ ๊ฐ์ธ๋ค์ ์ด๋ฏธ ํ๋ฅผ ์ ๋ด๊ณ ์ถฉ๋์ ์ผ๋ก ํ๋ํ๋ ๊ฒฝํฅ์ด ์๋๋ฐ, ์ด๋ฌํ ์ง๋์๋ค์ ๋ฐ์ธ์ ๊ทธ๋ค์ ํ๋์ผ๋ก ์ฎ๊ธฐ๊ฒ ํ๋ ์ด๋งค์ ๊ฐ ๋ฉ๋๋ค.
๋ํ โ์๋์์ ์๋ก(bottom-up)โ ์ ๋ฌ๋๋ ์์ฌ๋ ์ค์ํ ์ญํ ์ ํฉ๋๋ค. ์์ ๋ฏธ๋์ด๋ ์ด๋ฌํ ์ํฉ์ โํ๋ฐ์ ๋ฅผ ๋ถ๋โ ์ญํ ์ ํฉ๋๋ค. ์๋ฅผ ๋ค์ด, ์ ๋์ดํฐ๋ ํฌ์ค์ผ์ด(United Healthcare) CEO๊ฐ ์ดํด๋ ํ, ์ฌ๋๋ค์ด X(๊ตฌ ํธ์ํฐ)์์ ๋ฒ์ธ์ ๊ณต๊ฐ์ ์ผ๋ก ์ฐฌ์ํ๊ณ ์์ ์ํ๋ ํ์์ด ๋ํ๋ฌ์ต๋๋ค. ์ด๋ ์ ์น ํญ๋ ฅ์ ์ง์งํ๋ ์์ฒ๋ง ๋ช ์ ๋ฏธ๊ตญ์ธ์ด ์์ ๋ฏธ๋์ด์ ์กด์ฌํ๋ฉฐ, ๊ทธ๋ค์ ๋ฐ์์ ์ ์ฌ์ ๊ณต๊ฒฉ์๋ค์๊ฒ โ์ฌํ์ ์น์ธโ์ ๋ํ ํ์ ์ ์ค๋๋ค. ๋๋๋ ํธ๋ผํ ์ฃผ๋์ด๊ฐ ํด ํ ๋ก์ ํผ์ต ์ฌ๊ฑด์ ํผ๋ฌ์ ์์์ผ๋ก ์กฐ๋กฑํ๋ ๊ฒ์๋ฌผ์ ์ฌ๋ฆฐ ํ ์์ฒญ๋ ์ง์ง๋ฅผ ๋ฐ์ ์ฌ๋ก์ฒ๋ผ, ์ํฅ์๊ณผ ํํฅ์ ์์ฌ๊ฐ ์ํธ์์ฉํ๋ฉฐ ํญ๋ ฅ์ ๋ถ์ถ๊น๋๋ค.
๊ทธ๋ฌ๋ ํ์ดํ ๊ต์๋ ์์ ๋ฏธ๋์ด๊ฐ ์ํฉ์ ์ ํ์ํค์ง๋ง, ๊ทธ ์์ฒด๊ฐ โ๋ถ(fire)โ์ ์๋๋ผ๊ณ ๊ฐ์กฐํฉ๋๋ค. 1920๋ ๋์ 1960๋ ๋์๋ ์์ ๋ฏธ๋์ด๊ฐ ์์์์๋ ๋ถ๊ตฌํ๊ณ ์์ฒญ๋ ์ ์น ํญ๋ ฅ์ด ์กด์ฌํ๊ธฐ ๋๋ฌธ์ ๋๋ค. ๊ทธ๋ ์ฐ๋ฆฌ๊ฐ ์ธํฐ๋ท๊ณผ ๊ฐ์ ๋ถ์ฐจ์ ์ธ ๋ฌธ์ ์๋ง ์ง์คํ์ฌ ๊ทผ๋ณธ์ ์ธ ์ฌํ ๋ณํ๋ผ๋ ํต์ฌ ๋ฌธ์ ๋ฅผ ์ธ๋ฉดํ๋ ๊ฒฝํฅ์ด ์๋ค๊ณ ์ง์ ํฉ๋๋ค.
