English

Based on "The child who learned to disappear is still running your adult relationships | Nicole LePera" from Big Think Watch the original video

The Echoes Within: How Your Childhood Shapes Every Adult Relationship – And How to Heal

Many of us navigate adult relationships and daily life grappling with anxieties, conflicts, or self-doubt that seem to defy logical explanation. We might attribute these struggles to our "personality" or current circumstances, yet the truth, according to Dr. Nicole LePera, also known as The Holistic Psychologist, is often far more profound. Our present-day challenges are frequently echoes of a past where emotional safety was inconsistent, and our truest selves were not fully seen. These are the footprints of our "inner child," a body-based memory formed in our earliest environments, still silently running the show in our adult lives.

Dr. LePera, a New York Times bestselling author and founder of SelfHealer Circle, a virtual membership community dedicated to empowering individuals in their healing journeys, argues that understanding and reparenting this inner child is key to lasting transformation. Through her work, including her new book Reparenting the Inner Child, she illuminates how early experiences, often subtle, shape our nervous systems and dictate our reactions, habits, and even our identities.

The Unseen Wounds: Redefining Trauma's Reach

When we hear the word "trauma," our minds often conjure images of catastrophic events: abuse, natural disasters, severe accidents. Dr. LePera challenges this narrow definition, revealing that trauma is not solely about what happened, but crucially, about the support we had to process those experiences. This expanded understanding highlights how trauma can arise from "small, subtle moments of a lack of emotional regulation, attunement, and repair."

Consider a child whose parents are divorcing. One child might have a parent who shows up, validates their distress, and allows them to cry, grieve, and feel upset, reassuring them it's not their fault. Another child, in the same divorce scenario, might lack a present parent, left to process their grief entirely alone. Identical external event, drastically different internal outcomes. Similarly, a child sharing a painful experience of being bullied at school might be met with a distracted parent scrolling on their phone, dismissively telling them to "just get over it." The unspoken message: "You are alone in your emotions." Repeated often enough, such moments can foster hyperindependence, teaching a child that vulnerability leads only to disappointment.

The impact of trauma, LePera explains, isn't confined to the individual's direct experience. Epigenetics reveals that the impact can traverse generations, with ancestors' traumatic experiences – such as food scarcity, abuse, or war – influencing how our genes express themselves in future generations, including our own. This pervasive reach means that while many of us may have had physically present parents and our basic needs met, we often lacked the emotional attunement necessary for our nervous systems to cope with life's big and small stressors. This makes trauma a unifying, albeit often unacknowledged, aspect of the human experience.

The good news, however, lies in neuroplasticity – the brain's remarkable ability to rewire itself. The survival habits and patterns formed in response to traumatic events can be changed throughout our lives, offering a profound sense of hope and agency.

When Survival Becomes "Personality": Unresolved Trauma in Adulthood

Our present-day struggles, LePera asserts, are rarely about the present day itself. The anxiety, conflict, or self-doubt that keeps us stuck are often strategies formed in childhood, long before we understood emotional safety. What once kept us safe in inconsistent or neglectful environments now masquerades as our "personality."

"We're still repeating the habits and patterns that once kept us safe," LePera observes. "We're doing so and we're calling it personality. We're believing it's just who we are because we haven't been able to create space from who we've had to become to survive those earliest environments."

These deeply ingrained coping mechanisms, while adaptive in childhood, become maladaptive in adulthood, sabotaging our relationships and well-being. For example, if shutting down during conflict offered safety as a child, an adult might continue this pattern even with a partner seeking understanding, closing off avenues for resolution. Or, if appeasing others maintained connections, an adult might habitually push aside their own needs, saying "yes" when they mean "no," quietly accumulating resentment.

Common maladaptive coping skills include:

• Dissociation or Disconnecting: If emotions were shamed or shut down in childhood, disconnecting from one's physical or emotional body becomes a protective barrier. In adulthood, this can lead to partners feeling unheard or misunderstood, as intimacy is sabotaged by the very distance created for protection.
• Overfunctioning or Hypervigilance: If a child had to physically or emotionally manage the household, they might become hypervigilant to others' needs, constantly pleasing and appeasing. This leads to exhaustion, burnout, and resentment, leaving no space for their own needs within relationships.
• Hyperindependence: If relying on others consistently led to hurt or disappointment, a child learns to deal with problems alone, suppressing emotions and never asking for help. In adulthood, this cuts them off from the connections they desperately need.
• People-Pleasing: Deferring to others' wants and needs might have been the easiest path to maintain early connections. But as adults, it disconnects us from ourselves, leading to resentment as we fail to express our authentic needs and opinions.

Decoding Your Past: The Six Archetypes of Childhood Trauma

Dr. LePera identifies six common archetypes of childhood trauma, each rooted in specific relational dynamics and manifesting in distinct adult patterns:

1. The Denied Reality: Growing up in a home where your perspectives or emotions were dismissed ("stop overreacting," "you're too sensitive") teaches you that your feelings are invalid and your internal world cannot be trusted. As an adult, this translates to second-guessing instincts, tolerating unacceptable behavior, and failing to speak up when necessary. Healing involves reconnecting with your instincts and learning to advocate for yourself.

2. The Invisible Child: This arises from having physically present but emotionally absent parents. A child showing off a drawing might receive a distracted "Oh, that's nice" without a glance. This fosters a feeling of invisibility, that what you have to say or do doesn't matter. In adulthood, this manifests as not speaking up in groups or feeling unheard. Healing means seeing and hearing yourself first, validating your own experience, and seeking relationships where your perspective is valued.

3. The Molded Self: Characterized by conditional love tied to performance or achievement. A child learns their worth depends on what they do. This leads to adult patterns of overwork, perfectionism, and an intense fear of criticism. Healing involves embracing imperfections, showing up authentically, and disconnecting your worth from external achievements.

4. The Blurred Boundaries: This archetype involves violations of physical or emotional space. A parent might confide in a child, relying on them for emotional support. This teaches the child that love and connection mean caring for someone else, leading to overextending oneself in relationships and feeling guilty for having personal needs. Healing requires identifying where boundaries are needed, learning to set and honor them, and creating separation by lovingly acknowledging, "That's yours, and this is mine."

5. The Appearance Obsession: In households where physical or family appearance mattered more than emotional connection, a child might receive criticism about their weight or clothing. This ties self-worth to external presentation, leading to adults who are overly focused on how they look, tirelessly working to present a certain image, or pursuing goals based on appearance. Healing involves separating your value from how you look and understanding your inherent worth.

6. The Erratic Regulator: This describes a parent who is unpredictable – calm one minute, then explosive or shut down the next. Slammed doors and silent treatments teach a child that emotions are dangerous, fostering hypervigilance. As an adult, this translates to heightened anxiety, overreactivity, and difficulty navigating one's own emotional responses. Healing involves understanding when your body is "flooded," pausing, and calming your nervous system to respond more intentionally.

Dr. LePera notes that the "Invisible Child" and "Erratic Regulator" archetypes are particularly pervasive because many parents themselves lacked the emotional tools to regulate their own feelings or hold space for another's.

The Inner Child: A Body-Based Memory

The "inner child" is often perceived as an abstract concept, but Dr. LePera clarifies it as a body-based memory – what psychologists call implicit emotional memory. Before logic and language, children experienced the world through sensation and reflex. This body-based learning forms deep "scripts" that run in the background, dictating feelings and actions even when our logical minds know better. This is why insight alone doesn't always translate into changed behavior; our bodies often speak quicker than our logical minds.

The inner child manifests not just in our reactions but also in our adult habits, patterns, and identities. An "overachieving child" might become an adult who feels "driven," not realizing it stems from conditional love. "Independence" might feel like a choice, but it was a protective stance when others couldn't be relied upon. "Sensitivity" might seem like temperament, but it could be born from an unpredictable environment demanding constant vigilance.

We often meet these survival strategies in moments of stress or dysregulation – mindlessly scrolling, eating, or using substances. We may confuse these actions with a lack of willpower, but they are our nervous system's best attempt to find safety or connection when it feels absent.

