The Slow Hunch

My conversation with Meltem Demirors, founder of Crucible.

Meltem has spent much of her career thinking about the relationship between scarce physical resources and the digital economy. She started in oil and gas, was early to Bitcoin, and is currently obsessed with the energy-to-compute value chain that underpins AI. 

That thread now runs through Crucible, where she invests across energy, compute, materials, and the markets that scale them. Compute and energy are becoming core commodities of the AI era, with entirely new markets forming around them. 

As agents become economic actors, those markets may look very different from the ones we know today.

Chapters: 

• 00:00 AI is a commodities business
• 01:50 Meltem’s slow hunch: energy, compute, and markets
• 06:25 From Bitcoin mining to the energy-to-compute thesis
• 10:03 The founding thesis behind Crucible
• 15:03 Why commodities and capital markets matter
• 20:18 The economics of the AI infrastructure boom
• 26:06 Debt, neo-clouds, and the coming compute shakeout
• 30:58 Compute and energy as the commodities of the AI era
• 34:43 Agents as traders and economic actors
• 37:47 What an agent-native economy might look like
• 40:34 Why markets may be the best way to distribute AI’s gains

My conversation with Jesse Walden, founder and managing partner of Variant. 

Jesse has spent most of his career pulling on one idea: technology is most interesting when it gives people more agency.

That idea first showed up for him in the early internet, through piracy, BitTorrent, and the independent music scene. It later led him to crypto, where ownership, permissionless networks, and composability promised a more open internet. 

He first tried to put that thesis into practice with Mediachain (which USV invested in back in the day) and later built Variant around a version of the same idea: that ownership could give users more control, access, and agency online.

Not all of that played out the way he expected. Ownership proved incredibly powerful in some parts of crypto, but much harder to generalize across the rest of the internet.

Today, Jesse thinks AI agents may supply some of what was missing - making it far easier to move between systems, compose different tools, and give users more control over how they interact with the internet.

• 0:00 Preview
• 01:15 Variant's new thesis
• 02:20 Automation ≠ Autonomy 
• 06:50 How piracy shaped Jesse's worldview
• 09:00 How the internet went from open to closed
• 15:45 How BTC inspired Jesse’s startup MediaChain
• 21:00 Why breaking the rules drives innovation
• 24:00 The Variant founding story 
• 29:30 Why crypto's ownership thesis didn’t fully play out
• 34:00 Could AI agents deliver on crypto’s original promise? 
• 37:00 Will AI become another walled garden?
• 39:00 “You can just do things” 
• 46:15 Will AI give us more agency or take it away?

In this episode of The Slow Hunch, I spoke with Clyde Lawrence, co-founder and co-band leader of Lawrence, about what it really means to build an independent career in music.

Our conversation traces how Lawrence slowly evolved into one of the most DIY operations at their level: handling touring, merch, accounting, and fan engagement in-house. We talk about how that hands-on approach led Clyde to uncover structural problems in the music industry, ultimately pushing him to testify before the Senate and publicly challenge industry norms.

I found this conversation especially interesting as someone who spends most of his time thinking about founders, incentives, and market structure. The music industry offers a clear case study in how misaligned incentives compound over time, and how different the system can look when incentives are actually aligned with participants.

Hope you enjoy!

Chapters:

• 00:00:00 Cold open & intro
• 00:02:26 What being "independent" actually means
• 00:05:23 What a record label really does (and why they Lawrence do it themselves)
• 00:07:17 Where the money is now: touring vs streaming
• 00:11:55 Starting scrappy: touring in a van + breaking even
• 00:13:08 The real “privilege”: funding the first record without giving up equity
• 00:20:24 College as a growth engine and fan flywheel
• 00:29:14 Signing (and quickly escaping) a major label deal
• 00:33:01 How Lawrence runs ops internally: tour, merch, social, accounting
• 00:34:19 Getting paid correctly is a full-time job
• 00:48:25 Discovering structural problems in the live music industry
• 00:53:22 Clyde breaks down the math of a live show 
• 00:57:21 Aftermath of the Senate testimony
• 00:01:10:37 Clyde's take on AI in music: will there be a human premium?  

In this episode of The Slow Hunch, I spoke with Matthew Prince, the co-founder and CEO of Cloudflare. Since 2010, Matthew and his team have built Cloudflare into one of the most important companies on the internet: powering and protecting vast portions of global traffic. 

Our conversation explores the through-line from Matthew’s initial hunch about fixing the flaws of the early internet, to Cloudflare’s present role as a foundational infrastructure provider.  

