The Grittiest Conversations of 2025: AI, Business & Beyond
Timestamps are as accurate as they can be but may be slightly off. We encourage you to listen to the full context.
Podcast Summary
This special year-end episode of the Grit podcast curates the most compelling moments from 2025's top conversations with industry leaders across finance, AI, software, and media. (00:10) Host Joubin Mirzadegan brings together insights from eight powerhouse guests including Carlyle co-founder David Rubenstein, HubSpot CEO Yamini Rangan, Mailchimp co-founder Ben Chestnut, Harvey CEO Winston Weinberg, Handshake co-founder Garrett Lord, Cohere CEO Aidan Gomez, Cloudflare co-founder Michelle Zatlyn, and Snap CEO Evan Spiegel. The episode explores how AI is fundamentally reshaping business models, the accelerating pace of technological change, and the future of human-computer interaction through AR glasses.
- Core themes include navigating political neutrality in business, cutting through AI hype to deliver real customer value, the challenges of scaling in rapidly evolving tech cycles, building trust-first AI products, the critical role of expert human data in training frontier models, the business model disruption facing web content creators, and the potential for AR to transform collaborative computing from single-player to multiplayer experiences.
Speakers
Joubin Mirzadegan (Host)
Partner at Kleiner Perkins and host of the Grit podcast, focusing on conversations with founders, operators, and builders across technology and finance. Known for deep-dive interviews exploring the intersection of business strategy and technological innovation.
David Rubenstein
Co-founder of Carlyle Group, one of the world's largest private equity firms. Has served as chairman of the Kennedy Center for 14 years, the Smithsonian for four years, and currently chairs the National Gallery of Art and Library of Congress Board. Known for his commitment to remaining politically neutral to maintain effectiveness across different administrations.
Yamini Rangan
CEO of HubSpot, the leading customer platform for scaling companies. Has extensive experience in sales and go-to-market operations, previously working at SAP. Known for her focus on solving for the customer and grounding AI initiatives in real customer value rather than hype.
Ben Chestnut
Co-founder of Mailchimp, which he built and recently sold after recognizing the accelerating pace of technological change. Early investor in AI and big data technologies, having used NVIDIA chips for abuse prevention as early as 2008. Currently relieved to no longer be running a tech company during the AI wave.
Winston Weinberg
Co-founder and CEO of Harvey, an AI company focused on the legal industry that has raised $300M. Previously worked on complex legal matters including major mergers like Microsoft-Activision. Known for building enterprise-grade AI solutions that can survive multiple generations of model improvements.
Garrett Lord
Co-founder of Handshake, the career platform connecting students and employers. His company now operates a network of millions of domain experts who help train and validate frontier AI models, representing a shift from generalist to specialist human data requirements in AI development.
Aidan Gomez
Co-founder and CEO of Cohere, and co-author of the foundational "Attention Is All You Need" transformer paper that sparked the current AI revolution. Former Google researcher who helped create the architectural foundation that powers modern large language models including GPT and other frontier systems.
Michelle Zatlyn
Co-founder of Cloudflare, which now serves about 20% of the web. Expert in cybersecurity, bot management, and the changing economics of web content as AI crawlers reshape how information is consumed and monetized. Leading efforts to create new business models for content creators in the AI era.
Evan Spiegel
Co-founder and CEO of Snap, which serves nearly a billion users with just 5,000 employees. Pioneer in augmented reality technology through Spectacles, focusing on making AR glasses useful through collaborative, multiplayer computing experiences rather than single-player interactions.
Key Takeaways
Political Neutrality Enables Greater Impact
David Rubenstein explains his strategic decision to remain completely apolitical, never donating to any political candidates or causes. (00:42) This neutrality allows him to work effectively across different administrations and bring people from both parties together for cultural and educational initiatives. By staying "right down the middle," he can host bipartisan Congressional dinners at the Library of Congress and serve on prestigious boards regardless of which party is in power. This approach demonstrates how avoiding partisan politics can actually amplify your ability to create positive change and build lasting institutions that transcend political cycles.
Ground AI Initiatives in Customer Value, Not Hype
Yamini Rangan emphasizes the critical importance of focusing on real customer outcomes rather than getting caught up in AI excitement. (04:58) She advocates for transforming "neat features" into "necessary features" by proving customer usage and repeat value. HubSpot's approach centers on their "solve for the customer" philosophy, measuring success through actual customer behavior rather than technical capabilities. This grounded approach helps companies navigate the challenging investment decisions around AI while avoiding the trap of building impressive technology that doesn't deliver real business value.
Technology Cycles Are Accelerating Beyond Most Companies' Ability to Adapt
Ben Chestnut provides a sobering perspective on the increasing pace of technological change, explaining why he's "so glad" to no longer be running a tech company. (11:26) He describes how reinvention cycles have compressed from every three years to something much faster, while companies simultaneously become larger and harder to pivot. This acceleration creates an almost impossible situation where businesses must make bigger bets on shorter timelines while having less agility to course-correct. His relief at exiting before the AI wave illustrates the mounting pressure on tech leaders to stay ahead of exponential change.
Build for Ten-Year Model Evolution, Not Current Capabilities
Winston Weinberg's strategy at Harvey focuses on tackling problems that will remain challenging even as AI models dramatically improve. (14:29) Rather than building solutions that current models might soon automate away, Harvey targets the "complete tail end" of model capabilities—work so complex and context-specific that it requires years of model advancement. This forward-thinking approach involves deliberately choosing harder problems today to build a defensible position for the next decade of AI development. The strategy accepts slower initial growth in exchange for building something that survives multiple generations of model releases.
Expert Human Data Is the New Competitive Moat
Garrett Lord explains how the AI industry has evolved from needing generalist human feedback to requiring domain experts for model improvement. (19:09) As models have absorbed most of the internet's content and surpassed generalist humans, the bottleneck has shifted to expert knowledge in fields like medicine, law, and advanced sciences. Companies like Handshake now connect frontier AI labs with PhD-level experts who can identify reasoning errors, provide ground truth answers, and validate complex model outputs. This shift represents a fundamental change in how AI systems improve, moving from broad data collection to highly specialized human expertise.
Statistics & Facts
- Salespeople spend only 25% of their time in front of customers, and even with perfect operations and systems, this only improves to 27% - a mere 2% increase despite optimal conditions. (06:59) This statistic from Yamini Rangan illustrates why AI automation of sales preparation tasks could dramatically improve productivity.
- Handshake experts earn over $125 per hour on average for providing specialized data to train AI models, representing a significant shift from lower-cost generalist labor to high-value domain expertise. (20:51) This demonstrates the premium placed on expert human knowledge in AI development.
- Cloudflare serves about 20% of the web, giving them unique visibility into AI crawler behavior, with some sites reporting 2,500 AI crawler visits per day while seeing reduced human traffic. (33:34) This highlights the scale of AI content consumption and its impact on traditional web business models.