A key indicator of an inner child reaction, LePera explains, is the size of the reaction. When responses are "really big, overwhelming, all-consuming, all or nothing," it often signals an older, unhealed past experience. In these moments, we experience "emotional flooding" – cortisol courses through our bodies, the amygdala (emotional center) is overactivated, and the prefrontal cortex (logical part) is underactivated. We literally become that reaction, unable to quickly return to a grounded state. For example, an email about an upcoming meeting might trigger a disproportionate spiral of anxiety about getting fired, even if logically, work is going well. This is the inner child bracing for an unpredictable past to repeat.

Understanding human development helps us shift from viewing these reactive moments as "dysfunction" to recognizing them as intelligent adaptations to earlier circumstances. By exploring what we needed in those early moments, we can begin to shift our present experience. Our attachment styles, too, are formed by the inner child's expectations, based on how consistently (or inconsistently) connection was available.

Reparenting: The Path to Lasting Transformation

Reparenting the inner child is the active process of stepping in as the compassionate, nurturing adult you might not have had. It's about showing up in a new way to meet those deeper, underlying unmet needs. This is distinct from coping, which is merely a means to get through a moment, alleviating discomfort without addressing the root cause. Reparenting, conversely, involves rewiring the nervous system to experience moments differently and make new choices.

Consider waiting for a text or email response. Coping might involve an escalating sense of urgency, firing off more messages, or overanalyzing what was last said, convinced the other person is upset. Reparenting, however, means noticing the urgency, relaxing tensed muscles, slowing your breath, and pausing before reacting. This teaches your body a new experience: distance doesn't necessarily mean disconnection or abandonment.

LePera highlights a crucial point: many of the habits formed out of childhood protection are paradoxically rewarded in modern society. The overachiever exhausting themselves for validation is seen as "driven." The person who shuts down in conflict is labeled "easy" or "low-maintenance." Reparenting challenges these societal norms by prioritizing genuine well-being over external validation.

The journey of reparenting, LePera emphasizes, begins "right here, right now." It's less about dissecting every detail of the past and more about understanding "what is happening" in the present moment. The foundational practice is daily conscious check-ins: setting an alarm or pairing the practice with a daily routine (like drinking coffee). In these moments, you pause to notice your body – how you're breathing, muscle tension, ease – and observe your thoughts. Practicing consciousness outside of reactive moments builds a bridge to those times when you need to show up differently, reminding your inner child that circumstances have changed and new resources are available.

It's important to dispel common misconceptions: Reparenting isn't about blaming parents; it's acknowledging the impact of the past. It doesn't keep us stuck in childhood; it updates old programs for a desired future. And it's not immediate; it's a gradual process requiring consistent new choices, reflecting the years it took for old patterns to form.

Embracing Wholeness: Practical Steps and Profound Shifts

To begin reconnecting with your inner child, LePera suggests a simple yet powerful exercise: find an old photograph of yourself as a child. Observing your size, vulnerability, and facial expression naturally evokes compassion and empathy, making it harder to dismiss that "little being" or label its feelings as "too dramatic."

If a photograph isn't available, try accessing your childhood through your senses. Close your eyes and recall a childhood home or a favorite room. Tune into the sights, sounds, and even smells. Notice your own body language in that memory. Then, ask yourself: "What did I need?" Was it attention, safety, protection? This practice cultivates empathy for that part of you that still resides within.

Daily practices for reparenting include:

• Grounding: Look around your current environment and identify three to five neutral or comforting objects. This reminds your nervous system of present-moment safety.
• Slowing Down: Consciously slow your breathing and movements. This activates the parasympathetic nervous system, sending signals of safety throughout your body.
• Acknowledge, Don't Shame: When you notice an intense reaction, acknowledge it as an "old reaction" without negating your feelings. Understanding, rather than shaming, fosters a compassionate state of awareness.
• Conscious Check-ins: Regularly assess your stress levels throughout the day by noticing muscle tension, breath rate, and heart rate. In moments of rising stress, pause and actively calm your body (e.g., walk, slow breath). This builds your capacity to handle stress.

The profound shifts that emerge from this reparenting journey are transformative. You gain increased capacity to navigate life's stressors, developing resilience to remain grounded and intentional regardless of external circumstances. You expand your awareness, embracing all parts of yourself, including the inner child, which is not just a repository of trauma but also a source of creativity, joy, and playfulness.

Emotionally and relationally, you become calmer and more grounded. You engage with others authentically, communicating your needs directly. You shed the old roles of caretaker or overachiever, focusing instead on genuine connection. The worry of losing people shifts to an assessment of whether you feel safe and authentic in their presence.

For those whose parents are still living, a natural byproduct of this journey can be a desire to share these insights with them. However, LePera wisely cautions that while we may desire validation and compassion from our parents, not all will be able to provide it, as they too are humans shaped by their own childhoods. The primary focus remains on our own healing and empowerment.

Ultimately, the goal of reparenting is not to separate ourselves from uncomfortable parts of our past, but to wholeheartedly embrace all that we are. By understanding the body's role in coloring our present through stress and reactivity, and by psychologically and emotionally making space for our past, present, and future selves, we embark on a journey of profound self-discovery and lasting transformation, empowering us to live more authentically and connectedly.

Based on "How He Built a 2-Year Moat Nobody Can Bet Against | Corgi, Nico Laqua & Emily Yuan" from EO Watch the original video

The Unconventional Path to an Unbreakable Moat: How Corgi Rebuilt Insurance from the Ground Up

In an era where startups often seek "capital-light" and "easy-to-scale" solutions, the story of Corgi, an AI insurance company, stands as a bold counter-narrative. Led by co-founders Nico Laqua and Emily Yuan, Corgi has not only achieved several hundred million in annual recurring revenue (ARR) in record time, becoming one of the fastest-growing B2B companies on the planet, but it has done so by deliberately choosing the hardest possible path: rebuilding a highly regulated, deeply entrenched industry from scratch. Their strategy has created a "two-year moat" so formidable that, in Nico’s words, "nobody can bet against."

The Philosophy of the Hard Thing

Nico Laqua, Corgi's CEO, developed his entrepreneurial philosophy early. Reflecting on his high school jobs, he realized he wanted to do "something big and important with my life," and starting a company seemed the best way to make a difference. This drive led him to question conventional startup wisdom. "Often times a young startup founder try to make products that you yourself would use," Nico observes, "which is normally pretty good advice, but the problem is that problems that they try to solve are often not as big as they could be."

Many founders, he believes, scale down big ideas to make them seem more achievable. Corgi did the opposite. "I think that people should do the opposite and think of the most ambitious version," Nico asserts. This isn't just about ego; it’s a strategic choice. Ambitious missions attract "really smart people" and "more likely to fund those sort of ideas." More importantly, "success actually changes the world in like an interesting way."

For Nico and Emily, a "good idea" is synonymous with a "difficult and hard" one. It's about doing something unique, something that makes you the category leader, not just the 50th app or the 500th restaurant. "In order to be category defining, you can't just do something very simple and very easy and very capital light," Nico explains. "You need to do something very ambitious and difficult and hard, kind of the craziest, the highest impact, the highest leverage one of whatever their idea is."

The Broken World of Insurance

Nico's personal experience solidified his conviction that insurance was ripe for disruption. When seeking insurance for his previous company, he encountered a labyrinth of inefficiency, exorbitant costs, and abysmal customer service. "The insurance policy was $60,000," he recalls, "and that's a lot of money to buy anything. I was making less than that in my salary."

The process was agonizingly slow, taking "several weeks to get the policy." Brokers were unresponsive, and the entire experience felt like a "scam." How could an industry representing 12% of the GDP – twice the size of the software market – operate with such glacial pace and opaque practices?

The answer, Nico realized, lay in its history and structure. Most major insurance providers were founded 40 years ago or more, fostering complacency and a resistance to change. Decades of regulation created huge barriers to entry, allowing incumbents to worsen product quality without fear of competition. Those working in the industry for 20 years, while experienced, often possessed "a solidified mindset of like this is how things are supposed to work." They might plug new tech into old infrastructure, but they rarely questioned the fundamental paradigm. Corgi saw this as a massive advantage: a chance to come in with "a brand new slate" and rebuild from an AI-native perspective.