๋ฌ๋ผ์ง ๊ณต๊ฒฉ์์ ํ๋กํ: ์ฐ๋ฆฌ ์์ง์ โ๊ดด๋ฌผโ
ํ์ดํ ๊ต์๋ ์ ์น ํญ๋ ฅ์ ์ ์ง๋ฅด๋ ์ฌ๋๋ค์ ํ๋กํ์ด ํฌ๊ฒ ๋ฐ๋์๋ค๋ ์ ์ ๊ฐ์กฐํฉ๋๋ค. ๊ทธ์ ์ฐ๊ตฌํ์ 1์ 6์ผ ๊ตญํ์์ฌ๋น ๋์ ์ฌ๊ฑด์ ๊ฐ๋ดํ ์ฌ๋๋ค์ ์ด๊ธฐ ํ๋กํ์ ์ง๋ ์์ญ ๋ ๊ฐ FBI ๋ฐ์ดํฐ์ ๋ํ๋ ์ฐ์ต ์ ์น ํญ๋ ฅ ๋ฒ์ฃ์๋ค์ ํ๋กํ๊ณผ ๋น๊ตํ์ต๋๋ค.
- ๊ณผ๊ฑฐ์ ํ๋กํ: ์ค์ ์, ๋ฏผ๋ณ๋์ ๋ฑ ์ฌํ์ ์ฃผ๋ณ๋ถ(fringe)์ ์ํ๋ ์ฌ๋๋ค์ด์์ต๋๋ค.
- 1์ 6์ผ ์ฌ๊ฑด ๊ฐ๋ด์ ํ๋กํ: ํ์ ํ ๋ฌ๋์ต๋๋ค. ์ค์ ๋ฅ ์ ์ ๊ตญ ํ๊ท ์ธ 7%์ ๋ถ๊ณผํ์ผ๋ฉฐ, ์ฌ์ ๊ฐ, ์์ฌ, ๋ณํธ์ฌ, CEO ๋ฑ ์ฃผ๋ฅ ์ฌํ ๊ตฌ์ฑ์์ด ๋ง์์ต๋๋ค. ๋ฏผ๋ณ๋ ์์์ 10%์ ๋ถ๊ณผํ์ต๋๋ค(๊ณผ๊ฑฐ์๋ 50%). ํ์ดํ ๊ต์๋ ์ด๋ค์ โ๋ฉ์ง ๊ต์ธ์ ์ฌ๋ ๋น์ฆ๋์ค ์ ์ฅ ์ฐจ๋ฆผ์ ๋ฐ๋์โ๋ผ๊ณ ๋ฌ์ฌํฉ๋๋ค.
์ด๋ฌํ ๋ณํ๋ ์ฐ๋ฆฌ๊ฐ ํญ๋ ฅ์ ์ธ ์ฌ๋๋ค์ ์๋ชป๋ ๊ณณ์์ ์ฐพ๊ณ ์์ ์ ์์์ ์๋ฏธํฉ๋๋ค. ํ์ดํ ๊ต์๋ ์ฐ๋ฆฌ๊ฐ โ๊ดด๋ฌผโ์ ์ฐ๋ฆฌ์๊ฒ์ ๋ฉ๋ฆฌ ๋จ์ด์ง ์ ๋น์ผ๋ก ์ฌ๊ธฐ๊ธฐ๋ฅผ ์ข์ํ๋ค๊ณ ์ค๋ช ํฉ๋๋ค. ์ ์น์ ์์ด๊ณผ ์ด์ธ์ ์ ์ง๋ฅผ ์ ์๋ โ๊ดด๋ฌผโ์ด ์ฐ๋ฆฌ์ ๋ฉ๋ฆฌ ๋จ์ด์ ธ ์๊ณ , ์ค๋ ์ผ์ด๋ ์ผ์ด ๋ค์๋ ์ผ์ด๋์ง ์์ ๊ฒ์ด๋ผ๊ณ ์๊ฐํ๋ ๊ฒ์ด ๋ง์ ํธํ๊ธฐ ๋๋ฌธ์ ๋๋ค. ์ด๋ฌํ โ์๋ชป๋ ํ๋กํ์ ๋ํ ๊ฑฐ์ง ์ด์ผ๊ธฐโ๋ ๋์ค์๊ฒ ์์ฌ์ ์ฃผ๊ธฐ ์ํด ์๊ฒจ๋๊ณ ์ธ๊ธฐ๋ฅผ ์ป์ง๋ง, ํ์ค๊ณผ๋ ๊ฑฐ๋ฆฌ๊ฐ ๋ฉ๋ค๋ ๊ฒ์ ๋๋ค.
