AI's new bottleneck: taste. eBay's corpse.

THE THROUGHLINE

1. Cross-Podcast Themes (3-5 themes)

AI's Cost Inversion: Implementation Is Cheap, Taste Is the Bottleneck

Paul Erlang, co-founder of FOMO, isn't seeing marginal gains. On 20VC, he argues that AI tools like Cloud Code and Codeax aren't just incremental improvements; they dramatically accelerate engineering output, letting teams build complex products in weeks, not months. His team built a trading app in one month, a pace he attributes directly to AI-enhanced workflows. Watch full episode

On Lenny's Podcast, Andrew Ambrosino, product lead for OpenAI's Codex app, reveals this rapid acceleration means the actual building of software is no longer the expensive part. The new bottleneck is "taste" and "curation"—the skill to sift through a multitude of AI-generated prototypes, identify what's good, and refine it into a superior product. Watch full episode

The AI Evaluation Crisis: Benchmarks Lie, Specialization Wins

An OpenAI model recently disproved the complex Erdos unit distance conjecture internally. On No Priors, OpenAI research scientist Noam Brown argues that traditional AI benchmarks are deceptive because they ignore "test-time compute"—the weeks or even months some models spend "thinking" to unlock their full capabilities. He explains a model's 'capability' is "a function of how much money you put into it" during evaluation, not just its intrinsic design. Watch full episode

On Latent Space, Matei Zaharia of Databricks warns against chasing general frontier LLMs, advocating instead for specialized models and systems that automate specific, complex tasks. He highlights their specialized vision model for document parsing, which is "100x cheaper" and delivers better accuracy than general-purpose models for that task. Watch full episode

Founders vs. The Board: Vision Wins Over Skepticism (and Inertia)

When Walt Disney dreamed up Disneyland, his own board of directors "hated the idea," seeing only a financially risky amusement park. Yet, as detailed on Acquired, Walt funneled his personal resources and even his own intellectual property through a separate company to fund its initial planning and bypass internal skepticism. Watch full episode

Ryan Cohen, known for his work at Chewy and GameStop, pulled no punches on All-In Podcast when discussing his $56 billion bid for eBay. Cohen, declaring “life is too short to do it small,”▶ 27:51 sees eBay as a stagnant e-commerce giant ripe for disruption, with his vision challenging its "professional management team" to overcome corporate inertia and adapt to new market realities.

The Creator Economy's Harsh Math: Costs Soar, Independence Fades

On Huberman Lab, Andrew Huberman and Dr. Paul Eastwick discuss how certain platforms create "kleptocracies" where a tiny fraction of users receive the vast majority of attention. This extreme inequality in markets like dating apps can be paralleled to the creator economy, where the "harsh math" of competition means only a few dominate, leaving most creators struggling to gain traction or maintain independence amidst fierce competition for attention. Watch full episode

On TBPN, John Coogan points out that by 2026, success in this low-bar format will be increasingly rare due to fierce competition, pushing creators towards high-production "shows." Watch full episode

2. Best Of the Week

20VC: Harry Stebbings wants founders to stop throwing money at marketing that vanishes, instead pushing for "immortal assets"—brand investments that generate value for years, like a "football shirt."

Acquired: Walt Disney's origin story wasn't a creative fairytale; it was a brutal business education, notably when he lost control over Oswald the Lucky Rabbit, ingraining an "ironclad commitment to IP ownership." Read more.

All-In Podcast: Ryan Cohen built Chewy into a multi-billion dollar business with an unrelenting dedication to efficiency and a deeply uncomfortable rule for suppliers, admitting his approach included "fiercely aggressive supplier negotiations." Read more.

Dwarkesh Podcast: Dwarkesh Patel questions whether training AIs in simulated environments can generalize to complex, real-world problems, noting that valuable knowledge gained 'in context' during deployment is "ephemeral" and wasted if not consolidated into the model's weights. Read more.

Huberman Lab: Andrew Huberman has "completely reversed his stance on artificial sweeteners," now stating he has “no problem with them whatsoever based on the current data,”▶ 20:11 after learning from Dr. Layne Norton that replacing sugary drinks with sweetened alternatives is a "net positive." Read more.

Latent Space: OpenAI's Mark Chen predicts AGI is "coming soon," fundamentally shifting innovation generation from humans to autonomous models, turning human researchers into "vibe researchers" focused on high-level idea generation. Read more.

Lenny's Podcast: Andrew Ambrosino advises ditching detailed 9-month product roadmaps in AI because model capabilities change too fast, calling any precision beyond a few months "false precision." Read more.

My First Million: Investor Nick Sleep's philosophy, highlighted by Shaan Puri, focuses on maximizing "consumer surplus" by companies like Costco intentionally passing scale savings onto customers, creating "an unbreakable value proposition." Read more.

No Priors: OpenAI research scientist Noam Brown reveals that current AI models, even public ones like GPT-5.5, hold significant untapped "latent capabilities," exemplified by a model internally disproving the complex Erdos unit distance conjecture. Read more.

TBPN: Meta's Ray-Ban smart glasses defied low expectations, selling 2 million units, proving that product-market fit for consumer hardware can hinge more on "a desirable aesthetic" and social integration (like a partnership with Kylie Jenner) than pure tech. Read more.

The Tim Ferriss Show: Anthropologist Wade Davis argues that prohibiting coca leaf access isn't merely drug policy; it's an act of "cultural genocide," severing the plant's intrinsic link to Andean identity and cosmology, where it's been a sacred plant for 8,000 years. Read more.

3. Most Quotable

"If you're using any other form of creatin I think you're wasting your money."

Dr. Layne Norton on Huberman Lab · June 28, 2026. A direct, no-nonsense dismissal of all creatine forms except for monohydrate, challenging popular supplement marketing.

"Life is too short to do it small."

Ryan Cohen on All-In Podcast · June 28, 2026. A bold declaration of ambition, explaining his $56 billion bid for eBay rather than incremental growth.

"AGI is coming soon, right?"

Mark Chen on Latent Space · June 28, 2026. OpenAI's Chief Research Officer drops a casual yet incredibly weighty prediction about the imminent arrival of artificial general intelligence.

Bottom Line: AI is rapidly reshaping development and evaluation, challenging traditional business models and corporate inertia, while founders and builders navigate rising costs and the changing nature of value creation.