The Pivotal Turn: From Broker to Carrier

Corgi didn't start with the intention of becoming a full-stack insurance carrier. Their initial approach, applying to Y Combinator already licensed as an insurance brokerage, was to embed with contract management companies and simply resell existing policies. "It was working pretty well," Nico admits. They sold "tens of thousands of dollars of premium," and revenue growth was solid.

However, a deeper dive into the insurance "stack" revealed a critical flaw: the problem wasn't just distribution; it was the underlying product itself. Dealing with traditional insurance carriers was a nightmare. "Making phone calls to them for every single policy," "sending faxes to them back and forth" – these were the realities of working with "100-20 billion dollar companies." Nico's search for a better carrier proved futile; none existed.

This realization led to a controversial and defining decision. They had to control the product, and the only way to do that was to become an insurance carrier themselves. "We decided to shut down what we were doing, which was working pretty well," Nico recounts. "That was not an obvious decision. That was a very controversial decision." They risked their YC standing, going from a "well-regarded" company to one "not doing well at all" and even missing demo day. But the vision was clear: "It's our vision not to build on top of, not to modernize, not to fix what's already there. We're building a new type of financial institution where we actually become these highly regulated financial entities and rebuild them from the ground up using AI."

Forging the Moat: The Gauntlet of Regulation and Capital

The decision to become an insurance carrier plunged Corgi into a multi-year, capital-intensive ordeal. "Our company almost ended on many, many occasions throughout this," Nico reveals. There were periods where they were "very default dead." It took "tens of millions of dollars and several years" to navigate the complex regulatory landscape, culminating in an $80 million pre-revenue raise. They didn't even make a pitch deck or engage in competitive fundraising; instead, they focused on making themselves "a hard company to bet against."

This wasn't just about money; it was about unwavering commitment. Emily Yuan, Corgi's co-founder and COO, highlights the dedication: "Nico actually lives in the office and a lot of our like team members is like live close to the office as well." This intense focus, born from a deep care for the problem and a desire to win, resonated with early investors. "They could see that we deeply cared about the problem. We wanted to be winners and we wanted to win and that the the end kind of destination would lead to the world being a better place."

Emily, whose background includes founding a non-profit, Paper Bridges, and running Stanford's entrepreneurship clubs, complements Nico's visionary drive with her exceptional execution skills. "Nico is very good at identifying good opportunities," she says, "and I think that actually complements my skill set. I'm very good at helping make sure that we can figure out how to like get all these different things done." Her ability to "throw like a very complicated thing at me and I'll go and figure out how to like get it done" was crucial for tackling the regulatory maze. She demystifies regulation: "It's not like some mysterious thing... you just make sure you follow the directions."

The "superpower that young people have is that they have a lot of time and a lot of commitments," Emily notes. "That time and that energy is worth a lot." This dedication was paramount for navigating the "very long process" of regulatory licensing, which took "almost 2 years" for their first product.

The Payoff: An Unassailable Advantage

Once Corgi secured its initial set of licenses, the inflection point arrived. "Our revenue started growing very quickly. We started looking a lot more healthier," Nico states. The strategy of doing the hard thing, of becoming the infrastructure rather than just reselling, paid off handsomely.

The result is a product "fundamentally better than what's available." For technology startups, Corgi offers an unparalleled insurance experience. "Unless someone else goes and makes an entire new carrier for startups, like there's just aren't that many other options."

This is Corgi's moat: the immense time, capital, and regulatory hurdles they overcame. "It's kind of hard for someone else to make like a Corgi 2.0," Nico explains, "because we had to spend so much time and so much money like getting the infrastructure set up. Most companies just can't do that."

The Corgi story is a testament to the power of audacious ambition and relentless execution. By choosing to confront the most difficult challenges in a legacy industry, Nico Laqua and Emily Yuan have not only built a rapidly growing company but have also established an enduring competitive advantage that few can hope to replicate. Their journey proves that sometimes, the only way to truly change the world – and build an unassailable business – is to do the things nobody else dares to touch.

Based on "673. What Is Money? | Freakonomics Radio" from Freakonomics Radio Network Watch the original video

The Economy Sings: Unpacking Adam Smith Through a Modern Oratorio

When Stephen Dubner, host of Freakonomics Radio, stumbled upon news of a new oratorio titled The Wealth of Nations, his "AI feeds you something a little too spot-on" alarm bells went off. Having recently produced a three-part series on Adam Smith, the 18th-century Scotsman widely regarded as the father of modern economics, and another on Handel's Messiah, the idea of a musical mashup of these two seemingly disparate worlds felt uncannily precise.

This wasn't just any oratorio; it was a world premiere by the New York Philharmonic, conducted by superstar Gustavo Dudamel. The composer was David Lang, a name that, while perhaps unfamiliar to Dubner at first, quickly revealed itself to be a significant force in contemporary classical music. Lang, a Pulitzer and Grammy winner who teaches composition at Yale, was embarking on a project that would transform Smith's dense, foundational text of capitalism into a vibrant, emotional sonic experience.

The Genesis of an Unlikely Oratorio

Lang's journey into the world of Adam Smith began, as many artistic endeavors do, with a commission. Having successfully "rewritten" Beethoven's opera Fidelio for the New York Philharmonic, stripping it down to its core prison narrative, Lang seized the opportunity to pitch another ambitious idea: setting The Wealth of Nations to music. He admits with a chuckle, "I wasn't going to read the book unless I had a gig to read the book."

Smith's 1776 tome, written in 18th-century English, is notoriously long and challenging. Lang's initial approach was to find a quirky thread, a "joke," to connect it to Messiah – specifically, sheep. "Sheep play a prominent role in the Messiah," he explains, and being from Scotland, Smith naturally used many sheep examples in his work, like the "woolen coat." While the sheep-based jokes eventually fell by the wayside, this initial search for thematic resonance proved crucial.

As he delved deeper, Lang's focus shifted from the obvious imagery of factories and the division of labor. He found himself drawn to a more profound insight: "the idea that trade connects us and that money itself doesn't really have any value, but money exists as a kind of token that goes from person to person as we are connected through trade." For Lang, money isn't an abstract concept; it represents the labor we invest, becoming a powerful symbol of human connection. This realization formed the emotional core of his oratorio.

David Lang: The Composer and His Craft

Lang's unique approach to composition is as intriguing as his choice of subject matter. He describes himself as "not a very good keyboard player," eschewing the traditional method of composing at the piano. Instead, he sings ideas to himself and then meticulously writes them down using a "Stone Age" software program called Encore, which he cherishes precisely because it doesn't auto-correct his "mistakes." "In my writing, I actually embrace all the mistakes," he states, highlighting a philosophy that values organic expression over polished perfection.

His music, exemplified by pieces like "Just" (lyrics drawn from the Hebrew Bible's Song of Songs), often mesmerizes with its haunting beauty. Lang's path to becoming a celebrated composer was not straightforward. He initially studied chemistry at Stanford, intending to go to medical school, much to his doctor father's delight and subsequent dismay. He recalls a poignant moment after a performance with the Cleveland Orchestra when his mother, instead of offering the long-awaited approval, tearfully suggested, "There's still time to go to medical school." It was a Pulitzer Prize, he jokes, that finally got his parents "off one's back."

Lang's journey also saw him embrace vocal music later in his career, a "huge part of something that I'd been missing." He explains the difference: "If I sing something myself that I know is going to be sung by someone else, I get to feel it. And somehow for me that makes it a lot more powerful."

Even his titling convention—all lowercase, no capital letters—is a deliberate choice. It's a "hopeless affectation" started in graduate school to relieve the pressure of living up to the "great composers from the past." Lowercase titles, he felt, allowed him to create without the oppressive weight of tradition, freeing him to simply "write the music."