ํด๊ฒฐ์ฑ : ๋ ๋จ๊ณ ์ ๊ทผ๊ณผ 75%์ ์ญํ
ํ์ดํ ๊ต์๋ ์ด ์๊ธฐ๋ฅผ ํด๊ฒฐํ๊ธฐ ์ํด ๋ ๊ฐ์ง ๋จ๊ณ์ ์ ๊ทผ ๋ฐฉ์์ด ํ์ํ๋ค๊ณ ์ ์ํฉ๋๋ค.
1. ์ฅ๊ธฐ์ ํด๊ฒฐ์ฑ (์ค๊ฐ์ ๊ฑฐ ์ดํ)
- ์ธ๊ตฌํต๊ณํ์ ๋ณํ ์๋ ์กฐ์ : ๊ตญ๊ฐ์ ์ธ๊ตฌํต๊ณํ์ ๋ณํ๋ฅผ ๋ง์ ์๋ ์์ง๋ง, ๊ทธ ์๋๋ฅผ ๋ฆ์ถ์ด ์ ์น ์ ๋๊ฐ ๋ฐ๋ผ์ก์ ์ ์๋๋ก ํ๋ ๊ฒ์ด ํ๋ช ํฉ๋๋ค. ์ด๋ ๋ณธ์ง์ ์ผ๋ก ์ค๋ฐ๋ง ์๋์ ์ด๋ฏผ ์ ์ฑ ์ ์ฑํํ๋ ๊ฒ์ ์๋ฏธํฉ๋๋ค.
- ๋ถ์ ์ง์ค ์ํ: ์์ 1%๋ก์ ๋ถ์ ์ด๋์ ๋ฉ์ถ๊ฑฐ๋ ์ญ์ ์ํค๋ ์ ์ฑ ์ด ํ์ํฉ๋๋ค. ํ์ฌ ์๋น ๋ชจ๋ ์ด ๋ฌธ์ ๋ฅผ ํด๊ฒฐํ์ง ๋ชปํ๊ณ ์์ผ๋ฉฐ, ๊ธฐ๋ถ์๋ค์๊ฒ ๋์ ์ฃผ๊ณ ์๋ ์ํฉ์ ๋๋ค. ์ฐ๋ฆฌ๋ ์ด์ ๋ํด ์์งํด์ผ ํฉ๋๋ค.
2. ๋จ๊ธฐ์ ํด๊ฒฐ์ฑ (์ค๊ฐ์ ๊ฑฐ๊น์ง)
- ์ง๋์๋ค์ ์ญํ : ๋น์ฅ ์ค๊ฐ์ ๊ฑฐ๊น์ง๋ ์์ ์ฅ๊ธฐ์ ์ ์ฑ ๋ค์ ๋ฐ๊ฟ ์ ์์ต๋๋ค. ๋ฐ๋ผ์ ์์ผ๋ก ๋ช ๋ฌ ๋์ ์ฐ๋ฆฌ๊ฐ ํ ์ ์๋ ์ ์ผํ ์ ์ฑ ์ ์ง๋์๋ค์ ๋ณํ์ ๋๋ค. ์ง๋์๋ค์ ํ์ฌ ๋ฐ์ํ๊ณ ์๋ ํญ๋ ฅ์ ์์ฉ๋์ด๋ฅผ ์ค์ผ ์ ์๋ ์ ์ผํ ์ค์ ์์ฒ์ ๋๋ค.
- ํ์ดํ ๊ต์๋ ํํด ์ ํ๋ฆฌ์ค(Hakeem Jeffries) ํ์ ๋ฏผ์ฃผ๋น ์๋ด๋ํ๊ฐ ๋๋๋ ํธ๋ผํ ์ ๋ํต๋ น์๊ฒ ์ฐ๋ฝํ์ฌ โ์์ชฝ ์ง๋์๋ค์ด ๋ชจ๋ ํํ์ ์ ์น ํญ๋ ฅ์ ๊ทํํ๋ ๊ณต๋ ์์ ์ฑ๋ช ์ ๋ฐํํ ์ ์์์งโ ์ ์ํ๋ ๊ณผ๊ฐํ ํ๋์ ํด์ผ ํ๋ค๊ณ ๋งํฉ๋๋ค.