We talk about the early experiments and risks that have shaped Cloudflare’s culture, and how those small bets compounded into a truly iconic company today. Matthew shares stories from the company’s pre-IPO days, the decision to make encryption free, and how Cloudflare’s infrastructure ended up running two of the internet’s thirteen root servers. Toward the end, we dive deep into the transition from a search-driven internet to an answer-driven one, and what that means for publishers, creators, and the future business model of the web.

It was especially fun to record this one with Matthew, who I’ve known since USV’s investment in Cloudflare’s Series C back in 2013. 

Hope you enjoy!

Chapters

• 00:00:00 Curiosity vs. focus; small bets culture
• 00:02:44 Pre-IPO mock earnings calls & learning to take hard questions
• 00:04:48 Matthew’s slow hunch
• 00:05:54 The Unspam origin story: legal mindset meets early internet problems
• 00:11:16 Passing trademark legislation in Utah
• 00:13:39 Meeting Lee (via Arthur Keller)
• 00:18:00 Lee moves to Utah; building from a basement
• 00:20:02 From Unspam to Cloudflare
• 00:20:25 Enter Michelle
• 00:28:19 Realizing how critical Clouflare’s role was (the 2017 outage)
• 00:29:07 Conducting experiments at scale: how small bets can become big lines of business
• 00:31:44 Making encryption free
• 00:33:26 From brittle deploys to Workers
• 00:36:00 Cost curve obsession & why lowest cost to serve always wins
• 00:38:00 Running 2 of the 13 internet root servers
• 00:41:31 Pakistan Telecom story: local demand opens networks
• 00:43:32 Principled decisions > spreadsheets
• 00:44:40 Shift from search engines to answer engines
• 00:48:00 Longing for a quirkier web
• 00:52:56 Incentivizing creators to fill LLM knowledge gaps
• 00:56:05 Designing an open, fair market (price by scale/MAU, not tokens)
• 01:01:15 Scarcity switch flips; next-gen models hit a plateau
• 01:04:00 Google’s role: should AI overviews fund creators?
• 01:06:35 GPUs & researchers commoditize; content becomes the moat
• 01:09:00 Reddit vs. NYT: the value of original/local/quirky content
• 01:10:49 Toward a golden age of content (less rage, more knowledge)
• 01:13:17 Counterintuitive optimism for human-made content

In this episode of The Slow Hunch, I spoke with Alex Komoroske, the co-founder and CEO of Common Tools. Alex has spent his career thinking about how individual incentives can add up to significant collective outcomes. 

Before starting Common Tools, he spent more than a decade at Google leading product management for the Chrome web platform, ambient computing, AR, and Search, and later served as Head of Corporate Strategy at Stripe.

We traced his slow (emergent) hunch from an early fascination with Wikipedia, through his years building internet-scale systems at Google, to his current work rethinking how AI is architected.

A big part of our conversation centered on emergence: why the most durable systems grow from the bottom up, and what that means for product design, org culture, and the future of technology - especially AI. 

We also spoke about the hidden security risks in today’s AI ecosystem: why “chat” may not be the defining paradigm for complex work, how fusing data to apps risks locking us into an AI monoculture, and why policies should travel with data if we want healthier emergent effects.

It’s always fun catching up with Alex. Hope you enjoy!

Chapters:

• 00:00:00 Cold open: the inevitability of transformers
• 00:03:32 Why emergence is so powerful
• 00:08:49 Alex’s early influences 
• 00:10:35 The emergent dynamics of Wikipedia
• 00:13:15 The role of “folksonomies”
• 00:17:33 Concave systems vs convex systems
• 00:20:41 Alex’s time at Google
• 00:24:27 How small signals scale
• 00:28:58 Evolutionary algorithms in AI
• 00:30:52 Understanding data bias and rethinking how AI is architected
• 00:41:02 The same-origin trap and the limits of app-centric software
• 00:47:42 The future of contextual apps
• 00:49:03 Aggregators and the tyranny of the marginal user
• 00:52:08 Why prompt injection is so dangerous
• 00:55:23 The inherent security risks of MCP and vibe coding
• 00:59:32 A new constitution for AI: policies attached to data
• 01:03:17 The promise of confidential compute
• 01:11:00 Why Alex is optimistic about AI's Future

In this episode of The Slow Hunch, I spoke with MC Lader and Marvin Ammori, who spent four years together helping build Uniswap into one of the most important companies in decentralized finance. MC was President and COO; Marvin served as Chief Legal Officer after a long career as one of the internet’s leading policy lawyers.

We traced their shared slow hunch that technology can shift power: first through the open internet, and later through open financial systems.

