YouTube Digest

June 18, 2026

Listen to this issue

Part 1
Part 2
Part 3

English

Based on “How GitHub Deals with 17 Million Pull Requests a Month” from Every Watch the original video

The AI Deluge: GitHub’s Battle on the Front Lines of Code’s Exponential Future

Episode summary

GitHub’s COO, Kyle Dagel, reveals how the platform is grappling with an explosion of activity, now processing 17 million agent-created pull requests monthly—a staggering increase from the mere 1 billion commits in all of last year. This surge is driven by a broadened definition of “developer,” with non-traditional users like legal and finance teams leveraging tools like GitHub Copilot to build apps and assets, fundamentally shifting product roadmap priorities towards accessibility and ease of use.

To manage this influx and prevent agent subscription costs from spiraling, GitHub is focusing on “hill climbing”—a continuous improvement loop using both hard metrics and user sentiment to refine models and workflows. The practical takeaway for users and enterprises is the development of “model routers” that intelligently select the most cost-effective AI models based on task complexity, ensuring efficient resource allocation rather than relying on expensive, general-purpose models for every minor task.

The world of software development is undergoing a seismic shift, driven by the relentless rise of artificial intelligence. At the epicenter of this transformation stands GitHub, the ubiquitous platform where the world builds software. With an unprecedented 17 million pull requests (PRs) pouring in every month, a significant portion of which are generated by AI agents, GitHub is grappling with a deluge of code that is redefining how we build, collaborate, and even conceive of “developers.”

Mike Taylor, head of tech consulting at Every, recently sat down with Kyle Dagel, COO of GitHub and Chief Marketing Officer of Developer for Microsoft, to peel back the layers of this new “agent economy.” Their conversation illuminated GitHub’s strategies for navigating this exponential growth, supporting its diverse user base, and envisioning a future where AI isn’t just a tool, but a fundamental partner in creation.

The New Face of the Developer

One of the most striking changes GitHub observes is the evolving demographic of its users. Historically a domain for professional software engineers, GitHub is now seeing a significant influx of “knowledge workers” – individuals from legal, finance, and other non-traditional tech backgrounds – leveraging AI tools to build applications and assets.

“For GitHub, in particular, we’ve always really had this really expansive view of what a developer is,” explains Dagel. He recounts his own journey, writing code to pay for art school, long before he would have called himself a “dev.” This personal history informs GitHub’s philosophy: to make it easier for anyone to choose to write some code, ensuring there’s always an on-ramp into software creation. Tools like GitHub Copilot are central to this democratization, enabling teams across an enterprise to build small apps and automate tasks, even without a formal computer science background.

While GitHub continues to serve the largest businesses and “serious developers,” it recognizes that the future of coding is inclusive. The product roadmap is now shaped by the needs of this broader audience, striving to build tools that empower both seasoned engineers and accidental coders.

Taming the Deluge: Supporting Open Source Maintainers

The sheer volume of AI-generated code presents a significant challenge, particularly for open-source maintainers who are, in Dagel’s words, “drowning” under a flood of pull requests. The question isn’t just about quantity, but also about quality and the burden of review.

GitHub is addressing this with agentic tools designed to streamline the code review and merge process. Features like Copilot Code Review, which can identify novel vulnerabilities and implement changes based on comments, significantly reduce manual effort. The “agentic merge” capability allows maintainers to set policies and let the AI handle the final processing steps, waiting for CI checks and other requirements to pass automatically.

However, GitHub maintains a delicate balance when it comes to open source. “Every community is choosing a slightly different way to approach the problem,” Dagel notes. Rather than imposing a universal standard, GitHub focuses on providing maintainers with granular control. This means offering tools to decide who can submit PRs, how much work is required to prove a contribution’s value, and the level of vetting needed. Dagel cites Mitchell Hashimoto’s “vouch system” as an example of a community-led solution that GitHub observes but doesn’t immediately roll out to everyone. The goal is to provide building blocks and enable communities to self-organize, cementing a system only if a clear, widely adopted practice emerges.

The Agent Economy’s Boom

The statistics are staggering. Last year, GitHub recorded a billion commits for the full year. This year, if growth were linear, it would be 14 billion. But the growth isn’t linear; it’s exponential, largely thanks to AI agents. In March alone, 17 million pull requests were created by agents.

This explosion of activity signals a fundamental shift: “It’s not just Kyle building, but it’s Kyle in one to N, you know, agents,” Dagel illustrates. While some might dismiss this as “slop” or low-quality code, Dagel argues that it represents a significant leap in productivity. Developers are now collaborating with AI partners, using their skills, resources, and context to generate code at an unprecedented pace. GitHub is investing heavily to prepare for the “next wave of growth,” recognizing that “no matter where you’re building or what tools you’re using to build, all of that code ends up on GitHub.”

Evolving Business Models: Freemium to Usage-Based?

The rise of 24/7 AI agents challenges traditional freemium business models. If agents are constantly working while humans sleep, how does pricing adapt? Dagel acknowledges that the ultimate model is still evolving. GitHub has historically adapted its offerings, moving from paid private repos to free ones for individuals. The current focus is on ensuring a “great core GitHub experience” for developers, which includes a baseline of agent usage.

However, for those who want to run “150 agents doing everything all at once,” as one developer famously put it, a usage-based model seems inevitable. The challenge lies in balancing the need to enable scaled agent activity with providing a predictable, valuable experience for individual developers and enterprises. This will involve careful consideration of API rate limits and how to meter the immense computational effort behind agent-driven development.

A Dual Mandate: Developer-First at Scale

Kyle Dagel’s unique dual role as COO of GitHub and Chief Marketing Officer of Developer for Microsoft offers a fascinating glimpse into the strategic alignment of two tech giants. Having been at GitHub for 13 years, Dagel emphasizes the company’s unwavering focus on the developer. “We’re not building for the buyers, we’re building for the developers 100%,” he asserts.

This developer-centric philosophy is now being extended across Microsoft’s broader developer tooling. Dagel’s aim is to ensure “holistic solutions that you can use that are authentic to developer experiences.” This influence was evident at the recent Microsoft Build conference, which took a dramatically different approach: relocating to San Francisco, focusing on hands-on sessions over sales pitches, and, crucially, inviting external community speakers into primary sessions. This shift reflects the belief that “software development is a team sport” and that no single company holds all the answers.

The Power of Choice in a Competitive Landscape

In a rapidly evolving and fiercely competitive market, GitHub’s differentiation lies in its steadfast commitment to “developer choice.” Dagel explains that while the industry sometimes trends towards “unintentional walled garden setups,” GitHub actively resists this. Instead, it aims to enable developers to use GitHub alongside other tools and partners with “everyone who’s bringing a model to market or a coding agent to market.”

This means not only investing in its own AI models (like the new Microsoft AI models) but also integrating and emitting through GitHub Copilot the capabilities of Anthropic, OpenAI, Google, and other providers. “That choice is core, and that’s something that I don’t think we’ll ever back down on,” Dagel declares. “Because if we do, developers will still choose. They’ll just be stuck in another kind of mouse trap… and we don’t want the world of software to be like that.”

Internal Innovation and “Hill Climbing”

GitHub’s internal decision-making process embraces a culture of “experimentation.” Dagel himself “dogfoods” across multiple operating systems (Mac, Windows, Linux) and uses various tools, even those not made by Microsoft, to understand the developer experience. This prevents “blind spots” that can arise from an overly narrow focus.

A core methodology driving GitHub’s progress is “hill climbing.” This iterative, data-driven approach involves continuously improving models and tools based on real-world usage data – from “thumbs up/thumbs down” feedback to acceptance rates of AI suggestions. “Every week we’re talking about the hill climbing results,” Dagel states. The emphasis is on small, continuous improvements rather than chasing “moonshots.” This involves looking at both “hard measures” (evals, rubrics) and “soft measures” (user sentiment), recognizing that sometimes an improvement in metrics can lead to a crash in user satisfaction if not carefully managed. The ultimate goal is to empower every developer with their own “hill climbing machine,” making continuous improvement accessible and automated.

The $200 to $2,000 Problem: Solving for Token Economics

The concern of a $200/month coding agent subscription ballooning to $2,000/month is a very real one in the agent economy. Dagel believes the answer lies in intelligent model routing and personalization. The solution isn’t just about better, cheaper models, but about automatically choosing the right model for the right task.

This involves “model routers” that can assess task intent and select the most efficient model – perhaps a smaller, cheaper model for a simple find-and-replace, and a more powerful, expensive one for a complex architectural problem. “The more and more that we can help you tell us a bit of like where your bars are, like this is an incredibly hard problem and I’m willing to go all the way to the top, or I don’t, I just kind of want to sit here, and let us help choose the models,” Dagel explains. This intelligent orchestration, combined with “Frontier Tuning” (personalizing models with individual context and memory), is crucial to optimizing token usage and preventing runaway costs, especially for enterprises.

The Personal AI Loop: A Surprising Revelation

In a lighter, yet profoundly insightful moment, Mike Taylor revealed he had created an AI clone of Kyle Dagel to practice the interview. Dagel’s response was even more surprising: he does something similar himself. Dagel uses a personal AI (affectionately named “Baxter”) to read everything he writes – emails, Slack messages, scripts – and provide daily “comms reports.”

“Kyle, you keep saying this. This isn’t super clear,” his AI might say, or offer clearer metaphors based on his speaking style. Dagel finds this “self-improvement loop as a human from these agents to be incredibly powerful.” He harks back to the early days of chat ops with Hubot, noting, “Humans are way more willing to take critical feedback from robots than other humans.” This less threatening feedback loop helps him recursively self-improve his communication. This personal application of AI highlights a future where agents not only augment our coding abilities but also our human skills, offering a unique form of introspective development.

Conclusion

GitHub stands at a critical juncture, navigating the unprecedented growth of AI-generated code and the evolving landscape of software development. Under Kyle Dagel’s leadership, the platform is committed to an inclusive, developer-first approach, offering choice, fostering community-led innovation, and embracing iterative improvement. From taming the deluge of pull requests to reimagining business models and even personal development, GitHub is not just observing the AI revolution; it’s actively shaping its future, ensuring that the next wave of code is built collaboratively, efficiently, and with the developer at its heart.

Screenshot of GitHub Copilot app showing code suggestions, illustrating AI's role in democratizing coding.

Screenshot of GitHub Copilot app showing code suggestions, illustrating AI’s role in democratizing coding.

Screenshot of a pull request with Copilot Code Review suggestions, demonstrating AI's role in streamlining code review.

Screenshot of a pull request with Copilot Code Review suggestions, demonstrating AI’s role in streamlining code review.

A chart or graphic illustrating the exponential growth of pull requests, highlighting the 'agent economy' boom.

A chart or graphic illustrating the exponential growth of pull requests, highlighting the ‘agent economy’ boom.

A photo of the Microsoft Build conference in San Francisco, showing a more hands-on, community-focused atmosphere.

A photo of the Microsoft Build conference in San Francisco, showing a more hands-on, community-focused atmosphere.

A diagram showing GitHub Copilot integrating with multiple AI models (Anthropic, OpenAI, Google), emphasizing developer choice.

A diagram showing GitHub Copilot integrating with multiple AI models (Anthropic, OpenAI, Google), emphasizing developer choice.

A screenshot of an AI 'comms report' with feedback on communication style, illustrating personal AI for self-improvement.

A screenshot of an AI ‘comms report’ with feedback on communication style, illustrating personal AI for self-improvement.


Based on “Generative AI’s race problem | It’s Been a Minute” from NPR Podcasts Watch the original video

The Unseen Battle: How Generative AI Exploits Race, Gender, and Our Deepest Biases

Episode summary

NPR’s “It’s Been a Minute” explores the “race problem” in generative AI with podcast host Bridget Todd, who argues that AI-generated content, particularly videos featuring Black women, often exploits racial stereotypes for profit and to confirm existing biases. A prime example is the viral, AI-generated song “Phenomenal Black Woman,” which, despite its manufactured origins, resonated deeply with many Black women due to a lack of genuine media representation, highlighting how AI is designed to elicit strong emotional responses. This extends to “scammy” videos of Black children making handmade items, which are actually drop-shipped goods, exploiting community solidarity for financial gain.

Viewers should be wary of content designed to provoke an instant, strong emotional reaction, as this is a key indicator of potentially AI-generated or manipulative material. The hosts also discuss how AI is used to spread misinformation and political discord by validating pre-existing fears, especially among those unfamiliar with a given context. While some platforms encourage AI disclosure, these are not requirements, leaving users susceptible to content that profits from engagement, even if it’s based on harmful caricatures or outright scams.

In the ever-evolving landscape of digital media, a catchy tune titled “Phenomenal Black Woman” recently captivated TikTok feeds, sparking joy and pride among many Black women and their families. The song, by an artist named China Styles, from an album called I Love Me, seemed to embody a powerful message of self-affirmation. Yet, beneath its seemingly positive surface lay a disquieting truth: the artist, the album, and the song itself are entirely products of artificial intelligence.

This anecdote, shared by NPR’s Britney L. Loose on “It’s Been a Minute,” serves as a stark introduction to a broader, more insidious problem that AI expert and “There Are No Girls on the Internet” podcast host Bridget Todd has been tirelessly ringing the alarm about. Generative AI, while promising innovation, is increasingly being leveraged to create content that traffics in harmful racial and gender stereotypes, often for financial gain or to sow discord. “Our platforms, a lot of them are designed such that unfortunately there is financial incentives to making this kind of content,” Todd explains.

The Phenomenon of “Phenomenal Black Woman”

The “Phenomenal Black Woman” song, despite its AI origins, quickly became a viral sensation. Loose recounts how she heard it countless times on her TikTok feed, and even Black “aunties” in her life embraced it. But as soon as Todd heard it, her “society senses” went up. “It sounds so fake. It sounds so cheesy. It sounds so corny,” she remarked, questioning if it was truly a Black woman singing or “Glen Kosa’s character from Deliverance.”

This “on the nose” quality, Todd explains, is a hallmark of racialized AI content. It’s so overtly crafted to appeal to a specific identity or emotion that it struggles to suspend disbelief. While the song moved many Black women, offering them a rare feeling of being “meaningfully seen” in a media landscape that often overlooks them, Todd points out its manufactured nature. “It’s literally been designed, manufactured… to elicit some kind of strong response,” she says, describing it as having “all of the caricature turned up to 11.” This exemplifies how AI can exploit genuine human needs and desires for representation, twisting them into commodified, inauthentic experiences.

When Empathy Becomes a Weapon: Scams and Stereotypes

The “Phenomenal Black Woman” song, while ethically dubious, pales in comparison to some of the more overtly harmful AI-generated content Todd has encountered. These videos range from low-level scams to deeply nefarious forms of exploitation.

The “Handmade Purses” Deception

One prevalent scam involves AI-generated videos purporting to show Black children lovingly making handmade items, often purses. The accompanying narrative typically features a tragic backstory: the child took their creations to a craft fair, only for “racist white kids” to push them to the ground. The video then shows the child diligently sewing and asks viewers to “support this young black visionary” by clicking a link to buy their products.