Money, Art, and "Enough"

Lang's personal relationship with money informs much of his work. He admits, "I'm not that interested in money, to be honest... I'm interested in having enough." This concept of "enough" is deeply personal, shaped by his upbringing in a family "without money" and his mother's saying, "enough is as good as a feast." He aligns with Mr. Micawber's definition from David Copperfield: "If you have 20 pounds of annual expense, and at the end of the year, you have 20 pounds and 1 pence, you are a rich person. And if you have 19 pounds and 19 shillings or whatever the calculation is, 1 penny less, you're impoverished." Living as a freelance artist, he's cultivated an attitude "comfortable with having less."

This personal philosophy extends to his broader views on economics. He sees money as a "social construct," a "social lubricant" that, despite its potential for inhumanity, connects people in ways that physical goods or violence cannot.

Reimagining Suffering: The Little Match Girl Passion

Before The Wealth of Nations, Lang earned his Pulitzer for "The Little Match Girl Passion," a choral piece that brilliantly recontextualizes suffering. A self-professed Bach lover but "not Christian," Lang sought to create a "passion" he could believe in. He drew inspiration from Bach's St. Matthew Passion, recognizing its power in inviting listeners to reflect on Jesus's suffering to become better people.

Lang's radical innovation was to replace Jesus with Hans Christian Andersen's tragic figure, the little match girl, who freezes to death on a cold street. He intercut her story with the crowd's responses from Bach's work. "So, I took Jesus out and I put the little match girl in," he explains. What he initially feared might be "blasphemous" and lead to "bricks through my windows" instead resonated deeply with audiences, earning him the prestigious award. It demonstrated his ability to tap into universal emotions of empathy and the human condition.

Democratizing Sound: Beyond the Ivory Tower

Lang is not comfortable with the label "classical music," preferring simply "music." He finds "contemporary" equally problematic, implying a fleeting relevance. His aim is to broaden the audience for his work, a "democratizing instinct" that has been a hallmark of his career.

In the late 1980s, he co-founded the Bang on a Can music festival, an ambitious "12-hour orgy of contemporary music" designed to expand who listens to experimental new works. This ethos of inclusion and generosity, rather than competition, remains central to his approach.

This democratizing impulse took an unusual turn during a trip to London, when Lang, not a sports fan, impulsively bought a ticket to an Arsenal football match. Amidst 50,000 to 60,000 fans singing "incredibly lewd songs," he observed a profound cooperation through music. "Everyone was cooperating through music," he realized, noting the absence of auditions, political litmus tests, or social stratification. The only requirement was belief in the team's victory.

This experience inspired his piece "crowd out," written for a thousand community members. He drew lyrics from auto-completed internet searches for "When I am in a crowd," filtering out the negative and commercial. The resulting text, a mix of anxieties ("I start to panic," "I feel so alone") and affirmations ("I feel more confident and calm," "I can fully submerge myself"), reflects the complex experience of individuality within a collective. The piece was designed to be "easy enough so that ordinary community members could do it, but hard enough so that they would have to rehearse a few times... They would meet their neighbors. They would end up learning how to depend on each other. And I really felt like that was the democracy-building part of this piece."

The Wealth of Nations: A Moral Compass in Music

Bringing this democratizing instinct to The Wealth of Nations meant making the "emotional weight of international trade" palpable. "Everybody deals with money and everyone has a totally messed up relationship with how money changes hands and how it lives in their lives. It's an emotional issue, right?" Lang argues.

One of his favorite passages from Smith, the "woolen coat," vividly illustrates the interconnectedness of global labor. Lang describes it: "Imagine the poorest laborer in the wool coat that that laborer wears. The sheep had to be sheared and the shears were smelted, the ore was smelted from, you know, places and the dye came on ships. And imagine who made the rope for those ships and who made the sails." He even sings a snippet, admitting his "terrible singer" status, to convey the raw power of the text: "The woolen coat which covers the laborer as coarse and rough as it may appear is the produce of the joint labor of a great multitude of workers."

Lang's oratorio isn't a mere recitation of Smith; it's a conversation. He weaves in other influential voices like Ralph Waldo Emerson, Edith Wharton, and, significantly, Frederick Douglass and Eugene V. Debs. Douglass's essay on wealth, linking wealth inequality to enslavement, and Debs's powerful courtroom speech as a conscientious objector to WWI, serve as crucial counterpoints. While Lang is a "pretty moderate political person" and removed specific socialist advocating from Debs's text, he was drawn to the "powerful angry but ultimately very optimistic statement about where our country can go and how we should live with each other." He seeks to "call out hypocrisy where you see it" and explore "how a virtuous person build a moral structure for commerce."

Crucially, Lang also adds his own voice. Smith's examples, even when discussing the "poorest person," assume everyone has a "woolen coat." Lang noted, "there actually are people who don't have a coat. Those people don't show up in this book." So, he wrote a movement called "Enough," a simple, poignant reflection: "If I could have a piece of bread, when I need a piece of bread, it would be enough. If I could have a coat to wear when I need a coat to wear, it would be enough." This personal text underscores the oratorio's focus on fundamental human needs and justice.

From Page to Performance: The Composer's Agony and Ecstasy

The transition from a composer's mind to a live performance is a unique blend of anxiety and anticipation. Lang describes the period just before rehearsals as "shockingly most empty," a "Schrödinger's cat" state where the piece "existed and also didn't exist." He's "just sitting around nervous," oscillating between pride and fear of "Titanic errors."

His coping mechanism is philosophical: he believes every piece will be played "a thousand times." This perspective liberates him from the pressure of any single performance, allowing for future revisions and evolution. "I don't want to ever think that the piece becomes set in stone," he explains.

The first vocal rehearsal with the New York Philharmonic Chorus was a revelation. Lang, a small man in black with chunky glasses and a shaved head, tried to stay out of the way, observing as 49 singers, led by Malcolm J. Merriweather, brought his music to life. Dubner, present at the rehearsal, found the music "totally arresting," noting Lang's use of "plain song, like a slightly modernized Gregorian chant." This was evident in the chorus's repeated singing of one of Smith's most famous lines: "It is not from the benevolence of the butcher, the brewer, or the baker that we expect our dinner... but from their regard to their own interest." Lang's music, Dubner observed, could be "lacerating and comforting in the same moment. Hypnotic, and then cathartic."

A few days later, with Gustavo Dudamel at the helm and the full orchestra assembled, the piece truly began to coalesce. Dudamel, a "rock star" conductor, worked with intense focus, occasionally consulting Lang on nuances like a "forte subito" (suddenly loud) versus a gradual crescendo. The collaboration, despite minimal prior communication, was seamless, a testament to the professionalism of all involved. As mezzo-soprano Fleur Barron, one of the soloists, noted, the initial surprise of setting The Wealth of Nations to music quickly gave way to its compelling artistic reality.

The Unfinished Symphony of Humanity

The Wealth of Nations oratorio is more than a musical performance; it's a profound commentary on humanity's enduring struggle with economics, morality, and justice. Lang's work implicitly asks: How far have we come since Adam Smith's 18th-century observations, or Eugene Debs's early 20th-century critiques?

Debs's stark declaration, "money is still so much more important than the flesh and blood of childhood... gold is God today," resonates with chilling prescience. Lang acknowledges, with characteristic understatement, "We could do better." He adds, "I think that that's the whole point, right? If we had actually paid attention to all the lessons we could have learned up to now... all the music would be about love and dancing. And the fact that we still have other things to write about means we have a little farther to go."

David Lang's The Wealth of Nations challenges audiences to reconsider a foundational text through an emotional, communal, and deeply human lens. It's a testament to the power of art to not just entertain, but to provoke thought, foster empathy, and illuminate the hidden connections that bind us all, much like the countless laborers contributing to a simple woolen coat. It reminds us that the quest for a more just and equitable world is an ongoing symphony, with many movements yet to be composed.

Based on "Baseten CEO Tuhin Srivastava on Custom Models, and Building the Inference Cloud" from No Priors: AI, Machine Learning, Tech, & Startups Watch the original video

The Unseen Engine of AI: Why Inference is the Next Frontier

The artificial intelligence revolution is in full swing, with new models and applications emerging at a dizzying pace. While much of the spotlight falls on the creation of these powerful AI systems, the real workhorse, the engine driving their everyday utility, often operates in the shadows: AI inference. This is the process where a trained AI model takes new data and makes a prediction or decision – from generating text and images to powering medical diagnoses and customer support.