We also spoke about the parallels between the net neutrality battles of the 2000s and the present-day struggle over how crypto is regulated, the challenge of building in the face of policy headwinds, and why stablecoins, programmable markets, and open protocols are placed to be the next rails for global finance.

This was a fun conversation, recorded at a moment when the policy climate for crypto is starting to thaw. 

Hope you enjoy!

Chapters:

• 00:00:00 Cold open - policy headwinds under Gary Gensler
• 00:05:50 Their shared slow hunch: technology as a force for redistributing power
• 00:14:47 Winning the net neutrality fight
• 00:18:36 First encounters with Bitcoin
• 00:21:14 Parallels between the open internet and DeF
• 00:22:58 Spotting early policy threats and forming the DeFi Education Fund
• 00:23:34 Marvin recruits MC to Uniswap Labs
• 00:29:27 Scaling Uniswap from a tiny team to a full-stack protocol
• 00:34:36 Navigating growth amid SEC opposition
• 00:39:11 Gary Gensler’s impact on US crypto entrepreneurship
• 00:40:45 Stablecoins as the “lily pad” for mainstream adoption
• 00:43:20 Shifting perceptions on Wall Street
• 00:46:12 What’s next: stablecoins, tokenized markets, and on-chain identity
• 00:47:00 Building open, permissionless financial infrastructure
• 00:51:52 Potential risks: fraud, systemic stability, surveillance
• 00:55:45 Stablecoins vs. the fragility of traditional banks
• 00:57:53 Privacy, regulation, and zero-knowledge proofs
• 01:00:00 From DeFi to AllFi: what moves on-chain first?
• 01:04:11 Building for consumers versus institutions
• 01:06:03 Making money feel more human
• 01:08:49 Access to capital as a pillar of opportunity

In this episode of The Slow Hunch, I spoke with Dan Romero and Varun Srinivasan, the co-founders of Farcaster. Farcaster is a social app and protocol that is open, programmable, and crypto-native. 

Before starting Farcaster, both Dan and Varun spent a few years at Coinbase. That experience deeply shaped their perspective on crypto infrastructure, user behavior, and what it takes to build a “sufficiently decentralized” experience at scale.

In this conversation we trace their slow hunch: the idea that social networks needed to be rebuilt from the ground up, as decentralized protocols with credible neutrality, shared state, and a design space open to builders. 

We talked about what they got wrong early on (too much focus on architecture, not enough on user acquisition), how crypto enables new interaction primitives like tipping and token-based identity, and why open programmability (not just ideology) is Farcaster’s biggest edge.

Hope you enjoy!

Chapters: 

• 00:00:00 Cold open
• 00:02:20 What makes Farcaster different
• 00:06:45 Early crypto days at Coinbase
• 00:10:30 Discovering a shared vision for decentralized protocols
• 00:16:07 Why the infrastructure is ready now
• 00:20:55 The social landscape in 2020: Twitter, Mastodon, Bluesky
• 00:23:00 Elon acquires Twitter, FTX, and the narrative shift in decentralized social
• 00:24:53 Designing for "sufficient decentralization"
• 00:29:26 Why the obsession over pure decentralization is a distraction
• 00:32:05 The Farcaster launch story - how they got their first users
• 00:34:30 Why social protocols take time to grow
• 00:36:28 Inventing new content primitives instead of choosing political sides
• 00:41:00 What crypto rails enable: Wallets, tipping, and programmable social UX
• 00:42:38 Reframing money as social interaction
• 00:43:56 Why crypto feels contrarian
• 00:45:59 Crypto as the last frontier of indie building
• 00:47:01 AI vs crypto as platforms for small creators
• 00:49:40 Hiding vs embracing crypto in UX
• 00:50:55 Dan and Varun’s evolving view on abstracting away the chain
• 00:54:00 The adjacent possible: mini-apps, embedded wallets, AI video
• 00:59:00 Using AI to surface context + trending content
• 01:00:54 What big platforms won’t do: programmable money
• 01:03:57 Crowdsourced Q&A – early Farcaster days
• 01:06:53 Why mobile UX is everything
• 01:07:00 The surprising difficulty of building other clients
• 01:08:43 Varun on shifting from text to video
• 01:13:00 Why they cut encrypted messaging
• 01:15:00 Closing thoughts

In this episode of The Slow Hunch, I spoke with Ben Leventhal, the founder and CEO of Blackbird. Ben has spent the past two decades reimagining the restaurant industry, having previously co-founded Eater and Resy. 

The throughline that connects his efforts is a strong belief that restaurants are universally loved but fundamentally broken businesses—and that there must be a better way to run what is a trillion dollar industry in the United States alone. 