A screenshot of an AI-generated video showing a Black child sewing a purse, illustrating the 'handmade purses' scam.

A screenshot of an AI-generated video showing a Black child sewing a purse, illustrating the ‘handmade purses’ scam.

This tactic is particularly insidious because it preys on community solidarity. “Whoever is making this content, they know that black folks, we ride for each other. We go up for each other,” Todd asserts. Celebrities like Viola Davis and Robin Dixon were among those initially taken in by these emotionally manipulative videos. The “handmade” items, of course, turn out to be mass-produced, drop-shipped junk from unethical sources, highlighting how AI is used to craft compelling, yet entirely false, narratives for financial gain.

Political Caricatures and Misinformation

Beyond scams, AI-generated content is a potent tool for political manipulation and the spread of misinformation. Loose recalls seeing AI-generated “political oppo videos” during a New York City mayoral race, designed to quickly elicit negative emotions about candidates by linking them to terrorism, violence, or crime.

Todd highlights even more disturbing examples, such as AI-generated videos that surfaced before SNAP funding pauses were implemented. These videos depicted “racially stereotypical black women” – often “heavy set” and behaving in exaggerated ways – having “big reactions” in grocery stores. This content, often bearing AI branding like “Sora or V3,” went viral despite its obvious artificiality.

“You don’t even need… the welfare queen Linda Taylor anymore,” Todd laments, referring to a historical figure used to perpetuate racist stereotypes about welfare recipients. “You don’t need an actual real life sort of scammer and schemer when you can just sort of generate these AI caricatures.” These videos effectively serve as “racist fanfiction,” designed to validate pre-existing biases and fears about public services and their recipients, influencing public opinion and potentially even news outlets.

Digital Blackface and Deeper Exploitation

The use of AI to create racially charged caricatures has been aptly termed “digital blackface” by some. Todd notes that these videos heavily traffic in racial and ethnic stereotypes, designed to provoke strong emotional reactions. For instance, AI-generated Black women with “impossibly dark skin” are sometimes depicted saying “racially degrading or racially inflammatory things,” with links in their bios leading to paid adult content sites. The audience for such content is clearly not Black people, but rather those who consume and are fueled by such derogatory depictions.

The “why” behind this exploitation ranges from “engagement farming” – keeping users on a page longer for potential platform payouts – to direct dropshipping scams and even explicit content. AI allows creators to “AB test” different stereotypes at scale, generating endless “slop” until they find the exact right content to resonate with a target demographic, often one predisposed to existing biases.

The Profit Motive: Why AI Exploitation Thrives

The explosion of such content is largely driven by its low cost and ease of production. Platforms like Google V3 (and formerly OpenAI’s Sora) allow for the rapid creation of these videos. This accessibility, coupled with platforms’ design that financially incentivizes viral content, creates a fertile ground for exploitation.

“If you can get a few bucks for making a viral AI generated video that pops off, whether it traffics in like racially inflammatory content or not, our platforms… are designed such that there is a financial motivation to do exactly that,” Todd explains. This system allows for the “commodification” of Blackness and other identities without the need to engage with or compensate actual people. Creators can bypass the objections of real individuals who might find such depictions racist or demeaning, producing content at scale that appeals to a market for “black caricature in conflict.”

The “Girlbossification” of AI: A Gendered Divide

The ethical minefield of AI extends beyond race to gender, sparking another contentious conversation. Reese Witherspoon recently drew criticism for advocating that more women get involved with AI, citing that women’s jobs are more likely to be automated. This push for women to adopt AI has been dubbed the “girlbossification of AI.”

Statistics reveal a significant gender gap: a CNBC survey found that 69% of men view AI as a “valuable assistant,” compared to just 61% of women. Half of women expressed suspicion, viewing AI use at work as “cheating,” a sentiment shared by only 43% of men.

Todd delves into the nuanced reasons behind women’s skepticism, challenging the simplistic narrative that women are “too stupid” or “afraid” of AI. She highlights that women are:

  • More sensitive to environmental concerns around AI.
  • Acutely aware of increased scrutiny faced by marginalized identities in the workplace. While a white man using AI might be seen as a “forward thinker,” a woman or person of color might have their work questioned or be accused of “cheating.”
  • Already burdened by significant emotional and domestic labor, making the expectation to learn yet another complex technology an additional, often overwhelming, demand.

Todd criticizes the “influencer” culture that frames AI adoption as an individual failing for women. “It’s the same thing that we always see of like blaming women for institutional or systemic issues that women are not the cause of,” she argues. This narrative deflects from the systemic biases embedded in AI and tech infrastructure, placing the onus on individual women rather than addressing deeper structural problems.

Celebrity endorsements, like Witherspoon’s or Mel Robbins’ advice for women to use AI for finances, are also viewed with skepticism. These powerful figures, often part of an “owner class,” see AI as a business advantage. They leverage their familiarity and “woman-to-woman” persona – like Witherspoon making a smoothie in her kitchen – to sell AI adoption to a general audience, despite likely relying on human experts for their own complex needs. “She’s definitely not using Microsoft Copilot to manage her finances,” Todd quips about Mel Robbins.

AI in the Literary World: The Goodreads Scandal and Beyond

The publishing industry has also grappled with AI’s disruptive force, a crisis foreshadowed by the “Goodreads scandal” of 2024. In that incident, an author was caught using fake aliases to “review-bomb” competing authors while giving her own book glowing reviews.

This paved the way for more recent controversies, such as author Mia Ballard’s book Shy Girl, which faced claims of being AI-written. The ensuing backlash was so severe that the publisher, Hatchet, canceled its US publication.

Todd points out that publishers often “want it both ways,” desiring to use AI when it benefits them but lacking clear policies, leaving individual authors vulnerable. “It’s not really a great system if we’re just setting up individual writers or people to absorb all the negative blowback that might come with using AI,” she states. She advocates for publishing companies to have “completely transparent and fleshed out” AI policies in their contracts, reflecting the realities of 2026.

The pervasive and often insidious uses of AI highlight that technology, far from being a neutral force, amplifies existing societal problems like racism, sexism, and misogyny. How do we account for these deeper harms that go beyond labor issues and plagiarism?

Todd offers a starting point: individual mindfulness. When scrolling through feeds, if content sparks a strong emotional reaction – whether pride, disgust, or even joy – pause and reflect. “Why am I experiencing it? Who is on the other side of this interaction who might be profiting from me feeling this way?” she advises. This awareness, she suggests, is crucial for navigating a digital world where content is explicitly designed to manipulate emotions.

However, Todd is quick to emphasize that the burden shouldn’t solely fall on individuals. “We should not have to do this,” she declares, acknowledging that people are often vulnerable while consuming media, scrolling half-asleep in bed. The expectation for everyone to maintain a constant, high level of critical thought about their media habits is unreasonable.

Ultimately, the conversation points to a profound need for better systemic solutions and greater accountability from platforms and AI developers. Until robust safeguards are in place, users must remain vigilant, questioning the origins and intentions behind the captivating, often unsettling, content generated by AI. The battle against exploitation in the age of generative AI is not just technological; it’s a social and ethical reckoning that demands collective attention.

On-screen text shows the song title 'Phenomenal Black Woman' by China Styles, from the album 'I Love Me'.

On-screen text shows the song title ‘Phenomenal Black Woman’ by China Styles, from the album ‘I Love Me’.

A still from an AI-generated video depicting a Black woman in a grocery store, with 'Sora' or 'V3' branding visible.

A still from an AI-generated video depicting a Black woman in a grocery store, with ‘Sora’ or ‘V3’ branding visible.

A bar chart from CNBC's survey showing a gender gap in AI perception: 69% of men vs. 61% of women see AI as a 'valuable assistant'.

A bar chart from CNBC’s survey showing a gender gap in AI perception: 69% of men vs. 61% of women see AI as a ‘valuable assistant’.

Reese Witherspoon making a smoothie in her kitchen, demonstrating how celebrities use a 'woman-to-woman' persona to promote AI.

Reese Witherspoon making a smoothie in her kitchen, demonstrating how celebrities use a ‘woman-to-woman’ persona to promote AI.

A screenshot of the Goodreads website, highlighting the platform central to the 'Goodreads scandal' where an author review-bombed competitors.

A screenshot of the Goodreads website, highlighting the platform central to the ‘Goodreads scandal’ where an author review-bombed competitors.


Based on “How Getting Stoned With My Dad Helped Us Heal” from New York Times Podcasts Watch the original video

From Smoke to Healing: A Son’s Journey to Reconcile with His Trickster Father

Episode summary

This Modern Love episode features writer and filmmaker Julian BraveNoiseCat, who shares how smoking weed with his father, Ed Archie NoiseCat, a renowned Indigenous artist, became a profound healing experience. Julian’s father, who struggled with alcoholism and was largely absent during his son’s childhood after leaving when Julian was six, reconnected with Julian when Julian moved back in at 28 to research a documentary on Indian residential schools. During their nights playing “bong hit Scrabble” and sharing stories, Julian learned about his father’s traumatic birth at St. Joseph’s Mission, where he was found in a trash incinerator, a stark example of the systemic abuse and infanticide at these institutions that profoundly impacted generations of Indigenous families.

The practical takeaway is that confronting intergenerational trauma, even through unconventional means like shared cannabis use, can foster deep understanding and reconciliation between family members. Julian, on the cusp of becoming a father himself, realized that understanding his father’s past, marked by the residential school system’s devastating impact on parenting skills, was crucial to breaking cycles of pain and building a more present relationship with his own future child, despite his earlier fears of repeating his father’s mistakes.

Father’s Day, for many, conjures images of backyard barbecues, fishing trips, and heartfelt greeting cards. But for writer and filmmaker Julian BraveNoiseCat, the reality of his relationship with his father, Edwin “Ed” Archie NoiseCat, was far more complex, marked by absence, intergenerational trauma, and an unconventional path to healing that involved late-night sessions of “Bong Hit Scrabble.” Julian’s story, shared on the New York Times’ Modern Love podcast, delves into the profound journey of understanding, forgiveness, and the hopeful breaking of cycles as he prepares to become a father himself.

The Unconventional Bonding Ritual

Julian’s father, Ed Archie NoiseCat, is a larger-than-life figure—a renowned Native artist, a “style icon” with “rizz,” and, as Julian affectionately puts it, “a good hang.” Their unique bonding ritual often involved smoking weed and playing games, most notably “Bong Hit Scrabble,” a peculiar blend of wordplay, mathematics, and cannabis consumption. Julian recounts his father’s legendary (and often non-dictionary) words, like “slopify,” highlighting a playful, trickster spirit that defined their interactions.

But beneath the laughter and the hazy camaraderie lay a deeper, more painful history. Julian didn’t grow up with his father; Ed left when Julian was six or seven, a ghost-like presence who would “not show up very often” and sometimes disappear entirely for years. This absence left a profound wound, fueling Julian’s anger and a deep-seated fear of repeating the same mistakes.

A Father “With the Wind”: Childhood Scars

Julian’s memories of his father from childhood are vivid but sporadic. He recalls watching his dad drive away, wondering when he’d see him again, or if he ever would. Ed, a “road warrior” with countless unpaid speeding tickets across North America, embodied a restless, almost mythical persona. Julian recounts a particularly striking memory from age 12, driving with his dad who was speeding, blasting Mike Jones, and confidently declaring, “Son, I have Crazy Horse medicine. The cops can’t see me. They can’t catch me.”

While Julian knew his dad was “full of shit,” he also couldn’t fully disbelieve the sentiment. There was an otherworldly quality to Ed, a sense of magic and survival that defied logic. This duality—admiration for his father’s charisma and resilience, coupled with the pain of his abandonment—defined Julian’s early years. His mother even prepared him at 12 for the “real possibility” that his father might die, a stark reflection of Ed’s self-destructive streak and struggles with alcoholism.

By adolescence, Julian actively sought to be “nothing like” his father. He recognized Ed’s immense artistic talent but also saw how “all this stuff that he haven’t dealt with,” particularly his drinking, consistently “got in his own way.” The fear of inheriting this “Achilles’ heel” and becoming the same kind of absent father became a driving force for Julian. At 16, he confronted his father, a tearful outpouring of anger and pain over the abandonment, though he admits his father’s history of survival often served as a “justification” for his actions.

Unearthing Intergenerational Trauma: The Residential School Legacy

The turning point in their relationship came when Julian, at 28, decided to move back in with his father. He was working on a book and a documentary, Sugarcane, which explored the legacy of Indian residential schools in Canada. The 2021 discovery of 215 potential unmarked graves at an Indian residential school in Kamloops sparked an international reckoning, and it resonated deeply with Ed. Julian’s father had grown up on the Canim Lake Indian Reserve in British Columbia, and there was a long-standing rumor that he had been born in Williams Lake and “found in a dumpster.”

Through his research for Sugarcane, Julian uncovered the shocking truth of his father’s birth. Ed was born at St. Joseph’s Mission, a residential school, on August 16, 1959. His grandmother, who had attended the school and was working there as a nurse, hid her pregnancy. After the baby was delivered, it was found in the trash incinerator. This infant, “Baby X” in the Williams Lake Tribune, was Julian’s father. The newspaper, even in 1959, questioned the “procedures and policies and practices at this institution” that could lead a young mother to such an act. Julian’s documentary revealed a pattern of infanticide at the school, highlighting the horrific realities of these institutions.

Julian painstakingly explains the devastating ripple effects of these schools: “kids who were taken away, who were abandoned when they were kids, turned around and they didn’t really know how to be parents.” This cycle of trauma, violence, and abuse explained much of his father’s struggles, including his alcoholism and his inability to be a present parent. Understanding this history provided crucial context for Ed’s actions, allowing Julian to begin to unravel the complex tapestry of their relationship.

The Coyote’s Shadow: A Cultural Lens for Understanding

Living and working alongside his father, Julian observed Ed’s creative process—carving to Led Zeppelin, taking bong rips, a “vibe” that contrasted sharply with Julian’s solitary writing. During this time, Julian delved into his own Indigenous culture, studying the trickster coyote, the first ancestor of his people. This figure, a “patriarch figure from the beginning of time who made the world… but was a trickster, and so he messed a lot of stuff up,” resonated deeply. The coyote was a “creator destroyer survivor deadbeat dad.”

Suddenly, looking from his laptop to his father in the carving studio, the connection clicked: “My father is the trickster coyote.” This realization provided Julian with a language, a narrative, and a character that helped him understand his father’s complexities—his charisma, his resilience, his talent, and his profound flaws. It was a way of understanding his dad that also helped him reclaim something lost by the residential schools, connecting his personal pain to a broader cultural narrative. While Julian acknowledges that this “coyote” lens might also offer his father “a way out” of accountability, he believes it “deepens the truth of the whole thing,” recognizing the ancestral patterns that manifest in human lives.