At the forefront of building this critical infrastructure is Baseten, an "AI inference cloud" that has witnessed an astonishing 30x growth over the past year. Tuhin Srivastava, founder and CEO of Baseten, offers a compelling perspective on the future of AI, arguing that inference is not just a burgeoning market but "the last market" – an ever-expanding domain of intelligence consumption.

The AI Explosion: From Labs to Enterprises

Srivastava emphasizes that the past 24 months have been pivotal, marked by a universal realization that "you can put AI everywhere." This pervasive adoption is fueled by a rich ecosystem of both closed-source and increasingly capable open-source models. A critical turning point, he notes, is that open-source models have "crossed some sort of chasm" in baseline capability, making advanced techniques like reinforcement learning (RL) and post-training for specialized models mainstream. This empowers customers to "own their inference more and more," driving demand for platforms like Baseten.

The Application Layer's Existential Fight

A central debate in the AI world revolves around whether independent application companies can truly thrive, or if frontier model labs will eventually consume the entire stack. Srivastava firmly believes in the enduring power of the application layer, citing several reasons:

1. User Signal as a Moat: Companies gather unique user signal – data about how their customers interact with their products and workflows – that is invaluable for specialization. When this signal is encoded in a model or, more importantly, in specific workflows, it creates a defensible competitive advantage.
2. Specialized Workflows: He offers the example of Abridge, an ambient scribe company used by physicians in US hospitals. Abridge's deep integration into clinician workflows, capturing edits and subsequent actions within Electronic Medical Records (EMRs), generates a unique "reward signal." Frontier model companies struggle to replicate this because they lack access to this proprietary user data. Over time, companies with this access can post-train models on their specific reward signals, developing "long horizon agentic models" that are highly specialized and efficient.

This dynamic, where unique user signal drives specialized model development, ensures that a vibrant application layer will continue to exist.

Enterprise Adoption: The Sleeping Giant

While AI-native application companies like Abridge, Decagon, and Open Evidence are currently Baseten's primary customers, Srivastava sees a massive wave of enterprise adoption on the horizon. "If you look by inference count, it'd be 99% the full," he states, highlighting that the majority of the market is yet to come online.

These AI-native companies, serving enterprises in mass, act as a crucial feedback loop for Baseten. They translate enterprise requirements – data retention, deployment locations, GPU types, latency tolerances, and model transparency – allowing Baseten to build an inference cloud that is inherently suited for future enterprise needs. "By serving companies like Abridge and Open Evidence, we're probably pretty well suited to go serve the healthcare system," Srivastava explains.

Custom Models: The Heart of Differentiated AI

A surprising insight from Baseten's operations is the prevalence of custom models. "It is all custom... 95% plus," Srivastava reveals, referring to the tokens served on their dedicated inference business. This means customers aren't just running vanilla open-source weights; they're making significant modifications.

Beyond Vanilla: The Power of Post-Training

Customers specialize models not only for quality but also for performance. They might compile models in different ways or fine-tune them with their own data for specific use cases. This deep customization is where the real value is unlocked, allowing companies to develop AI capabilities that are "better, faster, and cheaper" than generic models.

To further support this trend, Baseten acquired PAS, a company specializing in post-training models. This strategic move brought in research expertise, allowing Baseten to accelerate the market's adoption of post-trained resources and get closer to customers earlier in their AI journey. Srivastava emphasizes the symbiotic relationship: "how linked inference and post-training are... even when you think about stuff like quantization... how training... affects how you need to quantize for inference and how paired these problems are." The ideal loop involves inference creating data, which then feeds into evaluations, post-training on a reward function, and back into improved inference.

When to Specialize

For companies considering custom models, Srivastava offers clear advice: "Go prove to yourself with the best-in-class model that you have something worth optimizing." He likens it to the early startup adage, "no GPUs pre-product market fit," now updated to "no post-training pre-product market fit." Only after establishing a clear user signal and demonstrating customer value should companies invest in the complex process of custom model development.

Navigating the Compute Minefield: Capacity, Geopolitics, and the Multi-Chip Future

The rapid growth of AI has created an unprecedented demand for specialized computing power, primarily GPUs. Srivastava minces no words about the severity of the situation: "I don't think people realize how bad it really is."

The Unrelenting Supply Crunch

Baseten operates large clusters at "uncomfortably high utilization" (mid-90s), spanning 18 different clouds and 90 clusters globally. Their technology, designed to create a unified runtime fabric across these diverse environments, has become critical for simply acquiring compute. They can integrate a new provider in a different country into their fabric in less than half a day, offering enormous flexibility.

Beyond the sheer scarcity, Srivastava points to a "grifty" element among some new suppliers who lack experience in running data centers or understanding inference SLAs (Service Level Agreements). This further constrains the reliable supply of compute, making operational excellence in managing distributed infrastructure a key differentiator.

The Cost of Capital and Long-Term Bets

The supply crunch is also reshaping financial dynamics. Acquiring significant capacity today often requires 3-5 year contracts with 20-30% of the Total Contract Value (TCV) prepaid. This necessitates having sufficient demand to justify the investment and a low cost of capital, potentially influencing decisions like going public sooner.

The Geopolitics of AI Models

The discussion also touched on the origin of open-source models, specifically the rise of high-performing Chinese models like DeepSeek. While acknowledging concerns about security or embedded biases, Srivastava states that Baseten has "never seen any real evidence" of such issues, except for very early models quickly identified by the community.

His pragmatic view centers on innovation and cost. If a model like DeepSeek can run at 20% of the cost of other frontier models with comparable or better performance, denying access to it would be a "massive loss" for innovation. He argues that the US must develop its own strong open-source models, seeing it as both "necessary" and "inevitable."

The Multi-Chip Horizon

While Nvidia's H100 remains a dominant force (a "great chip" that is "four and a half years old" with prices still rising), Srivastava believes in "diversification everywhere." He anticipates a "multi-chip world" with inference-specific chips, acknowledging efforts like Groq's LPUs. However, he stresses Nvidia's current insurmountable advantage due to its supply chain, CUDA ecosystem, and developer mindshare. "The ability... to me like one of the most important things the infrastructure company in this moment is how fast you can move and you can move fastest with Nvidia today," he asserts.

The Operational Crucible: Scaling and Staying Ahead

Scaling 30x in a year presents unique operational challenges. Baseten continuously invests in optimizing its runtime for various workloads.

Workload Evolution

Key areas of investment include:

• Diffusion Transformers: Handling complex generative AI tasks.
• Coding Agents and Sandboxes: Providing secure environments for AI agents that interact with code.
• Speculation Techniques: Accelerating inference.
• KV Cacheware Routing: Optimizing memory usage for large models.
• Prefill and Decode Disentanglement: Treating pre-computation and token generation as separate problems for efficiency.

Beyond runtime, Baseten focuses on creating a seamless loop between inference and post-training, partnering with companies like BrainTrust for evaluation and building APIs for continuous learning.

Surprises at Scale

Operating at such immense scale reveals unexpected "edge cases." Srivastava recounts a kernel panic caused by a Fluent Bit worker generating too many logs on a single node – a systems-level problem that only manifests under extreme load. He notes that even LLM runtimes are "pretty immature," and Baseten is uncovering limitations that will drive the next generation of primitives for scale, security, and performance.

The Capacity Obsession

What keeps Tuhin Srivastava up at night? "Capacity," he states unequivocally. The insatiable demand for compute, coupled with its scarcity, means the mandate is always to "go bigger, go faster." He believes there simply isn't enough compute to unlock the full potential of LLMs in the next 5-10 years without significant new inventions.

Building a High-Performance Culture

Scaling a company 30x isn't just about technology; it's about people. Srivastava reflects on Baseten's journey from a flat, engineer-centric structure to embracing leadership.