์ด๋ฌํ ์ง๋์๋ค์ ์์ฌ๊ฐ ํจ๊ณผ๊ฐ ์์๊น์? ํ์ดํ ๊ต์๋ 2024๋ ํธ๋ผํ ์ ๋ํต๋ น์ ๋ํ ์ฒซ ๋ฒ์งธ ์์ด ์๋ ํ ์กฐ ๋ฐ์ด๋ ๋ํต๋ น์ด ๋ฐฑ์ ๊ด ์ง๋ฌด์ค์์ ํญ๋ ฅ์ ๊ทํํ๋ ์ฐ์ค์ ํ์ ๋, ๋ช ๋ฌ ํ ์ ์น ํญ๋ ฅ ์ง์ง์จ์ด ์ ๊ตญ์ ์ผ๋ก 20% ๊ฐ์ํ๋ ์ฌ๋ก๋ฅผ ์ธ๊ธํฉ๋๋ค. ๋ ๋ฒ์งธ ์์ด ์๋ ํ ๋ฐ์ด๋ ๋ํต๋ น๊ณผ ์นด๋ง๋ผ ํด๋ฆฌ์ค ๋ถํต๋ น ํ๋ณด๊ฐ ๋ค์ ๊ฐ์ ์์ฌ๋ฅผ ์ฌ์ฉํ์ ๋๋ ์ง์ง์จ์ด ์ถ๊ฐ๋ก ํ๋ฝํ์ต๋๋ค. ์ด๋ ์ง๋์๋ค์ ๋ฉ์์ง๊ฐ ์ต์ํ ์ํฉ์ ์ ํ์ํค์ง ์๊ณ ๊ธ์ ์ ์ธ ์ํฅ์ ๋ฏธ ๋ฏธ์ณค์์ ์์ฌํฉ๋๋ค.
๊ตญ๋ฏผ์ ์ ํ: 75%์ ์นจ๋ฌต์ ๊นจ์ผ ํ ๋
ํ์ดํ ๊ต์๋ ์ด ๋ชจ๋ ๊ตฌ์กฐ์ ์์ธ๊ณผ ์๋ ฅ์๋ ๋ถ๊ตฌํ๊ณ **โ์ ์น ํญ๋ ฅ์ ๋ฏธ๋๋ ์ ํ์ ๋ฌธ์ โ**๋ผ๊ณ ๊ฐ์กฐํฉ๋๋ค. ์ฐ๋ฆฌ๋ ์ด๋ฌํ ๊ตฌ์กฐ์ ์๋ ฅ์ ๋ฌด์ํ ์๋ ์๊ณ , ์ดํดํ๊ณ ์์ผ๋ก ๋์๊ฐ ์๋ ์์ต๋๋ค.
๊ทธ๋ ์ ์น ํญ๋ ฅ์ ์ง์งํ๋ ์์ฒ๋ง ๋ช ์ ์ฌ๋๋ค์๊ฒ ์ด์ ์ ๋ง์ถฐ์์ง๋ง, ๋ ์ค์ํ ๊ฒ์ **โ๋ฏธ๊ตญ ์ธ๊ตฌ์ 75%๊ฐ ์ ์น ํญ๋ ฅ์ ํ์คํ๋คโ**๋ ์ฌ์ค์ ๋๋ค. ์ด 75%๋ ํ๋์ ๋์์ผ ํฉ๋๋ค. ๊ทธ๋ค์ ์ ์น ํญ๋ ฅ์ ๋ํด ๋ถํํ๊ณ , ์์์์, ํ์์์, ์์ฅ, ๋ํต๋ น์๊ฒ ํธ์ง๋ฅผ ๋ณด๋ด์ผ ํฉ๋๋ค. โ๋ฏผ์ฃผ๋น๊ณผ ๊ณตํ๋น ์ง๋์๋ค์ด ๊ณต๋์ผ๋ก ์ ์น ํญ๋ ฅ์ ๊ทํํด์ผ ํ ๋์ ๋๋คโ๋ผ๊ณ ๋ชฉ์๋ฆฌ๋ฅผ ๋ด์ผ ํฉ๋๋ค.
โ์ฐ๋ฆฌ๋ ๋ ๋์ ๋ฏธ๋๋ฅผ ๊ฐ์ง ์ ์์ต๋๋ค. ๊ทธ๊ฒ์ ์ ํ์ ๋๋ค. ๊ทธ๋ฌ๋ ๊ทธ๊ฒ์ ์ฐ๋ฆฌ ๊ตญ๋ฏผ์ด, ๊ทธ๋ฆฌ๊ณ 75%์