An Apology and the Hope for a New Generation

It wasn’t until after Julian’s book was published, and his father finally read an advanced copy, that the long-awaited apology came. Ed expressed understanding for the pain he had caused and apologized for it, also sharing his immense pride in Julian’s artistic achievement. Julian sees his own art-making, though different from carving, as an echo of his father’s legacy, fulfilling a desire for sons to become “little versions of them out in the world.”

Now, Julian BraveNoiseCat is about to become a father himself, welcoming a son at the very same age his own father was when Julian was born: 33. This serendipitous timing adds another layer of significance to his journey. He admits to a deep-seated fear of repeating his father’s mistakes, of “messing things up” and not being present. Yet, he holds onto the hope of “getting it right,” of making “fewer mistakes than the last generation,” and ultimately, of “bring[ing] our people back from the genocide.”

Julian dreams of a “magical relationship” with his son, filled with cultural activities—hockey, dancing, and, most poignantly, bringing his son into the studio to sit with “his Ed” and learn to carve, hoping the artistic legacy will “skip a generation.” When asked how he counters his fears, Julian humorously admits, “I usually take out a pre-roll.” But he also looks at the “fullness of the story with my own father,” a narrative that has transformed from one of “brokenness and pain” to one of “love and repair.” Their relationship, despite the messiness and past transgressions, has become one where they are “best friends,” making “little coyote trickster adventures of our own, in a beautiful way.”

Julian believes his father will be an “amazing grandparent,” a “rock” who will “nail it” in his “pa era.” Indeed, in April of this year, Julian and his partner, Joan, welcomed their son, Copper Alexander Martin Noise Cat, a “little bear of a kid.” Ed Archie NoiseCat was present in the delivery room, perhaps the most nervous person there, witnessing the continuation of a lineage now steeped in both historical pain and profound hope. Julian’s story is a powerful testament to the enduring human capacity for healing, the complex interplay of personal and cultural identity, and the transformative power of love across generations, even when the path is unconventional and paved with smoke.


Based on “Are you anxious, avoidant, or secure? | The Gray Area” from Vox Watch the original video

The Hidden Power of Secure Attachment: Rewriting Your Relationship Story

Episode summary

Vox’s “The Gray Area” features Amir Levine, author of Secure, who asserts that attachment styles (anxious, avoidant, secure, fearful-avoidant) are not fixed categories but a spectrum, changeable throughout life. He cites a 30-year longitudinal study showing parental behavior contributes only 3% to adult attachment style, while early friendships contribute 11%, challenging the notion that childhood dictates adult attachment.

Levine argues that feeling secure is about establishing a “social baseline” of safety, which is achievable through “secure priming therapy.” This involves identifying and focusing on “secure kernels”—past or present positive relational experiences often overlooked—and rewriting self-defeating “scripts.” He emphasizes that seemingly insignificant micro-interactions, like a non-response to a text, are crucial, as the brain, wired for tribal survival, interprets social exclusion (the “cyber ball effect”) as a threat, regardless of conscious reasoning.

We are inherently social creatures, wired for connection. This fundamental need shapes our experiences of both profound suffering and deep healing. At the core of this human experience lies our attachment system – an intricate hierarchy in our brains that dictates how we seek comfort, navigate intimacy, and respond to perceived threats in our relationships. As Dr. Amir Levine, a psychiatrist and neuroscientist, explains, a single word or a hug from someone with whom we share a secure attachment has a power no medication can rival, instantly making us feel better. Conversely, insecure relationships are often the most potent instigators of emotional distress. Understanding and cultivating secure attachment, therefore, isn’t just about better relationships; it’s about the very basis of our well-being.

Decoding Your Attachment Style: Anxious, Avoidant, or Secure?

Initially identified in babies, attachment styles are now understood to persist throughout our lives, profoundly influencing our adult relationships. Dr. Levine, co-author of the influential book Attached, and now Secure, breaks down the primary styles based on two dimensions: our comfort with intimacy and our sensitivity to potential danger in a relationship.

A diagram illustrates two axes: 'Comfort with Intimacy' and 'Sensitivity to Danger,' which define attachment styles.

A diagram illustrates two axes: ‘Comfort with Intimacy’ and ‘Sensitivity to Danger,’ which define attachment styles.

  • Anxious Attachment: Individuals with an anxious attachment style desire closeness but are constantly vigilant for signs of rejection or abandonment. “If we love to be close but we are constantly worrying about, is the partner available to me? Why are they not answering my call? What’s happening with them? Where did they go?” Dr. Levine describes. They often feel relationships are unstable, prone to worry and seeking reassurance.
  • Avoidant Attachment: While avoidant individuals also desire relationships (as humans are “highly, highly social species”), they become uncomfortable with too much closeness once in one. They tend to maintain a measure of distance, often driven by a life script emphasizing independence and self-sufficiency. “Oh my God, this person is too close to me, like stay away,” describes their internal experience.
  • Secure Attachment: These are the individuals who “love closeness but also are not too sensitive for potential danger.” They are comfortable with both intimacy and autonomy. If a partner needs space or time alone, secure individuals don’t perceive it as a threat to the relationship. They are characterized by warmth, love, and a fundamental trust in the stability of their bonds.
  • Fearful-Avoidant (Disorganized) Attachment: A rarer combination of anxious and avoidant traits, these individuals fantasize about closeness but, once in a relationship, become highly sensitive to potential danger. They experience an internal push-pull: “with one hand you say come close and the other hand stay away, stay away all the time.”

Beyond Labels: Attachment as a Dynamic Spectrum

While these categories provide a useful framework, Dr. Levine emphasizes a crucial, hopeful insight: attachment is not a fixed trait. “It’s not like a category that we fall into. It’s a spectrum,” he asserts. We can exhibit different attachment styles with different people, and importantly, we have the capacity for change. This understanding forms the bedrock of his new work, Secure, which posits that “we can all learn to live in secure mode.”

Security, in this context, is less about the absence of anxiety and more about stability. It’s our brain’s quest for a social baseline, a “safe baseline,” much like a biological system striving for homeostasis. Dr. Levine’s new questionnaire helps individuals map their “attachment topography,” providing a personalized understanding of their social world from which to build a “roadmap to greater security.”

The journey to security involves understanding the “secure brain”—what it needs to feel safe and what makes it insecure. Many societal norms, Dr. Levine suggests, are actually misaligned with the fundamental needs of our social brains, contributing to widespread insecurity.

Challenging the Past: The Limited Influence of Childhood

One of the most surprising and liberating insights from Dr. Levine’s work is the re-evaluation of childhood’s causal role in adult attachment. While traditional therapeutic narratives often trace adult patterns directly to early experiences, Dr. Levine, who pivoted from psychotherapy to molecular neuroscience, questions this neat causality. “When I became a scientist, I discovered how hard it is to establish causality,” he explains.

Recent longitudinal studies, some following individuals for 30 years, reveal a surprisingly small impact of parental behavior on adult attachment style. Maternal effect, for instance, contributed only about 3%. Early childhood friendships showed a slightly higher influence at 11%. This isn’t to say childhood experiences are irrelevant, but rather that the “causal weight” often placed on them might be “too neat.”

A bar chart shows the surprisingly low percentage impact of maternal effect (3%) and early childhood friendships (11%) on adult attachment style.

A bar chart shows the surprisingly low percentage impact of maternal effect (3%) and early childhood friendships (11%) on adult attachment style.

The danger of overemphasizing childhood causality, Dr. Levine argues, is that it can lead individuals to feel “damaged goods” and unable to change. Instead, he highlights that “a lot of the change that happens happens if we actually change the brain, change the environment for the brain in the here and now.” This perspective shifts the focus from an unchangeable past to the empowering potential of the present.

Rewriting Your Story: The Power of Secure Kernels

Attachment styles, Dr. Levine explains, are essentially “scripts” we tell ourselves about ourselves and our relationships: “I can’t trust anyone,” or “I can’t say anything because the relationship will break up.” Our brains, sophisticated machinery that they are, then actively seek out evidence to reaffirm these scripts, often ignoring contradictory information.

Secure Priming Therapy aims to disrupt this cycle by helping individuals identify “secure kernels”—those past or present experiences where they felt truly safe and connected. Dr. Levine shares a personal anecdote of a vacation with his sister and her “very, very secure” friend’s mother, Ruth, a memory he frequently returns to for its warmth and security. “Many of us have those experiences,” he notes, “but our brain chooses to ignore them or they don’t take center stage.” The therapy encourages refocusing on these secure occurrences, “summoning out that potential that I believe resides in the vast majority of us.”

We are, as the host observes, “narrative creatures,” and while this allows us to make sense of the world, it can also become a curse when we cling to stories that make us miserable. The ability to rewrite these internal scripts is key to cultivating greater security.

The Sting of Exclusion: Our Deeply Social Brain

Why do seemingly minor social slights, like being ignored or ghosted, sting so much? Dr. Levine points to the powerful and often unconscious workings of our “social brain.” He illustrates this with the “Cyberball effect,” a series of experiments where participants playing a virtual game of catch are suddenly excluded by the other players. Even when offered money for exclusion or told the other players are members of a despicable group like the KKK, the brain’s response is consistent: areas associated with painful distress and self-scrutiny light up. Psychologically, it leads to decreased self-esteem, a reduced sense of control, and a feeling that life is less meaningful.

A screenshot from the 'Cyberball' game shows three stick figures playing catch, with one player being excluded.

A screenshot from the ‘Cyberball’ game shows three stick figures playing catch, with one player being excluded.

This profound reaction, Dr. Levine explains, has deep evolutionary roots. His experience on an African safari provided a vivid analogy: walking single file, guides constantly urged the group to “close the gap” between them. A gap meant vulnerability, an opening for predators. Similarly, for our emotional brain, formed in a time when tribal belonging meant survival, being ostracized was literally a matter of life or death. Our brain, therefore, interprets social exclusion, even minor slights, as a significant threat, activating ancient survival mechanisms. “The brain’s not making any distinctions here,” between a life-or-death threat and a friend not calling back for three days, he notes.

Building a Secure Life: The CARP Framework

If exclusion is so potent, what nurtures security? Dr. Levine introduces the CARP acronym, representing the five pillars of a secure life: Consistent, Available, Responsive, Reliable, and Predictable.

He discovered the power of these elements through “reverse Cyberball” experiments, where individuals are “hyper-included” (e.g., always receiving the ball). This experience leads to increased self-esteem, a greater sense of control and meaningfulness, reduced inflammation, and even greater longevity. The CARP framework aims to replicate this hyper-inclusion in real-life relationships.

Predictability, in particular, is often misunderstood. It’s not about being boring, but about being a steady presence. Just as a baby occasionally checks for a parent’s presence to feel safe enough to play, adults need “momentary check-ins” to ensure the “thread that connects us” is still intact. When this thread feels secure, the relationship fades into the background, providing a stable foundation for life. Unpredictability, conversely, triggers the brain’s radar, leading to self-scrutiny and anxiety.

These “seemingly insignificant minor interactions of everyday life” are incredibly important for the brain to register and build a secure mode.

For those seeking to cultivate more secure relationships, Dr. Levine offers two key “rules of secure engagement,” rooted in understanding attachment logic:

  1. Only one person is allowed to be upset at a time: This rule acknowledges that secure relationships are fundamentally about helping each other feel calmer. When one person is distressed, the other’s role is to provide comfort and support, not to become upset themselves. This mutual commitment to emotional well-being is the hallmark of a secure bond.
  2. The Mia Kulpa Rule (Fallback Rule): If both individuals become upset, both must apologize. This isn’t about who is “right” or “wrong,” but about acknowledging that the “covenant of a secure bond” – the shared responsibility for each other’s emotional well-being – has been momentarily broken. Prioritizing emotional resolution and reconnection over winning an argument allows the attachment system to settle down. “If you apologize and you can connect, then tomorrow you can discuss who’s right and who’s wrong,” Dr. Levine advises.

Practical Tools for Managing Insecurity

For individuals with anxious attachment, Dr. Levine identifies a common pattern: the Protest-Regret Cycle. This involves lashing out when feeling insecure, leading to defensiveness from the partner, creating distance. The anxious individual then regrets the outburst and apologizes, but the apology focuses on the lashing out, not the underlying “non-CARP behavior” that triggered the initial distress. This perpetuates a loop where genuine attachment needs are never addressed.

To break such cycles and “rightsize” relationships, Dr. Levine offers a tool called Wall Tennis with Love. When dealing with a friend or partner who is often inconsistent or unresponsive, the anxious individual becomes “the wall.” They respond immediately and lovingly when the other person initiates contact, maintaining the attachment thread without initiating too much themselves. “Whatever you dish the wall, the wall will return with maybe a tiny less velocity in the same direction. It never initiates,” he explains. This strategy prevents the anxious individual from experiencing the “Cyberball effect” of unreturned texts or calls, allowing their attachment system to stay quiet, achieving the desired “homeostasis.”

The Promise of Change: Agency Over Fate

The message is clear: we are not trapped by our past or by labels like “anxious” or “avoidant.” While the host reflects on wishing he had this information earlier in life, Dr. Levine affirms that change is always possible. “Didn’t you just say that change in your 30s and you were able to do more?” he playfully challenges.

The journey to greater security begins with acceptance. “Accepting okay, this is my biology, this is the animal that I am. Now how do I work with it to actually find a place that’s more convenient and that feels good to me?” This self-knowledge, combined with a willingness to reorganize our social environment and practice new behaviors, empowers us to cultivate security. It’s about agency, not fate. By understanding the deep-seated needs of our social brain and applying practical strategies, we can genuinely rewrite our relationship stories, fostering deeper connections and a more secure, fulfilling life.

On-screen text displays the CARP acronym: Consistent, Available, Responsive, Reliable, and Predictable.

On-screen text displays the CARP acronym: Consistent, Available, Responsive, Reliable, and Predictable.

A diagram illustrates the 'Protest-Regret Cycle,' showing how lashing out leads to distance, regret, and apologies that don't address core issues.

A diagram illustrates the ‘Protest-Regret Cycle,’ showing how lashing out leads to distance, regret, and apologies that don’t address core issues.

A graphic depicts a person playing tennis against a wall, symbolizing the 'Wall Tennis with Love' strategy.

A graphic depicts a person playing tennis against a wall, symbolizing the ‘Wall Tennis with Love’ strategy.


한국어

“How GitHub Deals with 17 Million Pull Requests a Month”Every 기반 기사 원본 영상 보기

월 1,700만 개의 풀 리퀘스트: GitHub, AI 시대 개발자 생태계의 미래를 그리다

에피소드 요약

GitHub의 COO 카일 데이글(Kyle Dagel)은 매달 1,700만 건의 풀 리퀘스트(Pull Request, PR)가 발생하는 등 기하급수적으로 증가하는 코드 생성량에 따라 개발자 도구의 지형이 변화하고 있음을 강조한다. 과거 개발자라고 불리지 않던 지식 노동자들까지 GitHub 코파일럿(Copilot) 같은 AI 도구를 활용해 코드를 작성하면서 개발자의 정의가 확장되고 있으며, 이는 GitHub의 제품 로드맵에도 영향을 미쳐 누구나 쉽게 코딩을 시작할 수 있도록 진입 장벽을 낮추는 데 집중하고 있다. 특히, 에이전트(Agent)가 생성하는 PR이 급증하면서 오픈소스 관리자들의 부담이 커지고 있는데, GitHub는 코파일럿 코드 리뷰(Copilot Code Review)와 에이전트 병합(Agentic Merge) 같은 도구를 통해 PR 처리 과정을 자동화하고, 관리자들이 PR 수락 여부나 기여자 자격 등을 스스로 결정할 수 있는 통제권을 제공하는 데 주력하고 있다.