From Flat to Focused Leadership

"You just need leaders," he recalls being told. This realization led to building a trusted leadership team. His philosophy centers on:

• Giving Whole Problems: Empowering leaders to own complete challenges rather than micromanaging.
• Clear Optimization: Defining what the company truly optimizes for beyond generic traits like "smart and hardworking." For Baseten, it's about "first principled work," kindness, collaboration, and a "very low ego" culture, eschewing a "hero culture." This clear rubric helps identify who fits and who doesn't.

The Operations Imperative

Srivastava emphasizes the distinct "operations culture" required for an infrastructure company. He recounts a meeting where senior AWS executives' pagers went off multiple times, illustrating the constant readiness required. For Baseten, inference simply "can't go down." This creates a culture where everyone, including leadership, is deeply attuned to operational alerts – a P0 (critical alert) is understood by a seven-year-old child.

Jevons Paradox and the Infinite Demand for Intelligence

A fascinating economic concept, Jevons Paradox, posits that increasing the efficiency with which a resource is used can actually lead to an increase in its consumption, rather than a decrease. Srivastava sees this playing out directly in AI inference.

Cheaper Intelligence, More Consumption

As the cost of inference goes down, developers embed "a hell of a lot more intelligence" into their applications. This manifests in agents running for longer, performing more complex tasks, and ultimately leading to "better user experience, more dollars, more revenue." The demand for intelligence, it seems, is elastic and almost infinite.

The "Last Market"

Srivastava's conviction is clear: "Inference going down just begets more [inference]... it is truly like I think we're kind of in a world that is... the last market, right? Like even if there's AGI, all that's left is inference." He sees no ceiling to the demand for intelligence; customers continuously seek "better answers" and richer experiences, driving an ever-increasing need for AI inference.

The journey of Baseten, spearheaded by Tuhin Srivastava, paints a vivid picture of the AI landscape: a market characterized by explosive growth, strategic challenges in compute and talent, and an insatiable demand for intelligence that promises to reshape industries for decades to come. The unsung hero, AI inference, is indeed taking center stage.

Based on "In a Small Iowa Town, a Solution to a National Crisis | 'The Opinions' Podcast" from New York Times Podcasts Watch the original video

The Riverdale Revelation: How a Small Iowa Town Forged a Lifeline Against a National Flood Crisis

The news cycles churn with alarming regularity: flash floods, once rare occurrences, are now a grim fixture of American life. Fueled by a changing climate, these sudden deluges overwhelm communities, destroy homes, and claim lives. Yet, even as the threat intensifies, the federal government, under recent administrations, has scaled back crucial support systems designed to help communities cope. Programs are cut, vital infrastructure is dismantled, and local officials are left feeling increasingly isolated in the face of an escalating national crisis.

But what if the answer to this overwhelming challenge isn't found in Washington D.C., but in an unassuming town of just 550 people in Iowa? What if a local innovation, born of necessity and ingenuity, offers a blueprint for the entire nation? This is the story of Riverdale, Iowa, and its part-time mayor, Anthony Hedleston, who discovered that saving lives and livelihoods doesn't require billions in federal spending, but rather a blend of smart technology, local expertise, and an unwavering commitment to community.

The Gathering Storm in Riverdale

Riverdale, Iowa, is a picturesque town nestled at the confluence of two significant waterways: the mighty Mississippi River and the more volatile Duck Creek. While the Mississippi typically offers days, sometimes weeks, of warning before a flood, Duck Creek is a "flashier stream," capable of rising eight feet or more in mere hours. This geographical vulnerability places Riverdale residents at constant risk, a reality Mayor Anthony Hedleston understands intimately.

Hedleston is an unusual mayor. Beyond his civic duties, he works as a civil engineer for the Army Corps of Engineers and possesses a self-professed "weather nerd" streak, complete with a personal weather station at his home. He's also a man of distinctive style, sporting long sideburns, a curled mustache, and patriotic glasses emblazoned with the American flag. More importantly, he's a mayor who takes his responsibility seriously, especially when it comes to the safety of his constituents.

For years, Riverdale, like many other small towns, relied on a federal gauge on Duck Creek. Housed in a small brick structure, this sensor measured the creek's height and posted the data online, providing local officials with critical information for evacuation decisions. But last year, Hedleston logged on to check the creek levels, only to find a chilling message: "This gauge has been discontinued." Further inquiry led him to an old acquaintance, Gary Johnson, a federal employee who had managed the gauges. Johnson, it turned out, had taken a "deferred resignation" offered by the Trump administration, a program that allowed tens of thousands of federal employees to leave their posts while still receiving pay for a few months. The federal gauge was gone, physically ripped out, and replacing it would cost Riverdale over $100,000 – an astronomical sum for a town of its size. "That is a big ask of residents of our community," Hedleston noted, acutely aware of the financial strain such an expense would place on his small community.

Iowa's Ingenious Answer: The Flood Center

Just as Riverdale found itself "up Duck Creek without a paddle," Mayor Hedleston discovered a lifeline – and a testament to Iowa's proactive approach to flood management. The Iowa Flood Center, an initiative born from a catastrophic flood in 2008, had developed an incredible tool. Larry Weber, the center's director, proudly demonstrated their innovation: a compact gauge, "the size of a big shoe box," with a sensor cylinder sticking out the bottom and antennas on top.

This Iowa-made gauge is a fraction of the size and cost of its federal counterpart. While the federal gauge would have cost Riverdale over $100,000 to replace, the Iowa Flood Center's version costs a mere $7,500. Local officials can easily attach it to a bridge, and its internal sensor measures water height, posting the information online. Crucially, it integrates with a predictive model, allowing officials to forecast how a rainstorm is likely to affect their waterways. While it doesn't feed into the national weather forecasting system like the federal gauges, it performs the essential local function for a fraction of the price.

When Mayor Hedleston realized his federal gauge was gone, he wasted no time. He purchased one of these smaller, smarter gauges from the Iowa Flood Center, bolted it to a bridge over Duck Creek, and just in time, as fate would have it.

A Night of Decision

The true test of Riverdale's new system came on a harrowing summer night. The rain had started that afternoon, and Mayor Hedleston, hoping for a "nice relaxing calm evening," had settled in with a brandy old-fashioned. But his personal weather station, an unerring sentinel, soon began to blare, signaling an alarmingly high rainfall rate. Flash flood warnings echoed through the air.

With his new Iowa gauge in place, Hedleston consulted its predictive model. The forecast was stark: Duck Creek was likely to rise eight feet sometime in the middle of the night. Eight feet. The mayor, driven by a deep sense of responsibility, grabbed a tape measure and headed to the creek. Standing at the flood wall, he meticulously counted the steps leading down to the water, each eight inches high. "Okay, one, two, three, four, right?" he recounted, doing the math. "Whoa, that's right. About eight feet."

The prediction matched the height of the flood wall. This meant potentially inundating homes, including that of an older resident who lived right next to the creek and used an oxygen tank – someone who would need extra time and care to evacuate. "Oh, that's a big question I need to ask myself there," Hedleston remembered thinking. "We're talking about potentially we need to evacuate people."

This was the first time Hedleston was relying solely on the new gauge during a potentially disastrous flood. He needed to be absolutely certain of its accuracy before making the life-altering decision to evacuate. He called the Iowa Flood Center, but it was Friday night, and they weren't a 24/7 operation. "There's nobody there to answer the phone," he recalled. The weight of the moment was immense. "It's just, it's so heavy," he confessed, describing moments of fury and the "heavy, heavy weight to put on somebody to have to make that decision." He felt he "earned my whole paycheck that night."

At 10:30 p.m., Hedleston went to the county's emergency operations center, contacting the Red Cross to arrange a shelter and drafting an emergency alert message. But the uncertainty lingered. The gauge predicted water levels right at the top of the flood wall, but what was the margin of error? What if it was underestimating?

In a stroke of modern-day luck, Mayor Hedleston turned to LinkedIn, tracking down Felipe, an engineer at the Flood Center. Felipe, recognizing the urgency, agreed to run a different model for a more precise answer. As Hedleston waited, he returned to the creek. The scene was terrifying. "Duck Creek is a raging inferno," he described. "It is a scary, high-velocity flood, and if it were to go over that flood wall and hit those houses, we're not talking about, you know, a soaked first floor and a ruined basement. We're talking about knocking things off foundations and people's lives being ruined and people's lives being lost."

