
One of my favorite "moves" is to give myself a day to process an important decision by "sleeping on it". While I feel like I generally tend to move quickly on things and I like being decisive, sometimes it's important to let a situation marinate and let some background processing happen.
This background processing is important. This article from MIT makes an interesting observation:
"as we enter a dreamlike state, the prefrontal cortex of the brain, responsible for keeping impulses in check, slowly grows less active. This is when there’s a spur in theta waves that leads to an unconstrained window of consciousness; there is little censorship from the mind, allowing for visceral dreams and creative thoughts."
(Why do we dream? by Shafaq Zia)
In other words, dreaming (sleeping on it) gives your brain a unique opportunity to process thinking, unbothered by your daily guard rails.
Interesitngly, AI systems are now forming up to do something similar. From the Anthropic / Claude docs on their dreaming feature:
"Agents write to their memory stores as they work, but these writes are local and incremental: over many sessions a memory store accumulates duplicates, contradictions, and stale entries.
Dreams let Claude clean that up. A dream reads an existing memory store alongside past session transcripts, then produces a new, reorganized memory store: duplicates merged, stale or contradicted entries replaced with the latest value, and new insights surfaced."
Both are a great reminder that some of your best thinking can happen when you're not even really trying. And, the more time / opportunity you can give yourself to sleep on it, the better. Clearly, this is core component of any Slow Hunch.

I just released a new episode of The Slow Hunch with Matthew Prince, CEO and co-founder of Cloudflare. Since we invested in their Series C back in 2013, I've watched Matthew and his team build one of the most critical pieces of internet infrastructure—protecting and accelerating vast portions of global web traffic.
Our conversation traces Matthew's journey from his early "slow hunch" that the internet was fundamentally broken and needed fixing. We start with his law school days in 2000, when he was annoyed by email spam and registered the domain "unspam.com." That legal mindset—thinking about jurisdiction problems the internet created—would eventually collide with technical brilliance when he met his future co-founders Lee Holloway and Michelle Zatlyn.
The origin story alone is worth the listen: Matthew literally pitched his trademark protection idea at an FTC hearing, got confronted by a professor who had built the exact same technical solution, and ended up hiring one of that professor's students (Lee) who would become Cloudflare's technical co-founder.
But the real meat of our conversation is about what's happening right now. We're living through a fundamental shift in how the internet works: moving from search engines to "answer engines"—from getting a "treasure map" of links to getting direct answers. This changes everything about how the internet's business model works.
Matthew argues we're at a critical juncture. The current model where publishers depend on traffic (and the ad revenue that comes with it) is breaking down because AI systems provide answers without sending users to the original sources. We could end up with a handful of AI companies controlling their own fleets of journalists and researchers—essentially returning to a medieval patronage system. Or we could build something better.
The path Matthew envisions involves creating a fair marketplace where AI companies pay content creators based on their scale and usage, similar to how Spotify compensates musicians. Matthew shares the story of a music creator making 40 million euros annually by writing songs for unfulfilled Spotify queries—imagine that model applied to knowledge creation.
What struck me most was Matthew's counterintuitive optimism. While many see AI as the death of human content, he believes we're heading toward a golden age of content creation. The key insight: Reddit got paid seven times more than the New York Times for the same number of tokens in AI licensing deals. Why? Because original, quirky, local content is more valuable than commodity news coverage.
We also dive deep into Cloudflare's culture of "small bets" and curiosity over focus, the decision to make encryption free (which eliminated their main paid feature), and how running two of the internet's thirteen root servers helped them expand globally.
Matthew's fundamental thesis remains the same as it was 25 years ago: the internet is more important than people realize, it's deeply flawed, and it needs to be fixed. The difference now is that the stakes are higher, the problems are bigger, but the opportunity to create positive change might be greater than ever.
Full episode here: Youtube, Spotify, Apple Podcasts

In this episode of the Slow Hunch, I spoke with Jay Graber, CEO of Bluesky.
Originally conceived as an initiative within Twitter under Jack Dorsey, Bluesky was designed to transform Twitter from a closed platform to an open, protocol-based network. For those of you who caught the episode with Brad & Fred, this is a subject we got into in some detail: years ago, USV had an intuition that Twitter might be better designed as an open protocol, but at the time, the tools didn't exist to make it a practical reality. In 2019, Twitter launched Bluesky (and the underlying AT Protocol) as a research initiative, and that's when Jay joined to lead the effort. She subsequently lead the spin-out of Bluesky into an independent effort, developed the app and the protocol, and launched the network. Since launching publicly in early 2024, Bluesky has since grown to roughly 30 million users.
I've always liked the core design of ATproto and Bluesky, which lets each user configure core elements of their social experience, pulling from an open library of feeds, filters and client apps; all tied together with a single, user-owned identity (my identity on Bluesky is actually my DNS domain name, nickgrossman.xyz). In our conversation, Jay talks about why this architecture matters so much to her: primarily by giving users the ability to vote with their feet if the platform makes changes they don't like.
Jay is one of the clearest thinkers I know on the relationship between protocol design, system architecture and user experience, and it's been pretty incredible to watch her build Bluesky from a high level idea into a scaled consumer product over the last few years.
I hope you enjoy our conversation!
As always, you can listen to this episode on Spotify, Apple, or wherever you consume podcasts. You can also watch the full video on YouTube.