이러한 변화는 비즈니스 모델에도 영향을 미쳐, 에이전트가 24시간 작동하는 환경에서 기존의 프리미엄(freemium) 모델 대신 사용량 기반(usage-based) 모델로의 전환 가능성을 시사한다. GitHub는 개발자들에게 필요한 핵심 경험을 제공하면서도, 더 많은 에이전트 활용을 원하는 사용자들을 위한 유료 모델을 모색 중이다. 또한, 마이크로소프트(Microsoft)의 개발자 마케팅 총괄(Chief Marketing Officer of Developer) 역할까지 겸하고 있는 카일 데이글은 개발자 중심의 철학을 마이크로소프트 전체에 전파하며, 다양한 AI 모델 및 코딩 에이전트 파트너들과 협력하여 개발자에게 선택의 폭을 넓혀주는 것을 핵심 차별점으로 삼고 있다. 궁극적으로는 토큰 가격의 효율성을 높이고 개인화된 에이전트 경험을 제공하기 위해 ‘언덕 오르기(hill climbing)’ 방식을 통해 지속적인 개선과 실험을 반복하며 장기적인 비전을 구축하고 있다.

인공지능(AI) 에이전트가 코딩의 최전선에 서면서 개발 생태계는 전례 없는 변화의 물결을 맞고 있습니다. 한 달에 무려 1,700만 개의 풀 리퀘스트(Pull Request, PR)가 AI 에이전트로부터 쏟아지는 GitHub는 이러한 변화의 최전선에서 새로운 경제의 흐름을 주도하고 있습니다. 최근 에브리(Every)의 기술 컨설팅 책임자 마이크 테일러(Mike Taylor)는 GitHub의 최고운영책임자(COO)이자 마이크로소프트(Microsoft) 개발자 부문 최고 마케팅 책임자(CMO)인 카일 데이글(Kyle Dagel)과 만나, 폭증하는 코드 생성과 진화하는 개발자 커뮤니티 속에서 GitHub가 나아가야 할 방향에 대해 심도 깊은 대화를 나눴습니다.

이번 인터뷰에서는 AI 에이전트의 급격한 성장이 사용자 경험과 오픈소스 유지보수자에게 미치는 영향, 그리고 월 200달러짜리 코딩 에이전트 구독료가 2,000달러로 치솟는 것을 막을 방법 등 개발자들의 뜨거운 관심사를 다뤘습니다. 특히, 마이크 테일러가 인터뷰 연습을 위해 카일 데이글의 AI 클론을 만들었다고 밝히자, 카일 데이글 역시 AI를 활용한 개인적인 자기 개선 경험을 공유하며 놀라움을 안겼습니다.

개발자 생태계의 변화와 GitHub의 대응

’개발자’의 정의 확장

카일 데이글은 GitHub가 오랫동안 ‘개발자’의 개념을 광범위하게 정의해왔다고 강조합니다. 그 자신도 컴퓨터 과학을 전공하지 않고 미술 학교 학비를 벌기 위해 코드를 작성했던 경험을 예로 들며, 이전에는 개발자 도구를 사용하지 않았던 많은 사람들이 이제 GitHub 제품을 사용하고 있다고 설명했습니다. “과거에는 스스로를 개발자라고 부르지 않았지만 코드를 작성했던 사람들이 많았습니다. 저도 미술 학교 학비를 벌기 위해 코드를 썼죠. 이제는 GitHub 코파일럿(Copilot) 앱을 사용하는 법무팀이나 재무팀 직원, 또는 지식 노동자(Knowledge Worker)들이 자신만의 작은 앱이나 자산을 구축하는 데 이 도구를 활용하고 있습니다.” GitHub는 여전히 전문 개발자 도구에 중점을 두지만, GitHub 코파일럿 앱과 같은 도구를 통해 누구나 쉽게 코딩을 시도하고 소프트웨어 개발의 길로 들어설 수 있도록 진입 장벽을 낮추는 데 집중하고 있습니다.

폭증하는 풀 리퀘스트, 오픈소스 유지보수자의 고충

AI 에이전트의 등장은 풀 리퀘스트(PR)의 양을 기하급수적으로 늘려, 특히 오픈소스 프로젝트 유지보수자들에게 막대한 부담을 안겨주고 있습니다. 카일 데이글은 이러한 문제를 해결하기 위해 GitHub가 제공하는 몇 가지 핵심 도구를 소개했습니다.

  • 코파일럿 코드 리뷰(Copilot Code Review): AI 에이전트가 코드 취약점을 찾아내고, 개발자의 피드백을 받아 자동으로 변경 사항을 구현하여 PR을 보다 쉽게 검토할 수 있도록 돕습니다.
  • 에이전트 기반 병합(Agentic Merge): PR을 처리하는 데 필요한 수동 단계를 자동화합니다. 개발자가 허용할 작업을 설정하면, 에이전트가 CI(Continuous Integration) 확인, 정책 준수 등 복잡한 과정을 거쳐 PR 병합을 완료합니다.

오픈소스 프로젝트의 경우, 누가 PR을 보낼지 통제하기 어렵다는 특수성이 있습니다. GitHub는 유지보수자들이 어떤 PR을 수락할지, 누구로부터 받을지, 기여자의 기여가 프로젝트에 얼마나 의미 있을지 결정할 수 있는 더 많은 도구를 제공하고 있습니다. 데이글은 “GitHub는 항상 유지보수자들에게 도구를 제공하고 권한을 부여하되, 최종적인 결정권은 그들에게 맡기려 한다”고 강조했습니다. 특정 표준을 강요하기보다, 각 커뮤니티의 자율성을 존중하며 최적의 방식을 찾아가도록 돕는 것이 GitHub의 철학입니다. 예를 들어, 미첼 하시모토(Mitchell Hashimoto)가 공유한 ‘바우치(Vouch)’ 시스템처럼, 커뮤니티가 자체적으로 검증 시스템을 구축하도록 지원하는 방식입니다.

에이전트 경제 시대의 GitHub

코드 생성의 폭발적 증가와 미래 준비

카일 데이글은 GitHub가 현재 경험하고 있는 성장을 “새로운 에이전트 경제(Agent Economy)의 최전선에 있는 것 같다”고 표현했습니다. 지난해 GitHub 유니버스(Universe)에서 연간 10억 건의 커밋(Commit)이 발생했다고 발표했지만, 올해는 선형적으로 성장하더라도 140억 건에 달할 것으로 예상됩니다. 특히 지난 3월에는 AI 에이전트가 생성한 풀 리퀘스트만 1,700만 건에 달했습니다.

데이글은 이러한 폭발적인 코드 생성에 대해 “이 모든 것이 그저 쓸모없는 코드는 아니다”라고 반박합니다. 오히려 이는 개발자들이 AI 에이전트와 함께 ‘카일 + 1에서 N개의 에이전트’ 형태로 협력하며 새로운 것을 구축하는 초기 단계를 넘어선 지점임을 시사합니다. GitHub는 이러한 성장이 둔화되지 않고 계속될 것으로 보고 있으며, 모든 개발자의 ‘에이전트 순간’을 지원하기 위해 대규모 투자를 진행하고 있습니다. 모든 코드가 결국 GitHub에 공유되고 협업이 이루어지기 때문에, GitHub는 이 새로운 시대의 핵심 인프라 역할을 할 수밖에 없다는 것입니다.

비즈니스 모델의 진화와 사용량 기반 요금제

AI 에이전트가 우리가 잠든 사이에도 계속해서 코드를 생성하는 상황에서, 기존의 프리미엄(Freemium) 비즈니스 모델은 어떻게 변화할까요? 카일 데이글은 아직 최종적인 모델은 알 수 없다고 말합니다. 현재는 개인 사용자가 라이선스를 갖거나 무료 GitHub.com을 사용하는 방식이지만, API 호출 제한(Rate Limit)과 같은 부분에서 에이전트 사용으로 인한 압력이 발생하고 있습니다.

GitHub의 목표는 피터 스타인버거(Peter Steinberger)의 말처럼 “150개의 에이전트가 모든 것을 동시에 처리하는” 것과 같이 훨씬 더 많은 작업을 수행할 수 있도록 지원하는 것입니다. 동시에 핵심적인 GitHub 경험을 훌륭하게 유지하는 것 역시 중요합니다. 데이글은 무료 공개 저장소에서 무료 비공개 저장소로 전환했던 과거의 사례를 들며, GitHub는 항상 산업과 커뮤니티의 변화에 따라 진화해왔다고 설명했습니다. 중요한 것은 개발자들이 성공하는 데 필요한 것을 제공하는 것이며, 기업 고객에게는 대규모 운영에 필요한 것을 제공하는 것입니다.

마이크로소프트 개발자 전략의 중심에서

GitHub COO와 마이크로소프트 개발자 CMO의 이중 역할

카일 데이글은 13년 동안 GitHub에서 개발자로서, 그리고 엔지니어링 팀을 이끌며 일해왔습니다. 그는 GitHub의 독특한 점은 100% 개발자를 위해 도구를 구축한다는 데 있다고 말합니다. 기업이 GitHub 제품을 구매하는 것은 좋지만, 구매자가 아닌 개발자를 위해 제품을 만든다는 원칙은 변함없습니다.

데이글은 현재 GitHub COO 역할과 함께 마이크로소프트 개발자 부문 CMO라는 이중 역할을 수행하고 있습니다. 그의 목표는 마이크로소프트의 모든 개발자 도구와 기술을 아우르며, 개발자 경험에 ‘진정성 있는(Authentic)’ 통합 솔루션을 제공하는 것입니다. 최근 샌프란시스코에서 열린 ‘빌드(Build)’ 컨퍼런스에서는 이러한 접근 방식이 잘 드러났습니다. 전통적인 컨퍼런스 홀 대신 샌프란시스코의 분위기를 살리고, 커뮤니티 연사들을 주요 세션에 초청하는 등 ‘피칭’보다는 ‘경험’에 초점을 맞췄습니다. 데이글은 소프트웨어 개발이 “팀 스포츠”와 같으며, 어떤 한 회사도 모든 질문에 답할 수 없다고 강조합니다. 오픈소스에 기반을 둔 개발 생태계에서 커뮤니티 구성원들을 적극적으로 참여시키는 것이 중요하며, 개발자들은 이러한 외부 관점에 큰 의미를 부여한다는 것입니다.

경쟁 시장에서의 차별화 전략

빠르게 변화하고 경쟁이 치열한 시장에서 GitHub는 어떻게 차별점을 만들어갈까요? 데이글은 GitHub의 뿌리인 ‘개발자 선택(Developer Choice)‘과 ‘빌더를 위한 구축(Building for Builders)‘에 집중한다고 말합니다. 과거에는 수많은 API와 액세스가 존재했지만, 의도치 않게 ‘벽으로 둘러싸인 정원(Walled Garden)‘처럼 특정 생태계에 갇히는 경향이 있었습니다.

GitHub는 개발자들이 GitHub와 함께 구축하면서도 다른 도구들을 자유롭게 사용할 수 있도록 지원하며, 모든 파트너와 협력하여 이를 최대한 간단하게 만들고자 합니다. 코드를 생성하는 측면이나 협업 및 검토 측면뿐만 아니라, 소프트웨어 구축의 모든 과정에 걸쳐 이러한 유연성을 제공하는 것이 GitHub의 진정한 ‘초능력(Superpower)‘이라는 설명입니다.

GitHub는 자체 기술(예: 새로운 마이크로소프트 AI 모델)에 투자하는 동시에, 엔트로픽(Anthropic), 오픈AI(OpenAI), 구글(Google) 등 시장에 모델이나 코딩 에이전트를 내놓는 모든 파트너와 협력할 것입니다. 이러한 ‘선택의 자유’는 GitHub가 결코 포기하지 않을 핵심 가치입니다. 만약 GitHub가 이를 포기한다면, 개발자들은 또 다른 ‘쥐덫(Mouse Trap)‘에 갇히게 될 것이고, GitHub는 소프트웨어 세상이 그렇게 되기를 원치 않는다고 데이글은 강조합니다.

AI 시대의 의사결정 및 개인적 활용

제품 개발과 ‘힐 클라이밍(Hill Climbing)’ 방법론

GitHub 내부에서는 새로운 모델이나 GitHub 코파일럿 데스크톱 앱 등 자사 제품을 ‘도그푸딩(Dogfooding, 자사 제품 사용)‘하는 것과 개발자들이 다른 도구를 실험하도록 허용하는 것 사이에서 어떻게 균형을 잡을까요? 카일 데이글은 모든 팀원이 다양한 도구를 사용한다고 말합니다. 그 자신도 맥북(MacBook), 윈도우 PC, 리눅스(Linux) 박스를 오가며 코딩을 하고, 특히 윈도우용 GitHub 코파일럿 앱만 사용하여 윈도우 개발자들의 경험을 직접 이해하려 노력합니다.

데이글은 “레이저처럼 한 곳에만 집중하는 것은 시야가 좁아질 수 있다(Myopic)“고 경고합니다. 비록 모든 도구를 매일 사용할 수는 없지만, 새로운 도구가 나올 때마다 왜 사람들이 좋은 경험을 하는지 이해하려 노력하고, 이를 통해 개발자 선택이라는 GitHub의 목표를 어떻게 달성할지 고민합니다.

제품 개발에는 ‘힐 클라이밍(Hill Climbing)’ 방법론이 핵심적으로 적용됩니다. 사티아 나델라(Satya Nadella), 무스타파 술레이만(Mustafa Suleyman), 제이콥 러시(Jacob Rush) 등 마이크로소프트 리더들이 강조하는 이 방식은 도구 사용 자체를 모델 개선의 핵심 동력으로 삼는 것입니다. ‘좋아요/싫어요’ 데이터, 코드 수락률 등 방대한 사용자 데이터를 분석하여 모델을 지속적으로 개선합니다. 매주 ‘힐 클라이밍’ 결과를 논의하며, 정량적 지표뿐만 아니라 사용자 감성(User Sentiment)과 같은 정성적 지표도 함께 고려합니다. 때로는 정량적 개선이 사용자 경험을 저하시킬 수도 있기 때문입니다. 데이글은 “이는 문샷(Moonshot)이 아니라, 그저 계속해서 오르고, 개선하고, 새로운 평가를 하고, 새로운 데이터를 반영하며 나아가는 과정”이라고 설명합니다. 특히 기업 고객에게는 마이크로소프트 365(M365) 데이터 기반의 ‘프론티어 튜닝(Frontier Tuning)‘과 같은 도구를 통해 이러한 개선 과정을 훨씬 쉽게 만들 수 있습니다.