We talked about what’s gone wrong with the restaurant business model, why most restaurants struggle to turn a profit despite enormous consumer love, and how each of Ben’s ventures has tried to close that gap—first with content (Eater), then with mobile (Resy), and now with crypto (Blackbird). 

Through Blackbird, Ben is using crypto rails to build a restaurant-native platform currency: one that rewards regulars, strengthens margins, and builds more intimate ties between diners and the places they love.

We recorded this conversation in my apartment in New York, just around the corner from a restaurant I paid for using Fly, Blackbird’s currency. Few founders have followed a hunch as consistently and creatively as Ben. 

Hope you enjoy!

Chapters:

• 00:00:00 Cold open: why restaurants are broken
• 00:01:10 Introducing Ben Leventhal
• 00:02:00 Ben’s slow hunch: the status quo is always wrong
• 00:05:00 Falling in love with restaurants as a kid
• 00:08:00 She Loves New York: the proto-Eater newsletter
• 00:10:30 Early blogging and New York’s indie media scene
• 00:14:00 Starting Eater with Lockhart Steele
• 00:16:00 Eater as “sports coverage” for restaurants
• 00:18:00 Why restaurateurs initially hated Eater
• 00:20:30 Scooping the New York Times
• 00:22:00 The adjacent possible and building with new tools
• 00:24:30 Leaving Eater and exploring new projects
• 00:25:50 The Resy origin story
• 00:27:30 Resy’s mobile-first wedge: outdoor seating and Notify
• 00:31:00 Selling Resy to Amex
• 00:33:00 Why Resy was restaurant-first (and OpenTable wasn’t)
• 00:38:00 The COVID reset: restaurants become brands
• 00:45:00 The idea for Blackbird takes shape
• 00:52:00 Introducing Fly: a platform currency for restaurants
• 00:56:00 How Fly helps restaurants recapture value
• 01:00:00 Restaurant regulars as shareholders
• 01:03:00 Designing Blackbird to feel like a consumer app
• 01:04:00 What’s next: AI and the future of restaurant marketing

In this episode of The Slow Hunch, I spoke with Juan Benet, the founder and CEO of Protocol Labs. Juan is best known for creating IPFS and Filecoin—two foundational technologies in the decentralized web. Through Protocol Labs, Juan wants to use decentralized protocols to unlock new ways of organizing capital, governance, and research. 

This conversation was recorded more than a decade after we first met, when USV seed funded Protocol Labs. Juan is one of the deepest thinkers I know, so naturally this conversation has a wide aperture. We touched on internet history, crypto network design, public goods funding, and the future of intelligence itself.

Hope you enjoy!

Chapters:

• 00:00 Cold open
• 00:07:39 Discovering the power of P2P tech
• 00:11:10 Decentralization versus central planning
• 00:17:44 Juan’s slow hunch
• 00:21:16 Protocol Labs as a way to plug the R&D funding gap
• 00:24:03 The genesis of IPFS and Filecoin
• 00:30:45 Layering protocols for success
• 00:33:18 New frontiers in governance
• 00:41:09 The role of public utilities
• 00:47:39 Juan’s thoughts on the future of work
• 00:52:41 Blockchains as a governance layer for AI agents
• 01:01:32 Evolving into a new species

In this episode of The Slow Hunch, I’ve pulled together some of the best moments from Season 1.

Across these conversations, what stood out was how many ‘inevitable’ ideas were, at one point, anything but.

Venture veterans Fred Wilson and Brad Burnham reflected on decades of investing, Bluesky CEO Jay Graber spoke about her vision for decentralized social, and author and NotebookLM co-creator Steven Johnson unpacked the future of networked thought.

We got a clearer view into how their slow hunches took shape — and where they’re headed next.

Hope you enjoy!

Chapters: 

• 00:00:00 Cold open 
• 00:02:53 Fraser Kelton (GP at Spark Capital, former Head of Product at OpenAI)
• 00:09:06 Dani Grant (CEO of Jam.dev)
• 00:20:02 Amir Haleem (Founder of Helium, CEO of Nova Labs)
• 00:25:10 Fred Wilson & Brad Burnham (Union Square Ventures)
• 00:30:40 Jake Heller (Co-founder & CEO of Casetext)
• 00:40:00 Steven Johnson (Author, Editorial at NotebookLM and Google Labs)
• 00:49:35 Muneeb Ali (Co-founder of Stacks, CEO at Trust Machines)
• 00:53:33 Jay Graber (CEO of Bluesky)
• 01:03:26 Zoe Weinberg (Founder, ex/ante)
• 01:09:25 Aaron Wright (Co-founder & CEO of Tribute Labs)