Finally, at 11:30 p.m., Felipe called back. The revised prediction: "It's only going to be four feet. I did the math. The model's overpredicting a little bit." Four feet. Far below the level of the flood wall. "I can breathe. I can sleep tonight. It's going to be okay," Hedleston remembered. He finally made it home around 1:00 a.m., took off his rain jacket, and walked over to his desk. His brandy old-fashioned, made hours earlier, was still there, the ice long melted. He drank it, a well-earned toast to a night of harrowing decisions and a community saved.

Beyond Riverdale: The Broader Impact

Mayor Hedleston knows how lucky Riverdale is to have the Iowa Flood Center. It’s not just the affordable, accurate gauges and their predictive models, but also the human expertise available in a crisis. Following that intense night, the Flood Center added its leaders' cell phone numbers to their website, ensuring Iowans can reach them anytime.

The benefits of Iowa's system extend far beyond preventing unnecessary evacuations. Local officials across the state laud the Flood Center and its online mapping system, which they use constantly. Rick Wolfkull, an emergency coordinator from Buchanan County, shared an anecdote: "We just had a proposal from a developer come through and they were proposing putting a 30-home subdivision right in the floodway. So the next meeting I went to, I said, 'You realize this area has been underwater every time a frog farts?' And that is literally what I said. So I am not about putting 30 families in harm's way." This system empowers officials to make informed decisions about building, preparing for extreme weather, and even guiding rescue efforts. During a major flood in 2024, Larry Weber provided real-time water velocity data for flooded streets, advising officials on the safest routes for boat rescues.

This stands in stark contrast to federal resources. The National Weather Service and FEMA provide flood maps, but they are notoriously limited. The interface is often clunky and outdated, with some maps not updated since 2010. Furthermore, Congress mandated federal flood maps for major rivers and coasts in the 1960s, but not for small tributaries or streams like Duck Creek. And while FEMA maps account for hurricanes, their models often fail to capture the intense, localized bursts of rain that cause flash flooding – a problem only set to worsen with climate change.

A Blueprint for the Nation

Larry Weber believes other states could easily replicate Iowa's success. He estimates the average state could implement a similar system for about $2 million to start, with an annual maintenance cost of approximately half a million dollars. This is a minuscule fraction of the tens of billions of dollars most states spend overall each year. Yet, for now, Iowa remains the only state with such a comprehensive, locally-focused system.

One expert suggested that communities typically don't act until after the "next catastrophic flood hits." However, as federal support continues to dwindle, as funding evaporates, and as storms intensify, the urgent need for local solutions becomes undeniable. The federal government's retreat, ironically, might just be the motivator communities need to embrace proactive climate adaptation.

Riverdale, Iowa, is more than just a small town with a mayor who enjoys a brandy old-fashioned. It represents a powerful testament to local innovation in the face of national neglect. Iowa has developed a plan, a practical, affordable blueprint for other states to follow. As flash floods become an increasingly common and destructive reality, this blueprint offers not just hope, but a vital strategy for protecting communities and saving lives across the country. The time for other states to adopt it is now.

Based on "How OpenAI Is Rewriting Its Future" from Hard Fork Watch the original video

OpenAI's High-Stakes Reset: Uncoupling, Lawsuits, and the Future of AI

OpenAI, the company that ignited the generative AI revolution, finds itself at a pivotal juncture. Far from a smooth ascent, the past few weeks have seen the company navigate a dramatic strategic reset, a bruising legal battle with co-founder Elon Musk, and a rapid redefinition of its market position. These shifts, unfolding against the backdrop of an insatiable demand for AI and a burgeoning medical revolution, paint a complex picture of a company rewriting its future in real-time.

From conscious uncoupling with its biggest investor to facing down accusations of "looting a charity," OpenAI's journey is a testament to the high-stakes, high-drama world of frontier AI development.

Uncoupling and Realigning: OpenAI's Evolving Partnerships

For years, Microsoft has been OpenAI's anchor, its largest investor with a stake valued at an astonishing $135 billion. Their partnership, while instrumental to OpenAI's rise, has also been characterized by underlying tensions. This week, the two tech giants announced a significant rewrite of their agreement, signaling what some describe as a "conscious uncoupling."

The core of the new deal allows OpenAI greater "promiscuity" in its partnerships. Previously, OpenAI was largely restricted to serving its models on Microsoft's Azure infrastructure. However, with cloud service providers like Microsoft increasingly maxed out on capacity due to soaring AI demand, this exclusive arrangement became a bottleneck for OpenAI's revenue growth. The revised terms now grant OpenAI the freedom to work with other cloud providers, such as Amazon Web Services (AWS) and Google Cloud Platform.

"OpenAI just had this real challenge, which was that until this week, they were really only allowed to serve their models on Microsoft's infrastructure," notes Casey Newton, co-host of Hard Fork. "One thing we talk about on the show a lot is just that a lot of the big cloud service providers, their infrastructure is just maxed out." This newfound flexibility is a massive win for OpenAI, enabling it to tap into a broader customer base tied to other cloud ecosystems.

Perhaps the most talked-about change in the Microsoft deal, however, is the removal of the "AGI clause." The original agreement stipulated that once OpenAI achieved Artificial General Intelligence (AGI)—a poorly defined but widely anticipated milestone where AI can perform any intellectual task a human can—Microsoft would cease receiving certain revenue share payments. Under the new agreement, OpenAI will continue sharing revenue with Microsoft until 2030, regardless of any AGI benchmarks. As New York Times columnist Kevin Roose quipped, "I for one will be sad to see it go because I think it was sort of the funniest clause in the entire AI world, right? It was like basically like, well, if we ever get to a point where OpenAI says the magic word, then the entire world changes and now they're not allowed to say the magic word anymore." The AGI clause, once a contractual curiosity, now exists only "as evaluated by Vibes."

This strategic loosening appears beneficial for both parties. Microsoft no longer has to share revenue with OpenAI from their joint ventures, while OpenAI gains the crucial ability to diversify its infrastructure and reach.

Finding Bedrock with Amazon

OpenAI wasted no time in leveraging its newfound freedom. Hot on the heels of the Microsoft announcement, it expanded its deal with Amazon, announcing that its models would be available through AWS's Bedrock AI platform. This includes making Codeex, OpenAI's coding model, accessible on Bedrock. Amazon also reportedly plans to invest $50 billion in OpenAI and develop customized models for its consumer-facing applications.

This move is particularly interesting given Amazon's existing close ties with Anthropic, a rival frontier model developer. OpenAI's entry into the Amazon ecosystem signals a direct challenge to Anthropic's favored status, highlighting the fierce competition for cloud provider allegiance. The CEO of AWS was reportedly "talking a really big game about this deal," with Newton likening it to a "the boy's mine situation" in the AI world.

These mega-deals underscore a critical reality in the AI boom: even the biggest companies lack the resources to serve the overwhelming demand for AI. The narrative has shifted from skepticism about AI's ability to generate demand to concerns about whether infrastructure can keep pace. "Even the biggest companies do not have the resources that they need to serve that demand," Newton explains, pointing to a profound shift in how the AI boom is perceived.

Stargate's Shifting Sands and Financial Realities

The ambitious "Stargate" project, OpenAI's proposed $500 billion infrastructure initiative, is also undergoing a significant recalibration. Recent reports indicate that OpenAI has halted planned data centers in the UK and Norway, declined to expand its flagship site in Abilene, Texas, and seen key figures tied to Stargate depart for rival Meta. Instead of building out all its own facilities, OpenAI is increasingly shifting to leasing capacity from third parties.

This pivot suggests a dose of reality intruding on initial, perhaps overly optimistic, projections. Roose recalls the early pronouncements: "Well, we're going to spend $1 trillion that we don't have to build 40 quadrillion data centers. And at the time, people said, 'That kind of seems like a lot. Can you guys actually live up to that?' And they said, 'Uh, yeah, just watch us.' Uh, well, guess what? They could. And now they're changing course."