에이전트 비용 관리와 모델 라우팅

월 200달러짜리 구독료가 2,000달러로 폭증하는 것을 막는 방법은 무엇일까요? 카일 데이글은 모델을 더 잘 이해하도록 ‘프론티어 튜닝’하는 것 외에도, 개발자들이 자동으로 모델을 선택할 수 있도록 돕는 것이 중요하다고 말합니다. ‘모델 라우터(Model Router)‘와 같은 기술이 그 해답이 될 수 있습니다. 이는 작업 의도(Task Intent)에 따라 적절한 모델을 자동으로 선택해주는 기능으로, 마이크로소프트 파운드리(Foundry)도 API 수준에서 이러한 모델 라우터를 제공합니다.

개발자들이 해결하려는 문제의 난이도(‘매우 어려운 문제이니 최고 성능 모델을 써도 좋다’ 또는 ‘간단한 작업이니 저렴한 모델을 써도 된다’)를 시스템에 알려주면, 도구가 자동으로 최적의 모델을 선택하여 토큰 비용을 절감할 수 있다는 것입니다. 데이글은 개인적으로도 어려운 문제를 해결하다가 갑자기 간단한 작업(예: 이름 바꾸기)으로 전환될 때, 비용 절감을 위해 모델을 수동으로 변경하는 경우는 드물다고 고백합니다. 그러나 이러한 자동화된 도구는 기업뿐만 아니라 개인 개발자에게도 큰 도움이 될 것이라고 강조했습니다.

AI를 통한 개인적 자기 개선

인터뷰 막바지에 마이크 테일러가 인터뷰 연습을 위해 카일 데이글의 AI 클론을 만들었다는 ‘약간은 이상한’ 고백을 하자, 데이글은 자신도 비슷한 일을 하고 있다고 밝혀 놀라움을 안겼습니다. 그는 자신이 작성하는 모든 것을 읽고 피드백을 주는 AI 클론(개인용 ‘클로(Claw)‘와 업무용 앱)을 가지고 있습니다. 매일 ‘카일, 당신은 계속 이런 말을 합니다’, ‘이 부분은 명확하지 않습니다’와 같은 커뮤니케이션 보고서를 받는다고 합니다. 특히 비유를 많이 사용하는 자신의 말하기/쓰기 습관에 대해 더 명확한 비유를 제안받기도 합니다.

데이글은 이러한 AI 에이전트로부터 얻는 ‘인간으로서의 자기 개선 루프’가 엄청나게 강력하다고 말합니다. 과거 GitHub의 챗옵스(ChatOps) 도구였던 ‘휴봇(Hubot)’ 시절부터, 사람들은 다른 인간보다 로봇에게서 비판적인 피드백을 훨씬 기꺼이 받아들인다는 사실을 알고 있었습니다. 그는 자신의 AI 클론 ‘백스터(Baxter)‘가 “내가 뭔가 형편없게 했다고 말할 때, 나는 훨씬 기분 좋게 ‘왜?‘라고 묻는다”고 말하며, 이 피드백을 이메일 작성, 스크립트 검토 등에 활용하여 글쓰기 및 말하기 습관을 개선하고 있다고 합니다.

결국 AI는 단순히 소프트웨어를 만들고 코드를 생성하는 도구를 넘어, 인간의 자기 개선을 돕는 강력한 파트너가 될 수 있음을 시사하는 대목입니다. 카일 데이글은 이처럼 ‘재귀적으로 자기 개선(Recursively Self-Improve)‘하는 것이 AI 시대의 중요한 과제이자 기회가 될 것이라고 강조하며 인터뷰를 마쳤습니다. GitHub는 이 격변의 시대에 개발자들의 창의성과 생산성을 극대화하는 동시에, 새로운 도전에 유연하게 대응하며 미래를 선도하는 역할을 계속해 나갈 것입니다.


“Generative AI’s race problem | It’s Been a Minute”NPR Podcasts 기반 기사 원본 영상 보기

생성형 AI, 인종적 편견을 먹고 자라다: 디지털 블랙페이스부터 사기까지, AI 시대의 어두운 그림자

에피소드 요약

NPR의 팟캐스트 ‘It’s Been a Minute’는 생성형 AI가 인종차별적 고정관념을 강화하고 착취하는 문제를 심도 있게 다룬다. 게스트 브리짓 토드는 AI가 생성한 ‘Phenomenal Black Woman’ 같은 노래나 흑인 아동이 수제 가방을 만들다 인종차별을 당하는 스캠성 영상처럼, AI 콘텐츠가 특정 감정을 유발하도록 설계되어 있음을 지적한다. 이러한 영상들은 흑인 커뮤니티의 연대감을 악용하거나, 특정 집단에 대한 기존의 편견과 두려움을 강화하며, 심지어는 유료 성인 콘텐츠 사이트로 연결되는 등 다양한 방식으로 금전적 이득을 취한다. 특히 AI가 인종적 고정관념을 대규모로 ‘AB 테스트’하여 가장 효과적인 콘텐츠를 찾아내는 방식은 과거 민스트럴 쇼(minstrel show)와 유사하게 작동하며, 실제 흑인 창작자를 배제하고 흑인 정체성을 상품화하는 데 일조한다.

이러한 문제에 대응하기 위해 시청자들은 소셜 미디어 피드에서 강한 감정적 반응을 일으키는 영상에 대해 ‘이것이 진짜인가, AI 생성물인가?‘라는 질문을 던지며 잠시 멈춰 생각하는 습관을 들여야 한다. 또한, AI 기술 채택에 있어 여성들이 남성보다 회의적인 태도를 보이는 현상에 대해, 리스 위더스푼(Reese Witherspoon) 같은 유명인이 AI 사용을 독려하는 것은 개인이 아닌 시스템적 문제의 책임을 여성에게 전가하는 ‘걸보스화(girlbossification)‘의 일환일 수 있음을 인지해야 한다. 즉, AI는 기술적 중립성을 넘어 사회적 편견과 불평등을 재생산하고 심화시킬 수 있으므로, 비판적 사고와 윤리적 접근이 필수적이다.

생성형 AI(Generative AI) 기술이 급부상하면서 우리 사회에 미치는 영향에 대한 논의가 활발합니다. 일자리 감소, 표절 문제 등 다양한 우려가 제기되는 가운데, NPR 팟캐스트 ‘It’s Been a Minute’에서는 특히 인종과 젠더 문제를 중심으로 생성형 AI의 어두운 이면을 심층적으로 다루었습니다. 문화 및 디지털 문화 전문가인 브리짓 토드(Bridget Todd)는 AI가 어떻게 인종적 고정관념을 악용하고, 사기성 콘텐츠를 확산하며, 심지어 정치적 불화를 조장하는 데 사용되는지 경고합니다. 이 글은 그들의 대화를 바탕으로, 생성형 AI가 사회에 미치는 복잡하고 종종 해로운 영향에 대해 탐구합니다.

‘경이로운 흑인 여성’과 AI의 불편한 진실

논란의 중심에 선 ‘Phenomenal Black Woman’

방송의 시작은 진행자 브리트니 루즈(Britney Loose)가 틱톡 피드에서 너무나 자주 접했다는 한 노래, ‘Phenomenal Black Woman’에 대한 이야기였습니다. 이 곡은 차이나 스타일스(China Styles)라는 아티스트의 2025년 앨범 ‘I Love Me’에 수록된 곡으로, 노래, 앨범, 심지어 아티스트까지 모두 AI가 생성한 것으로 밝혀져 큰 화제가 되었습니다.

이 노래는 최근 몇 달간 흑인들이 자신의 어머니, 이모, 할머니에게 들려주며 반응을 살피는 소셜 미디어 트렌드의 중심에 서기도 했습니다. 테니스 선수 코코 고프(Coco Gauff)의 어머니가 이 노래를 듣고 처음에는 짜증 섞인 표정을 짓다가 “이제 됐어. 충분히 들었어”라고 말하는 영상은 특히 인기를 끌었습니다. 브리트니는 많은 흑인 여성들이 이 곡에서 의미를 찾고 감동을 받는다는 사실에 마음 아파하면서도, 한편으로는 ‘의미 없는’ 기원을 가진 노래가 사람들에게 깊은 울림을 준다는 역설에 주목합니다.

브리짓 토드는 이 노래에 대한 솔직한 감정을 드러냈습니다. “솔직히 이 노래 싫어요. 너무 가짜 같고, 너무 치즈 같고, 너무 진부해요. 이 노래를 부르는 사람이 진짜 흑인 여성인지, 아니면 영화 ‘딜리버런스(Deliverance)‘에 나오는 글렌 코사(Glen Kosa) 캐릭터인지 모르겠어요.” 그녀는 노래가 너무 ‘직설적’이고 ‘제조된’ 느낌이 강하다고 지적합니다. 이는 인종차별적 AI 콘텐츠에서 흔히 나타나는 현상으로, AI가 생성한 인물이나 인플루언서가 전하고자 하는 메시지가 지나치게 명확하고 과장되어 오히려 불신감을 불러일으킨다는 것입니다. 사람들은 이것이 흑인 정체성의 진정한 표현이라고 믿기 어렵습니다.

감정을 자극하는 AI 콘텐츠의 전략

브리트니와 브리짓은 ‘Phenomenal Black Woman’ 현상을 통해 생성형 AI 콘텐츠의 핵심 전략을 파악합니다. 즉, AI 콘텐츠는 특정 감정을 유발하도록 설계된다는 것입니다. 자부심, 공감, 즐거움과 같은 긍정적인 감정뿐만 아니라 혐오감, 짜증, 분노 같은 부정적인 감정도 마찬가지입니다. 브리짓은 “캐리커처가 11로 증폭된 것 같다”고 표현하며, 이러한 콘텐츠가 의도적으로 극단적인 반응을 끌어내려 한다고 설명합니다.

소셜 미디어 피드를 스크롤하다가 갑작스럽고 강렬한 감정적 반응을 일으키는 영상을 접한다면, 그것이 AI 생성 콘텐츠일 가능성을 의심해 봐야 한다는 것이 브리짓의 조언입니다. 이러한 콘텐츠는 우리가 무의식적으로 느끼는 감정을 자극하여, 그것이 진짜인지 AI가 생성한 것인지 잠시 멈추고 생각해 볼 기회를 주지 않기 때문입니다.

인종적 편견을 악용하는 AI 콘텐츠의 그림자

’손수 만든 가방’ 사기: 감성 팔이 뒤에 숨겨진 진실

브리짓 토드가 발견한 AI 생성 영상 중에는 ‘저수준 사기성 콘텐츠’가 많았습니다. 대표적인 예는 흑인 여성이나 아이들이 직접 만든 수공예품을 보여주는 영상입니다. 예를 들어, 어린 흑인 소년이 손수 만든 가방을 들고 공예 박람회에 갔는데, 백인 아이들이 와서 가방을 바닥에 내팽개치는 장면을 연출합니다. 영상은 소년이 가방을 정성껏 바느질하는 모습과 함께 “내 손수 만든 가방이 못생겼나요? 이 어린 흑인 비저너리를 지지해 주시겠어요?”라는 메시지를 던지며 구매 링크를 제공합니다.

이러한 영상은 바이올라 데이비스(Viola Davis)나 로빈 딕슨(Robin Dixon)과 같은 유명인들까지 댓글을 달 정도로 많은 사람들의 마음을 움직였습니다. 이 콘텐츠를 만드는 사람들은 흑인 커뮤니티가 서로를 지지하고 보호하려는 강한 유대감을 가지고 있다는 것을 정확히 알고 이를 악용하고 있었습니다. 브리트니는 이러한 사기가 “테무(Temu)나 아마존(Amazon)에서 찾을 수 있는 대량 생산된 저렴한 상품”을 팔기 위한 것일 뿐이라고 지적합니다. 즉, 영상 속 이야기는 조작된 것이며, 실제로는 드롭쉬핑(Drop Shipping) 사기에 불과합니다. 진짜 ‘손수 만든’ 제품일지라도 영상에 나오는 사람과는 무관하며, 비윤리적인 환경에서 생산되었을 가능성이 높습니다.

정치적 선전과 잘못된 정보 확산

AI 영상은 정치적 목적이나 잘못된 정보 확산에도 악용됩니다. 브리트니는 뉴욕 시장 선거 당시 테러, 폭력, 빈곤, 범죄 등 부정적인 이미지를 담은 AI 생성 정치 광고를 예로 들었습니다. 이러한 광고는 뉴욕 시민보다는 뉴욕에 살지 않는 사람들을 대상으로, 그들의 기존 편견이나 두려움을 자극하고 특정 정치인에 대한 반감을 조장하려는 목적이 있었습니다.

브리짓은 SNAP(Supplemental Nutrition Assistance Program) 혜택 중단 논의가 한창이던 시기에 나타난 AI 생성 영상들을 언급했습니다. SNAP 혜택이 실제로 중단되기도 전에, 소셜 미디어에는 흑인 여성들이 식료품점에서 EBT 카드(전자 혜택 이체) 사용이 중단되자 과장된 반응을 보이는 영상들이 넘쳐났습니다. 이 영상들은 종종 Sora나 V3 같은 AI 생성 도구의 브랜딩이 명시되어 있었음에도 불구하고 빠르게 확산되었습니다. 이는 ‘복지 여왕(welfare queen)‘과 같은 인종차별적 고정관념을 강화하고, 공공 서비스 수혜자에 대한 부정적인 인식을 심어주려는 의도를 담고 있었습니다. AI는 더 이상 린다 테일러(Linda Taylor)와 같은 실제 ‘복지 사기꾼’이 필요 없이, 이러한 캐리커처를 손쉽게 생성하여 대중의 인식을 조작할 수 있게 된 것입니다.

디지털 블랙페이스와 그 목적

브리트니는 AI가 인종 및 민족적 고정관념을 이용하는 방식에 대해 “디지털 블랙페이스(digital blackface)“라는 용어를 언급하며 질문을 던집니다. AI가 생성한 흑인 여성 영상이 어떤 반응을 유발하려 하며, 게시자에게 어떤 가치가 있을까요?

브리짓은 이것이 강한 감정적 반응을 유발하려는 시도이며, 이는 미국 역사 속 민스트럴 쇼(minstrel show)와 유사하다고 설명합니다. 과장된 인종적 비유와 스테레오타입, 거짓 묘사가 효과적으로 작동한다는 것을 알고 있기 때문입니다. AI를 통해 이러한 콘텐츠를 대규모로 제작할 수 있다는 점이 더욱 위험합니다. 연구자 제레미 카라스코(Jeremy Carrasco)는 AI 생성 콘텐츠 제작자들이 A/B 테스트처럼 다양한 버전을 내놓으며 어떤 콘텐츠가 가장 효과적인지 찾아낸다고 설명합니다.

이러한 콘텐츠의 목적은 다양합니다. 단순히 ‘참여 유도(engagement farming)‘를 통해 플랫폼으로부터 수익을 얻거나, 앞서 언급된 드롭쉬핑 사기를 통해 돈을 버는 것입니다. 심지어 불가능할 정도로 검은 피부를 가진 AI 생성 흑인 여성 영상의 링크가 성인 콘텐츠 사이트로 연결되는 경우도 있습니다. 인종차별적이고 모욕적인 내용을 담은 AI 생성 흑인 여성 영상의 주된 소비층은 다른 흑인 여성이 아님이 분명하며, 이는 결국 유료 성인 콘텐츠를 홍보하는 수단으로 사용될 수 있습니다.

소셜 미디어 플랫폼들은 AI 콘텐츠 사용 시 표기를 권장하지만 의무 사항은 아닙니다. 브리트니는 AI 영상이 너무 터무니없거나 혼란스러워서 혹은 ‘불쾌한 골짜기(uncanny valley)’ 효과 때문에 계속 보게 된다고 말합니다. 하지만 이러한 ‘계속 시청’ 행위 자체가 알고리즘을 훈련시켜 더 많은 유사 콘텐츠를 보게 만들고, 게시자에게는 이득이 될 수 있습니다. 브리짓은 자신이 ‘못생기게 우는 아이’ 사기 영상을 많이 봤더니, 이제 자신의 피드에는 AI 생성 ‘못생기게 우는 사람들’이 가득하다고 농담처럼 말합니다.

AI 시대, 여성에게 전가되는 책임

AI 도입을 둘러싼 성별 격차의 오해

AI는 인종 문제뿐만 아니라 젠더 문제에서도 복잡한 양상을 보입니다. 리즈 위더스푼(Reese Witherspoon)이 여성들에게 AI 사용을 독려하며 여성의 직업이 자동화될 가능성이 높다고 언급해 논란이 되었습니다. CNBC의 설문조사에 따르면 남성의 69%가 AI를 ‘가치 있는 조수이자 협력자’로 보는 반면, 여성은 61%만이 동의했습니다. 여성의 절반은 AI를 의심의 눈초리로 바라보며 직장에서 AI를 사용하는 것이 ‘부정행위’ 같다고 느꼈습니다.

브리짓 토드는 이러한 성별 격차가 ‘여성이 AI를 사용하기에 너무 멍청하거나 두려워한다’는 식으로 단순화되는 것에 반대합니다. 여성들은 AI와 관련된 환경적 우려에 더 민감하며, 특히 소수자 정체성을 가진 사람들이 직장에서 AI를 사용할 때 더 많은 감시를 받게 된다는 현실을 잘 이해하고 있습니다. 백인 남성이 AI를 사용하면 ‘선구자’로 보일 수 있지만, 여성이나 유색인종이 AI를 사용하면 같은 대우를 받지 못할 것이라는 인식이 있다는 것입니다.

또한, 많은 여성들은 남편과 함께 가정에서 많은 감정 노동과 가사 노동을 책임지면서 동시에 직장 생활을 합니다. 이러한 상황에서 AI 학습이라는 또 다른 부담을 지는 것은 사실상 세 가지 풀타임 직업을 수행하는 것과 다름없다는 것입니다.

시스템적 문제의 개인화

브리짓은 리즈 위더스푼과 같은 인플루언서들이 여성들에게 AI 사용을 독려하며 “뒤처질 것”이라고 경고하는 것이, 본질적으로는 시스템적 문제를 개인의 책임으로 돌리는 방식이라고 비판합니다. 여성들이 AI에 대해 회의적인 이유들은 대부분 제도적이고 시스템적인 문제에서 비롯되는데, “AI를 충분히 빨리 배우지 않는 것이 당신의 잘못”이라고 말함으로써 그 책임을 개인에게 전가한다는 것입니다. 이는 AI를 만드는 사람들, AI 기술 인프라가 여성을 제대로 고려하지 않거나 심지어 해를 끼치도록 설계된 문제와는 무관하게, 학습하지 않는 여성 개인의 문제로 치부됩니다.

리즈 위더스푼은 부엌에서 스무디를 만들며 자신의 독서 클럽 이야기를 하는 등 ‘여성 대 여성’, ‘엄마 대 엄마’의 친근한 방식으로 AI 사용을 권유합니다. 하지만 그녀는 사실 ‘헬로 선샤인(Hello Sunshine)‘이라는 미디어 회사를 블랙스톤(Blackstone)에 매각한 사업가입니다. 브리짓은 이러한 친근한 포지셔닝이 그녀의 실제 위치(수백만 달러 규모의 비즈니스를 운영하는 소유주 계층)를 가리고, AI를 자신의 사업 이윤을 높이는 수단으로 보는 관점을 숨긴 채 대중에게 AI를 ‘친구처럼’ 팔아넘기는 수사적 장치라고 지적합니다. 멜 로빈스(Mel Robbins)가 여성들에게 재정 관리를 위해 AI를 사용하라고 권유하는 것도 마찬가지입니다. 그녀 자신은 인간 재정 고문과 변호사를 고용할 가능성이 높지만, 대중에게는 AI 사용을 권장하며 특정 AI 회사로부터 대가를 받을 수 있기 때문입니다.

제도적 보호와 개인의 책임 사이에서

출판계의 AI 스캔들: ‘굿리즈 사건’부터 ‘샤이 걸’까지

AI는 출판계에서도 논란의 중심에 서 있습니다. 2024년 ‘굿리즈 스캔들(Goodreads scandal)‘은 한 작가가 가짜 계정을 만들어 같은 출판사 다른 작가들의 책에 나쁜 리뷰를 남기고 자신의 책에는 좋은 리뷰를 달아주었던 사건입니다. 브리짓은 이 사건이 미아 발라드(Mia Ballard) 작가의 ‘샤이 걸(Shy Girl)’ AI 논란의 전조였다고 봅니다. ‘샤이 걸’은 AI가 작성한 것으로 의심받아 미국 출판이 취소되는 사태를 겪었습니다.

브리짓은 출판사들이 AI 사용에 대한 명확한 정책을 가지고 있지 않아 작가들이 그 영향을 고스란히 흡수해야 하는 불공정한 상황이라고 비판합니다. 할리우드 제작사든 대형 출판사든 기관들은 AI 사용에 따른 비난으로부터 보호받는 반면, 미아 발라드와 같은 개인 작가들은 모든 부정적인 파급 효과를 감당해야 하는 구조가 문제라는 것입니다. 브리짓은 2026년에는 출판 계약서에 AI 사용에 대한 기대치를 명확히 명시해야 할 의무가 있다고 강조합니다.

더 나은 시스템을 향한 길

브리짓은 AI가 일자리를 빼앗고, 뇌를 썩게 하며, 우리를 더 외롭게 만들 수 있다는 우려에 공감하면서도, 기술 도입에도 불구하고 여전히 해결하지 못한 인종차별, 성차별, 여성 혐오와 같은 문제들을 AI가 어떻게 더욱 심화시킬 수 있는지에 주목합니다.

그녀는 이 복잡한 질문에 대한 만족스러운 답은 없지만, 부분적으로는 ‘개인’으로부터 시작해야 한다고 말합니다. 소셜 미디어 피드를 스크롤하다가 강렬한 감정적 반응을 일으키는 콘텐츠를 본다면, 잠시 멈추고 “내가 왜 이런 감정을 느끼는가? 이 상호작용의 반대편에 누가 있고, 누가 나의 이런 감정을 통해 이득을 얻으려 하는가?”를 스스로에게 물어보라고 조언합니다. 인터넷을 대하는 우리의 마음가짐과 인식을 훈련하는 것이 중요하다는 것입니다.

하지만 그녀는 동시에 “우리가 이럴 필요는 없다”고 단호하게 말합니다. 아침에 눈을 뜨자마자 비몽사몽간에 피드를 스크롤하는 사람들에게 항상 그 정도의 비판적 사고와 분석 능력을 기대하는 것은 비합리적이라는 것입니다. 따라서 개인의 노력도 중요하지만, 우리는 훨씬 더 나은 시스템을 가질 자격이 있으며, 그러한 시스템이 구축될 때까지는 스스로를 보호해야 한다고 강조하며 대화를 마무리합니다.

생성형 AI는 분명 혁신적인 기술이지만, 그 이면에는 인종적 편견을 악용하고, 사기를 조장하며, 사회적 불화를 심화시킬 수 있는 어두운 잠재력이 존재합니다. 이러한 문제에 대한 인식을 높이고, 개인적인 경계심을 갖는 동시에, 기술 기업과 사회 전체가 더 윤리적이고 공정한 AI 시스템을 구축하기 위한 노력을 게을리하지 않아야 할 것입니다.


“How Getting Stoned With My Dad Helped Us Heal”New York Times Podcasts 기반 기사 원본 영상 보기

세대 간 상처를 넘어선 부자(父子)의 치유: 줄리안 브레이브노이즈캣이 아버지와 관계를 회복한 특별한 여정

에피소드 요약

이 에피소드는 작가이자 영화 제작자인 줄리안 브레이브노이즈캣(Julian BraveNoiseCat)이 아버지와 함께 대마초를 피우며 관계를 치유한 경험을 다룬다. 줄리안은 어린 시절 아버지가 알코올 중독과 개인적인 문제로 인해 종종 사라지고 양육에 소홀했던 아픔을 겪었다. 하지만 성인이 되어 다큐멘터리 ‘슈가케인(Sugarcane)’ 제작과 책 집필을 위해 아버지와 재회하면서, 그의 아버지가 캐나다 원주민 기숙학교(Indian residential school)에서 겪었던 트라우마와 그로 인한 대물림된 고통을 이해하게 된다. 특히 아버지가 1959년 세인트 조셉 미션(St. Joseph’s Mission) 기숙학교에서 태어나 쓰레기 소각장에서 발견될 뻔했던 충격적인 출생 비밀을 알게 되면서, 아버지의 행동 이면에 있는 깊은 상처를 공감하게 된다. 이 과정에서 밤마다 함께 대마초를 피우며 스크래블(Bong Hit Scrabble) 같은 게임을 하고 이야기를 나누는 시간이 서로의 상처를 치유하고 유대감을 형성하는 중요한 계기가 되었다.

이 이야기는 부모의 이해할 수 없는 행동 뒤에 숨겨진 복잡한 역사적, 개인적 트라우마가 있을 수 있음을 시사한다. 부모 자식 간의 관계 회복은 단순히 문제 해결을 넘어, 상대방의 과거와 고통을 이해하려는 적극적인 노력과 공감을 통해 이루어질 수 있음을 보여준다. 줄리안의 사례처럼, 때로는 기존의 방식에서 벗어난 새로운 소통 방식(이 경우 대마초를 통한 솔직한 대화)이 오랫동안 닫혀 있던 마음의 문을 여는 열쇠가 될 수 있다. 이는 부모와의 관계에서 어려움을 겪는 이들에게, 비록 쉽지 않더라도 서로의 상처를 들여다보고 이해하려는 용기가 치유와 진정한 연결로 이어질 수 있다는 희망적인 메시지를 전달한다.

아버지의 날은 많은 이들에게 미묘한 감정을 불러일으키는 기념일입니다. 함께 바비큐를 굽고 낚시를 가거나, 넥타이 매는 법을 배우는 따뜻한 ‘그리팅 카드’ 속 모습도 있지만, 현실은 늘 드라마틱합니다. 어떤 이들은 이해하기 어렵거나 함께하기 힘든 아버지와 자라며, 또 어떤 이들은 아버지의 부재 속에서 성장합니다. 유명 작가이자 영화 제작자인 줄리안 브레이브노이즈캣(Julian BraveNoiseCat)의 경우가 그랬습니다. 그는 어린 시절, 위대한 원주민 예술가이자 삶보다 더 큰 존재였지만, 마치 바람처럼 사라지곤 했던 아버지 밑에서 자랐습니다.

줄리안은 자신의 아버지가 왜 평범한 아버지처럼 곁에 있어주지 못했는지 이해할 수 없었습니다. 하지만 그가 성장하여 작가이자 영화 제작자가 되고, 그의 다큐멘터리 ‘슈가케인(Sugarcane)‘이 오스카 후보에 오르면서, 그는 아버지와 자신의 관계, 그리고 아버지를 멀어지게 만들었던 고통스러운 역사에 대한 답을 찾기 시작했습니다. 최근 뉴욕타임스 매거진에 실린 그의 에세이 ‘아버지와 함께 대마초를 피우며 회복한 우리(How Getting Stoned with My Dad Helped Us Heal)‘는 이러한 치유의 여정을 담고 있습니다. 마침 줄리안이 곧 아버지가 될 예정이라는 소식을 듣고, 뉴욕타임스 팟캐스트 ‘모던 러브(Modern Love)‘는 그와 심도 깊은 대화를 나누었습니다.

‘대마초 스크래블’로 시작된 치유의 밤

“아버지와 함께 대마초를 피우는 건 어떤 기분인가요? 저는 제 아버지와는 상상조차 할 수 없는 일인데요.” ‘모던 러브’ 진행자 애나 마틴(Anna Martin)의 질문에 줄리안은 웃으며 답했습니다. “저희 아버지는 정말 함께 있으면 즐거운 사람(good hang)이에요. 그는 제 ‘대마초 공급책(pusher man)‘이기도 하고요.”

줄리안과 그의 아버지가 함께 즐기는 가장 독특한 놀이는 바로 ‘대마초 스크래블(Bong Hit Scrabble)‘입니다. 스크래블 게임 도중 단어를 완성할 때마다 봉(bong)으로 대마초를 피우는 방식입니다. 밤이 깊어질수록 단어의 난이도는 낮아지지만, 유머는 더욱 깊어집니다. 아버지는 한번은 사전에 없는 ‘slopify’라는 단어를 만들어내기도 했습니다. 줄리안은 자신의 아버지의 전설에 비하면 자신의 이야기는 보잘것없다며 웃었지만, 이 특별한 밤들은 그에게 깊은 의미를 지녔습니다.

이러한 시간은 단순한 유흥을 넘어 부자 관계를 재건하는 중요한 매개체가 되었습니다. 줄리안에게 아버지는 단순한 부모가 아니라, 그의 인생에서 가장 큰 영향을 미친 ‘메인 캐릭터’였지만, 대부분의 장면에 등장하지 않는 존재였습니다.

‘트릭스터 코요테’ 아버지의 그림자

줄리안의 아버지, 에드 아치 노이즈캣(Ed Archie NoiseCat)은 캐나다 브리티시컬럼비아의 카님 레이크 원주민 보호구역(Canim Lake Indian Reserve)에서 자랐습니다. 그는 심지어 원주민 기숙학교(Indian residential school)에서 태어났습니다. 이 학교들은 원주민 문화를 말살하고 아이들을 기독교와 서구 사회에 동화시키기 위해 설립된 곳이었습니다. 자신의 부모에게 길러지지 못한 세대는 자녀를 양육하는 방법을 알지 못했고, 이것이 바로 에드에게도 해당되는 이야기였습니다. 그는 알코올 중독으로 고통받았고, 줄리안이 여섯 살 무렵 가족을 떠났습니다.