While this might seem like a retreat, it's more likely a strategic adjustment. OpenAI, reportedly eyeing an initial public offering (IPO), needs to get its financial house in order. Moving massive infrastructure costs off its balance sheet by leasing capacity is a savvy move to present a more attractive financial profile to investors.

This tension between grand ambition and financial pragmatism is a recurring theme within big AI companies. Roose identifies two camps: the "indefinite optimists" who believe demand for AI is infinite and all investment will be repaid many times over, and the "number crunchers" who demand clear financial projections and revenue plans. OpenAI's shift on Stargate, coupled with reports of missed internal user and revenue targets for 2025, suggests the number crunchers are gaining influence as the company prepares for public scrutiny.

The Ad-Supported Future of AI?

OpenAI is also recalibrating its subscription strategy. Internal projections suggest a massive growth in its $8-a-month "ChatGPT Go" ad-supported subscription, potentially reaching 112 million users this year, while its $20-a-month "Plus" subscriptions are projected to fall significantly. This mirrors Netflix's strategy of offering cheaper, ad-supported tiers.

This indicates a market splitting into two distinct segments:

• Casual users: Those who use AI chatbots for "souped-up Google queries" or email assistance a few times a day are likely to opt for cheaper or free, ad-supported tiers.
• Professional users: For whom AI is mission-critical, willing to pay significantly more than $20 a month for access to the latest models and higher rate limits.

"All of the companies now are sort of doing this kind of experimentation with how much can we charge the professional users without losing them to a rival company, and how cheap can we make the kind of lower-end subscriptions or the free tiers so that people who are more casual users won't be tempted to go use Google instead," Roose explains.

Amidst these strategic shifts, a peculiar anecdote emerged: the need for "safety guardrails to prevent goblins from taking over our coding app." This seemingly absurd detail highlights the unexpected challenges in AI development, blending serious safety concerns with a touch of the surreal.

Elon Musk's Legal Gauntlet: A Battle for OpenAI's Soul

Adding another layer of drama is the long-awaited trial of Elon Musk's lawsuit against OpenAI, which began this week in Oakland. Musk, a co-founder who provided initial funding, left OpenAI in a power struggle and later launched his own AI company. He is now suing OpenAI, alleging "unjust enrichment" and "breach of charitable trust."

Musk claims that OpenAI was "only ever supposed to be a nonprofit" and that its transformation into a multi-billion-dollar for-profit entity constitutes "looting a charity." He testified that "It is not okay to steal a charity," warning that if OpenAI is allowed to "get away with this, it will give license to looting every charity in America."

OpenAI's lawyers, however, paint a different picture, accusing Musk of being "bitter that the company has succeeded without him." Lead counsel William Savit stated, "We are here because Musk didn't get his way at OpenAI. My clients had the nerve to go on and succeed without him. Mr. Musk did not like that." They also point to emails from 2017-2018 where Musk himself discussed turning OpenAI into a for-profit entity, and even wanted to fold it into Tesla, undermining his current "principled stand."

The stakes are considerable. Musk seeks to redirect over $150 billion from OpenAI's for-profit arm back to its nonprofit foundation, which could create significant "headaches and roadblocks" for OpenAI's ambitious projects like Stargate. While legal experts suggest the case is unusual and Musk's chances of a full victory are slim, the ongoing litigation is a major distraction.

For journalists and observers, however, the trial has been "very valuable," revealing early emails and communications that illuminate the "inner workings" and "early dynamics" of OpenAI. It underscores the degree to which these projects are often "fueled by grudges," with personal animus and rivalries playing a significant role alongside grand visions of the future. "A shocking percentage of the AI industry is just people who decided they didn't want to work with Sam Altman and who now have their own companies," Newton observes.

AI in Medicine: A Rapid Revolution

Beyond the corporate drama, AI is quietly—and rapidly—transforming another critical sector: medicine. Dr. Adam Rodman, an internal medicine physician at Beth Israel Deaconess Medical Center and assistant professor at Harvard Medical School, describes AI as "probably the fastest adopted medical technology of all time."

Doctors, including "normies" not on the bleeding edge, are routinely integrating AI into their weekly practice. The most common tools include:

• AI Scribes: Voice-to-text algorithms that listen to patient-doctor conversations and generate a first draft of medical notes. These have become a "commodity" in less than two years, beloved by doctors for saving time and by patients for allowing more direct interaction with their physicians.
• Decision Support Software (e.g., Open Evidence): This free tool, adopted by close to half of U.S. doctors, uses retrieval-augmented generation to search medical literature and identify high-quality sources for clinical queries. Older doctors use it as a "souped-up way to search the literature," while younger doctors increasingly ask it for "second opinions" or "what could be going on?" questions.

While there's pushback against directly integrating AI with patient health records due to privacy concerns, this is already happening in areas like billing. EHR vendors like Epic are also building native AI features, such as AI-generated message drafts for patients.

Patients and AI: A New Competency for Doctors

Patients are also embracing AI. Approximately a third of Americans now report turning to AI for healthcare information. This means doctors are increasingly encountering "someone else in the exam room"—ChatGPT. This necessitates a "new competency" for physicians: discussing AI with patients.

Dr. Rodman categorizes AI uses for patients with a "green light, yellow light, red light" system:

• Green Light (Safe Uses): General health questions (e.g., "Can you help me come up with a diabetic diet?"), preparing for clinic visits by formulating questions, and analyzing wearable data (which a doctor might not have time to review).
• Yellow Light (Proceed with Caution): Exploring new symptoms or seeking second opinions from chatbots. This can be helpful for preparation, but patients must understand it's "not a replacement for a doctor" and is merely a "first step to talking to a human being."
• Red Light (Avoid): Asking for medical management decisions (e.g., "Is this the right chemotherapy option for my cancer?"). These decisions are too nuanced, and models can be "sickopantic," convincing users they are right even when wrong, potentially leading to dangerous advice.

A concerning trend among "health maxers" in places like San Francisco involves uploading extensive lab data from premium health services into chatbots like Claude or ChatGPT, treating them as "first-line medical professionals." Dr. Rodman cautions against this, warning it can drive people into a "cyber worry hole" without improving health outcomes. "The evidence is not there yet that the sort of large routine testing, functional medicine, and putting it into an LLM does anything to improve health outcomes," he states.

Regarding specific integrations, while projects like ChatGPT Health (analyzing Apple Watch data) and ChatGPT for Clinicians exist, Dr. Rodman remains cautious. Privacy, messy health record data (full of errors and copied information), and the current limitations of LLMs mean "there's like no advantage to just dumping everything in an LLM."

The discussion also touched on a trial in Utah where an AI agent autonomously renews prescriptions for nearly 200 routine drugs, with some human review. While Roose presented a "devil's advocate" argument, suggesting such automation could bypass "rent-seeking behavior" of doctors requiring office visits for refills, Dr. Rodman emphasized the importance of human review due to potential dangerous side effects that only a doctor might catch during a brief interaction.

Despite these caveats, AI in medicine is yielding genuinely promising results. The Mayo Clinic recently announced "Red Mod," an AI system capable of identifying subtle changes in routine CT scans up to three years before a pancreatic cancer diagnosis—a significant leap in early detection.

The Future of AI and OpenAI: A Rising Tide?

Despite the corporate drama, legal battles, and strategic pivots, the outlook for OpenAI and the broader AI industry remains largely optimistic. While some predict OpenAI's demise or failure to IPO, a more nuanced view suggests the company is making smart, necessary adjustments to solidify its position.

The idea that there will be "only one winner" in the AI race is a "fallacy," argues Roose. Instead, a "rising tide of AI adoption will sort of lift all boats." Companies with top-tier models are likely to thrive together, suggesting that OpenAI's current "weird enterprise" is navigating a complex but ultimately expanding landscape.

As OpenAI continues to rewrite its future, its journey offers a fascinating glimpse into the rapid evolution of technology, the intense competition for dominance, and the profound societal impact of artificial intelligence—from the boardroom to the doctor's office.