하지만 에드는 단순한 결핍의 존재가 아니었습니다. 그는 ‘유명한 원주민 예술가’라는 모순적인 타이틀을 가질 만큼 독보적인 존재였습니다. 어린 시절 줄리안은 종종 잡지 표지에서 거꾸로 쓴 캉골 모자와 보라색 록스타 선글라스를 낀 아버지를 보았습니다. 그는 여전히 스타일 아이콘이자 엄청난 카리스마(rizz)를 지닌 인물이었습니다.

아버지가 떠난 후, 줄리안은 아버지를 거의 볼 수 없었습니다. 1년에 한두 번, 혹은 아예 보지 못하는 해도 있었습니다. 그는 아버지가 시야에서 사라질 때까지 눈을 떼지 못하며, 언제 다시 볼 수 있을지, 어쩌면 다시는 볼 수 없을지도 모른다는 불안감에 시달렸습니다.

어린 시절의 줄리안에게 아버지는 신화적인 존재였습니다. 12살 때, 아버지는 과속을 하며 랩 음악을 듣는 와중에 “아들, 나는 크레이지 호스 메디신(Crazy Horse medicine)을 가지고 있어. 경찰은 날 볼 수도, 잡을 수도 없어.”라고 말했습니다. 크레이지 호스(Crazy Horse)는 미군에 맞서 싸운 유명한 오글라라 라코타(Oglala Lakota) 전사였지만, 결국 붙잡혀 살해당했습니다. 그럼에도 불구하고, 어린 줄리안은 아버지의 말에 완전히 동의하지는 않았지만, 그 안에 어떤 마법 같은 힘이 있다고 믿었습니다. 아버지는 수없이 죽음의 문턱을 넘나들었고, 법망을 피해 다녔으니 말입니다.

하지만 사춘기를 거치면서 줄리안의 아버지를 향한 시선은 바뀌었습니다. 아버지는 놀라운 재능을 가졌지만, 알코올 중독과 같은 문제들로 스스로를 파괴하는 모습을 보였습니다. 줄리안은 자신만큼은 아버지처럼 되지 않겠다고 다짐했습니다. 특히, 자신도 언젠가 아버지처럼 무책임한 아버지가 될까 봐 두려웠습니다. “이 남자는 제 유전자의 절반을 차지하고 있으니까요.”

잃어버린 과거를 찾아서: 원주민 기숙학교의 비극

줄리안은 28세가 되던 2021년, 10년 만에 아버지와 함께 살기 위해 동부에서 고향으로 돌아왔습니다. 그 해 캐나다 캠루프스(Kamloops)의 원주민 기숙학교 부지에서 215개의 잠재적인 미확인 무덤(unmarked graves)이 발견되면서 국제적인 뉴스가 되었습니다. 줄리안은 이 사건을 계기로 원주민 기숙학교와 아버지의 삶, 그리고 원주민 정체성에 대한 책과 다큐멘터리를 만들기로 결심했습니다. 그는 “주제에 최대한 가까이 다가가야 한다”고 생각했습니다.

줄리안이 아버지와 함께 살며 연구를 진행하는 동안, 그들은 아버지의 출생에 얽힌 충격적인 진실을 알게 되었습니다. 아버지는 성 요셉 선교회(St. Joseph’s Mission)라는 원주민 기숙학교에서 태어났습니다. 줄리안의 할머니는 미혼모였기 때문에 임신 사실을 숨겼고, 아버지는 태어난 직후 쓰레기 소각장에서 발견되었습니다. 당시 지역 신문인 ‘윌리엄스 레이크 트리뷴(Williams Lake Tribune)‘은 이 아기를 ‘베이비 X(Baby X)‘라고 불렀으며, 이러한 비극적인 사건이 발생한 기관의 정책과 관행에 대해 의문을 제기했습니다. 연구를 통해 줄리안과 그의 아버지는 이러한 영아 살해(infanticide) 사건이 한 번이 아니라 여러 차례 발생한 패턴이라는 것을 알게 되었습니다.

원주민 기숙학교의 만연했던 성적 학대와 폭력은 단순히 개인의 비극으로 끝나지 않았습니다. 부모에게 길러지지 못한 아이들은 부모가 되는 방법을 알지 못했고, 폭력의 피해자는 폭력의 가해자가 되는 악순환에 빠졌습니다. 에드 또한 어린 시절 버려지고 학대를 겪으며, 자신이 아버지가 되었을 때 어떻게 행동해야 할지 몰랐던 것입니다. 줄리안은 이 이야기를 통해 가족과 공동체의 상처를 치유하고 새로운 결말을 만들고자 했습니다.

‘코요테’ 메타포로 이해한 아버지

아버지와 함께 살며 낮에는 각자의 예술 활동에 매진했습니다. 아버지는 차고 겸 작업실에서 레드 제플린(Led Zeppelin) 음악을 들으며 조각을 하고, 몇 분마다 봉으로 대마초를 피우며 ‘힙한’ 분위기 속에서 작업했습니다. 반면 줄리안은 노트북 앞에 웅크려 책을 쓰고 다큐멘터리를 기획했습니다.

어느 날 줄리안은 원주민 문화 연구 도중 ‘트릭스터 코요테(Trickster Coyote)‘라는 첫 조상 이야기에 빠져들었습니다. 코요테는 세상을 창조했지만, 동시에 많은 것을 망쳐놓은 존재였습니다. 강에 연어를 채웠지만, 그 연어를 이용해 모든 마을에 장가들려 했고, 수많은 아이를 낳은 후에는 그들을 버렸습니다.

줄리안은 문헌 속 코요테와 작업실에서 조각하는 자신의 아버지를 번갈아 보며 깨달았습니다. “맙소사, 제 조상들은 제 아버지 같은 사람에 대해 이야기하고 있었어요. 제 아버지는 바로 트릭스터 코요테였어요!” 이 깨달음은 줄리안에게 아버지를 이해하는 새로운 언어와 서사를 제공했습니다. 아버지를 ‘창조자이자 파괴자, 생존자이자 무책임한 아버지’로 이해하는 이 비유는 그의 문화로부터 온 것이었기에, 잃어버렸던 문화적 유산을 되찾는 의미도 있었습니다.

코요테 메타포는 아버지에게 ‘빠져나갈 구멍’을 제공하는 동시에, 관계의 복잡한 진실을 더욱 깊이 이해하게 했습니다. 아버지는 늘 책임에서 벗어나려는 ‘코요테 같은’ 면모를 보였고, 줄리안은 아버지가 자신을 ‘전설적인 인물’로 만들도록 교묘하게 속였을지도 모른다고 유머러스하게 말했습니다.

늦은 사과, 그리고 새로운 시작

함께 살면서도 아버지는 줄리안에게 과거의 부재에 대해 사과하지 않았습니다. 줄리안은 아버지가 겪었던 고통스러운 경험들을 생각하면, 자신의 행동을 정당화하기 쉬웠을 것이라고 이해했습니다. 사과는 줄리안의 책이 출간된 후, 아버지가 책을 읽고 나서야 비로소 이루어졌습니다.

아버지는 줄리안에게 자신이 얼마나 큰 고통을 주었는지 비로소 이해했으며, 정말 미안하다고 사과했습니다. 동시에 아들은 자신의 이야기를 예술로 승화시킨 것에 대한 깊은 자부심을 표현했습니다. 줄리안은 자신이 아버지처럼 조각가는 되지 못했지만, 자신만의 방식으로 예술을 창조하는 것이 “모든 아버지가 바라는 것”일지도 모른다고 생각했습니다.

놀랍게도 줄리안은 아버지가 자신을 낳았던 나이인 33세에 아버지가 될 예정이었고, 심지어 아들을 낳게 됩니다. “이것은 마치 마법 같은 코요테 이야기 같지 않나요?” 그는 수천 년 동안 이어져 온 원주민의 이야기에 진실이 담겨 있을지도 모른다고 말했습니다.

줄리안은 항상 아버지가 되기를 원했지만, 동시에 아버지의 실수를 반복할까 봐 두려워했습니다. 하지만 그는 “이번에는 제대로 해낼 기회”라고 생각합니다. “세대마다 지난 세대보다 더 적은 실수를 하고, 궁극적으로는 우리 민족을 학살의 고통에서 되살려내는 것이 우리의 사명입니다.”

그는 아버지가 그랬던 것처럼, 아들에게도 “마법 같은 관계”를 만들어주고 싶다고 말합니다. 자신처럼 춤추는 사람이 되고, 하키 선수가 되고, 예술가가 되기를 바랍니다. 아버지가 아직도 조각을 하고 있으니, 언젠가 아들을 작업실에 데려가 작은 조각칼을 쥐여주고 싶다는 꿈을 꾸고 있습니다.

물론 여전히 두려움은 존재합니다. 아버지처럼 가족 생활을 망치고, 아들에게 곁에 있어주지 못하게 될까 봐 걱정합니다. 하지만 그는 자신의 아버지와의 관계가 한때는 “깨어지고 고통스러운 이야기”였지만, 이제는 “사랑과 회복의 이야기”가 되었다는 것을 상기합니다. 아버지가 어린 자신을 라스베이거스 호텔 방에 혼자 남겨두고 미술 전시회에 갔던 일과 같은 극단적인 사건들도 있었지만, 이제는 “가장 친한 친구가 되어 함께 예술을 만들고, 같은 지붕 아래서 멋진 시간을 보내며, 우리만의 작은 코요테 트릭스터 모험을 아름다운 방식으로 만들어가고 있다”고 말합니다.

“물론 실수를 하겠지만, 아버지의 실수만큼 심각하지는 않겠죠. 하지만 관계가 아무리 복잡해도, 충만한 삶과 충만한 관계를 살아가는 것은 충분히 가능합니다.”

그렇다면 줄리안의 아버지는 어떤 할아버지가 될까요? 줄리안은 확신합니다. “정말 멋진 할아버지가 될 거예요. 그에게는 너무 쉬운 일이죠. 그냥 나타나서 존재하기만 하면 돼요. 버킷햇을 쓰고 조각칼을 든 채로요. 그는 분명 해낼 겁니다. 그의 ‘할아버지 시대(pa era)‘가 될 거예요.”

결론: 치유는 계속된다

줄리안 브레이브노이즈캣은 인터뷰 이후 아버지가 되었습니다. 2024년 4월, 그의 아들 코퍼 알렉산더 마틴 노이즈캣(Copper Alexander Martin Noise Cat)이 태어났습니다. 줄리안은 아버지가 출산실에 함께 있었고, 아마도 그곳에서 가장 긴장한 사람이었을 것이라고 전했습니다. 아버지가 된 후 가장 놀랐던 점은 아들이 “얼마나 많이 오줌을 싸고 똥을 누는지”였다며 웃음 섞인 소감을 덧붙였습니다.

아버지의 날, 우리 모두는 각자의 아버지와 복잡하고도 아름다운 이야기를 가지고 있습니다. 줄리안의 이야기는 세대 간의 고통과 상처가 어떻게 예술과 이해, 그리고 용서를 통해 치유될 수 있는지, 그리고 그 과정이 새로운 세대에게 어떤 희망을 선사하는지를 보여줍니다. 아버지의 그림자를 넘어, 줄리안은 이제 자신만의 빛으로 새로운 아버지가 되어가고 있습니다. 그의 치유와 사랑의 여정은 계속될 것입니다.


“Are you anxious, avoidant, or secure? | The Gray Area”Vox 기반 기사 원본 영상 보기

불안한 관계는 이제 그만! ‘안정형 애착’으로 가는 뇌 과학적 로드맵

에피소드 요약

이 영상은 불안형(anxious), 회피형(avoidant), 안정형(secure)이라는 세 가지 애착 유형을 통해 인간관계의 핵심을 탐구하며, 안정적인 애착이 고통의 근원과 치유의 기반이 된다는 핵심 주장을 펼친다. 특히, 성인 애착 유형이 유년기 경험에 의해 결정된다는 통념에 반론을 제기한다. 아미르 르빈(Amir Levine)은 30년간 사람들을 추적 관찰한 연구 결과를 인용하며, 부모의 행동이 성인 애착 유형에 미치는 영향은 약 3%에 불과하며, 오히려 유년기 초기 친구 관계가 11%로 더 큰 영향을 미친다고 설명한다. 이는 애착 유형이 고정된 범주가 아니라 스펙트럼이며, 우리가 죽을 때까지 애착 방식을 계속해서 가질 수 있고, 심지어 다른 사람들과 다른 애착 유형을 가질 수도 있음을 시사한다. 궁극적으로 우리는 모두 안정적인 방식으로 살아가는 법을 배울 수 있다는 희망적인 메시지를 전달한다.

안정적인 애착은 우리의 뇌가 사회적 안정감, 즉 안전한 기준선(social baseline)을 찾는 생물학적 과정과 밀접하게 연결되어 있다. 이는 거창한 관계가 아닌 일상생활의 작고 사소한 상호작용, 즉 ‘겉보기에는 사소한 일상의 작은 상호작용(seemingly insignificant minor interactions of everyday life)‘을 통해 형성된다. 예를 들어, 메시지에 대한 짧은 응답이나 가벼운 스킨십 같은 미세한 상호작용들이 우리가 타인과 연결되어 있다는 안전감을 제공한다. 이러한 작은 연결고리는 마치 아이가 부모를 확인하며 안전하게 놀듯이, 우리의 뇌가 안전함을 느끼고 삶의 의미와 통제감을 유지하는 데 필수적이다. 따라서 우리는 과거의 부정적인 ‘스크립트’에 갇히기보다, 삶 속에서 안정적인 순간들(‘secure kernels’)을 재조명하고 현재의 관계에서 긍정적인 미세 상호작용을 통해 뇌를 ‘안정 모드(secure mode)‘로 전환하는 연습을 해야 한다. 이는 우리가 가진 ‘이야기’에 갇히지 않고, 더 나은 삶을 위한 새로운 서사를 만들어가는 과정이다.

우리는 관계 속에서 기쁨과 고통을 동시에 경험합니다. 때로는 따뜻한 포옹이나 진심 어린 한마디가 어떤 약물보다 강력한 위로가 되지만, 불안정한 관계는 가장 큰 정서적 고통의 원인이 되기도 합니다. 복스(Vox) 채널의 ‘더 그레이 에어리어(The Gray Area)’ 에피소드에서 정신과 의사이자 신경과학자인 아미르 라빈(Amir Lavine) 박사는 우리의 애착 유형(Attachment Style)이 삶의 고통과 치유의 근간을 이룬다고 강조합니다. 그의 저서 ‘애착(Attach)‘과 ‘안정형(Secure)‘을 통해 우리가 불안정한 관계 패턴에서 벗어나 더 안정된 삶을 살아갈 수 있는 과학적인 길을 제시합니다.

당신의 애착 유형은 무엇인가요? 세 가지 핵심 스타일

라빈 박사는 애착 유형이 우리가 타인과 얼마나 편안하게 친밀감을 느끼는지, 그리고 관계 속에서 잠재적 위험에 얼마나 민감하게 반응하는지를 나타내는 지표라고 설명합니다.

애착 이론의 발견과 성인 애착

애착 이론은 원래 아기들을 대상으로 연구되었지만, 라빈 박사는 우리가 평생 동안 애착 스타일을 유지하며 살아간다고 말합니다. 성인기의 애착 유형은 크게 세 가지로 나눌 수 있습니다.

불안정형(Anxious), 회피형(Avoidant), 안정형(Secure)

  • 불안정형(Anxious): 친밀해지는 것을 좋아하지만, 상대방이 자신에게 얼마나 ‘가용(available)‘한지에 대해 끊임없이 걱정합니다. “왜 내 전화를 받지 않지?”, “그들은 어디로 간 걸까?”와 같은 불안감에 시달리며 관계의 안정성에 의문을 품습니다.
  • 회피형(Avoidant): 관계를 원하지만, 너무 가까워지면 불편함을 느낍니다. “나는 독립적이어야 해”, “스스로 충분해야 해”와 같은 삶의 스크립트를 가지고 있으며, 관계 속에서 일정한 거리를 유지하려 합니다.
  • 안정형(Secure): 친밀감을 사랑하지만, 잠재적인 위험에 대해 지나치게 민감하지 않습니다. 상대방이 혼자만의 시간을 갖거나 독립적인 행동을 해도 이를 위협으로 여기지 않으며, 따뜻하고 사랑스러운 태도를 유지합니다.

드문 유형: 공포-회피형(Fearful-Avoidant)

이 세 가지 외에 불안정형과 회피형의 조합인 ‘공포-회피형(Fearful-Avoidant)‘도 있습니다. 이들은 친밀한 관계를 갈망하면서도 일단 관계에 들어가면 잠재적 위험에 매우 민감하게 반응합니다. “나는 네가 필요하지만 너와 함께할 수는 없어”라는 모순적인 태도를 보이며, 한 손으로는 다가오라고 하고 다른 한 손으로는 밀어내는 경향이 있습니다.

애착 유형은 스펙트럼, 운명이 아니다

라빈 박사는 이러한 애착 유형이 고정된 ‘범주(category)‘가 아니라 ‘스펙트럼(spectrum)’ 위에 존재한다고 강조합니다. 이는 우리가 다른 사람들과 다른 애착 스타일을 가질 수 있으며, 심지어는 시간이 지남에 따라 변화할 수 있다는 희망적인 메시지를 담고 있습니다. 그의 새 책 ‘안정형(Secure)‘은 우리 모두가 ‘안정형 모드(secure mode)‘로 살아가는 법을 배울 수 있다는 아이디어에 기반을 두고 있습니다.

뇌가 갈망하는 ‘사회적 안정’: 안정형 애착의 본질

그렇다면 심리적, 정서적 안정감의 가장 중요한 차원은 무엇일까요? 라빈 박사는 이를 ‘안정성(Stability)‘이라고 말합니다.

심리적 안정은 ‘안정성(Stability)‘에서 온다

인간은 생물학적 종으로서 모든 생물학적 시스템과 마찬가지로 항상 ‘항상성(homeostasis)‘을 추구합니다. 즉, 돌아올 수 있는 안정적인 ‘사회적 기준선(social baseline)‘을 찾습니다. 안정형 애착은 바로 이 안전한 사회적 기준선을 뇌가 찾는 과정과 연결됩니다.

‘안정형 뇌(Secure Brain)‘를 이해하기

라빈 박사는 그의 책에서 ‘안정형 뇌(Secure Brain)‘가 무엇을 필요로 하는지, 그리고 우리의 ‘사회적 뇌(social brain)‘가 싫어하는 것이 무엇인지 깊이 탐구합니다. 사회에서 흔히 받아들여지는 많은 규범들이 사실은 우리의 사회적 뇌와 잘 맞지 않는다는 사실을 발견할 수 있습니다. 이를 이해함으로써 우리는 타고난 애착 스타일이 어떻든 간에 ‘안정형 모드’로 살아가는 법을 배울 수 있습니다.

잘못된 통념: 어린 시절의 영향 과대평가

흔히 우리의 성인 애착 패턴이 어린 시절의 가족 역학에 의해 결정된다고 생각합니다. 하지만 라빈 박사는 이러한 인과 관계에 너무 많은 무게를 두는 것이 위험하다고 지적합니다. 그는 과학자로서 인과 관계를 확립하는 것이 얼마나 어려운지 깨달았고, 많은 연구가 이 ‘어린 시절 결정론’이 사실이 아님을 보여준다고 말합니다.

예를 들어, 30년 동안 사람들을 추적한 연구에 따르면, 부모의 행동이 성인 애착 스타일에 미치는 영향은 매우 미미했습니다. 어머니의 영향은 성인 애착 스타일에 약 3% 기여했고, 어린 시절의 우정은 11% 기여했습니다. 이는 어린 시절의 경험이 전혀 중요하지 않다는 의미가 아니라, 그것이 우리의 현재를 ‘결정짓는’ 유일한 원인이 아니라는 점을 시사합니다. 변화는 현재의 뇌 환경을 바꾸는 것으로부터 시작됩니다.

‘이야기(Script)‘에 갇히지 마라

우리는 자신에 대한 ‘이야기(narrative)’ 또는 ‘스크립트(script)‘를 만들어냅니다. “나는 아무도 믿을 수 없어”, “내가 뭔가 말하면 관계는 깨질 거야”와 같은 스크립트들이죠. 우리의 뇌는 이러한 스크립트를 재확인하는 신호만을 찾고 다른 증거들은 무시하는 경향이 있습니다. 라빈 박사는 ‘안정형 점화 치료(Secure Priming Therapy)‘를 통해 이러한 스크립트에 도전하고, 우리가 무시해왔던 ‘안정형 잠재력(secure potential)‘을 인식하도록 돕습니다. 우리는 여러 스크립트를 동시에 가질 수 있으며, ‘안정형 씨앗(secure kernels)‘은 우리 안에 이미 존재합니다.

고통스러운 ‘사회적 배제’와 ‘초과 포함’의 힘: 사이버볼 효과

우리의 뇌는 사회적 연결에 대한 깊은 욕구를 가지고 있으며, 배제당하는 것에 극도로 민감하게 반응합니다.

사회적 배제는 왜 아픈가? 사이버볼 효과(Cyberball Effect)

라빈 박사는 ‘사이버볼 효과(Cyberball Effect)‘라는 일련의 실험을 통해 사회적 배제가 우리에게 미치는 영향을 설명합니다. 가상의 공놀이 게임에서 참여자가 다른 두 명의 플레이어와 공을 주고받다가 갑자기 자신에게 공이 오지 않게 되는 상황을 상상해 보세요. fMRI 스캔 결과, 이러한 사회적 배제 상황에서는 신체적 고통과 자기 비판과 관련된 뇌 영역이 활성화되었습니다. 심리적으로는 자존감 저하, 통제감 상실, 삶의 의미 감소로 이어졌습니다. 놀랍게도, 금전적 보상을 주거나 배제하는 사람들이 ‘끔찍한’ 사람들이라고 알려줘도 뇌의 반응은 변하지 않았습니다. 즉, 우리의 뇌는 배제되거나 무시당하는 것을 극도로 싫어하며, 이는 이성적인 판단으로 통제하기 어려운 본능적인 반응입니다.

뇌는 사소한 상호작용에 민감하다

우리는 흔히 깊은 관계만을 애착과 연결시키지만, 애착은 본질적으로 ‘안전 레이더(safety radar)‘와 같습니다. 아이들이 장난감을 가지고 놀다가도 주기적으로 부모를 확인하는 것처럼, 우리도 중요한 사람들이 우리 곁에 있다는 것을 확인하고자 합니다. 문자 메시지에 대한 무응답, ‘읽씹’과 같은 사소한 ‘미세 상호작용(micro-interactions)‘이 큰 상처가 되는 이유도 여기에 있습니다. 우리의 뇌는 이러한 작은 신호들을 통해 연결의 끈이 여전히 단단한지 확인하며, 이는 생존과 직결되었던 진화적 과거(무리에서 배제되는 것이 곧 죽음이었던)와 연결되어 있습니다. 뇌는 생사를 가르는 위협과 친구의 답장 없는 문자 메시지를 신경화학적으로 구분하지 않습니다.

‘초과 포함(Hyper-inclusion)‘의 놀라운 이점: CARP 원칙

사이버볼 실험의 반대 버전인 ‘역 사이버볼(Reverse Cyberball)‘에서는 참여자가 계속해서 공을 받는 ‘초과 포함(hyper-inclusion)’ 상황을 만듭니다. 이때 뇌는 ‘천국’에 있는 것처럼 반응하며, 자존감, 통제감, 삶의 의미가 증가하고 염증 감소 및 장수에도 기여하는 엄청난 이점을 보입니다.

라빈 박사는 이러한 초과 포함 효과를 일상생활에서 만들어낼 수 있는 다섯 가지 원칙을 제시합니다. 바로 ‘CARP’입니다.

  • Consistent (일관적인): 예측 가능하고 꾸준한 태도
  • Available (가용한): 필요할 때 항상 곁에 있어 주는 존재
  • Responsive (반응적인): 상대방의 신호에 민감하게 반응
  • Reliable (신뢰할 수 있는): 믿을 수 있고 의지할 수 있는 존재
  • Predictable (예측 가능한): 갑자기 사라지거나 변하지 않는 안정감

이러한 CARP 환경을 조성함으로써 우리는 뇌의 과도한 경계심을 낮추고, 더 안정적인 기준선을 찾도록 도울 수 있습니다. 여기서 ‘예측 가능성’은 지루함을 의미하는 것이 아니라, 상대방이 필요할 때 그 자리에 있을 것이라는 안정감을 제공하는 것을 의미합니다.

관계를 변화시키는 실용적인 전략: ‘안정형 애착’ 연습하기

그렇다면 어떻게 이 ‘안정형 모드’로 살아가는 법을 배울 수 있을까요? 라빈 박사는 몇 가지 실용적인 규칙과 도구를 제안합니다.

‘안정형 참여의 두 가지 규칙(Two Rules of Secure Engagement)’

안정적인 관계는 우리의 감정을 진정시키는 데 도움이 됩니다. 이 원리에 기반하여 라빈 박사는 관계의 갈등을 해결하는 두 가지 규칙을 제시합니다.

  1. 규칙 1: 한 번에 한 사람만 화낼 수 있다 (Only one person is allowed to be upset at a time).

    • 어떤 문제가 발생했을 때, 한 사람이 화가 나면 다른 사람은 그 사람을 진정시키고 기분 좋게 해주기 위해 노력해야 합니다. 이것이 바로 안정형 애착의 핵심입니다. 우리는 본질적으로 사회적인 존재이며, 신뢰하는 사람의 따뜻한 말이나 포옹은 어떤 약물보다 강력한 치유 효과를 가집니다.
  2. 규칙 2: ‘미아 쿨파(Mia Culpa)’ 규칙 (The Mia Culpa rule).

    • 만약 두 사람 모두 화가 났다면, 둘 다 사과해야 합니다. 이는 ‘우리는 서로의 정서적 안녕에 책임이 있다’는 안정적인 관계의 약속을 어겼기 때문입니다. 사람들은 종종 ‘내가 옳다’는 주장에 매달리지만, 애착은 누가 옳고 그른지에 관심이 없습니다. 연결되어 있다고 느낄 때 비로소 모든 것이 옳습니다. 먼저 사과하고 감정적 해결을 찾으면, 누가 옳았는지는 다음날 중요하지 않게 됩니다.

관계 재정립: ‘사랑을 담은 벽 테니스(Wall Tennis with Love)’

불안정형 애착을 가진 사람들에게 라빈 박사는 ‘사랑을 담은 벽 테니스(Wall Tennis with Love)‘라는 도구를 제안합니다. 벽 테니스는 벽이 던지는 대로 돌려주는 것처럼, 상대방이 먼저 상호작용을 시작하면 사랑으로 반응하되, 먼저 과도하게 관계를 시작하지 않는 방식입니다.

예를 들어, 친구가 문자를 보내면 즉시 답장하여 ‘CARP’ 원칙을 지키며 연결의 끈을 유지합니다. 하지만 그 친구가 자주 답장을 하지 않거나 연락이 끊기는 경향이 있다면, 내가 먼저 연락하여 ‘사이버볼 효과’를 유발하지 않습니다. 대신, 나의 애착 시스템을 안정적으로 유지하기 위해 안정형인 다른 사람들과 더 많이 소통합니다. 이는 특정 사람을 ‘제거(purge)‘하는 것이 아니라, 관계에 대한 기대치를 ‘적정화(rightsize)‘하여 나의 애착 시스템이 조용히 유지되도록 돕는 방법입니다.

자기 연민을 넘어 변화로

많은 치료 문화에서 자신의 애착 패턴을 ‘운명’처럼 받아들이고 ‘화해’하는 것을 강조합니다. 하지만 라빈 박사는 이러한 태도가 ‘주체성(agency)‘을 위한 공간을 충분히 남겨두지 않는다고 비판합니다. 그는 ‘수용(acceptance)‘이 변화의 첫걸음이라고 믿습니다. “이것이 나의 생물학적 특성이고, 이것이 내가 가진 동물적 본능이다. 이제 어떻게 하면 이 특성과 함께 더 편리하고 기분 좋은 삶의 방식을 찾을 수 있을까?”라는 질문에서 진정한 변화가 시작됩니다.

결론

아미르 라빈 박사의 통찰은 우리의 애착 유형이 고정된 운명이 아니라, 학습과 실천을 통해 변화할 수 있는 역동적인 스펙트럼임을 보여줍니다. 우리의 뇌는 사회적 안정을 갈망하며, 일상 속의 작고 사소한 상호작용들이 우리의 애착 시스템을 형성하고 변화시키는 데 결정적인 역할을 합니다.

‘CARP’ 원칙을 통해 안정적인 관계 환경을 조성하고, ‘안정형 참여의 두 가지 규칙’과 같은 실용적인 전략을 통해 관계의 갈등을 현명하게 해결하며, 때로는 ‘사랑을 담은 벽 테니스’처럼 관계에 대한 기대치를 재조정하는 연습이 필요합니다.

자신의 애착 지형을 이해하고, 과거의 이야기(스크립트)에 갇히지 않으며, 현재의 환경을 적극적으로 재구성하려는 노력을 통해 우리는 누구나 ‘안정형 모드’로 전환하여 더 건강하고 만족스러운 관계를 만들어갈 수 있습니다. 불안정한 관계의 고통에서 벗어나 진정한 치유와 성장을 경험할 수 있는 길은 우리 안에 있습니다.