Love seeing more networking startups pop up. If you want to know whether the founders can actually solve the problems they're tackling, just ask them to walk you through this diagram. You’ll have your answer in 30 seconds

Compute is getting closer to the user and Amazon is quietly abandoning the cloud. Read this. I have heard this same thing from former Amazon employees. DePIN is an inevitability.

The idea of a Virtual Power Plant (VPP) is legit fascinating. My good friend @jasonbadeaux explained it to me, and it blew my mind because it’s so simple yet the impact is profound. Instead of generating new electricity, a VPP coordinates existing "surplus" power sources from thousands of households, making them behave collectively like one large, unified power plant. Here's a simple example, imagine a future where every home has a battery bank: when the grid has surplus energy, batteries charge up, when the grid faces shortages, these batteries discharge, providing the necessary juice to maintain the grid. But here's the mind blowing part Jason showed me: "surplus" doesn’t only come from batteries. It also comes from simply choosing NOT to use power at a critical moment. If I decide not to run my 1kW washing machine during a heatwave that's threatening a grid shutdown, I've just provided a real, tangible surplus to the grid. And that surplus is valuable to power companies who would gladly pay for it at those critical moments. The profound implication: what if we could aggregate all these small decisions into a token-incentivized platform? That’s exactly what Jason's protocol, @daylightenergy_ is building. Daylight provides the incentives and intelligent orchestration to scale VPPs. Whether it's coordinating battery storage or smart sensors adjusting power consumption of devices like our Black Box running AI workloads, Daylight is setting the stage to be the defacto power orchestration platform of the future. With digital infrastructure moving closer to homes (as our Black Box exemplifies), the ability to intelligently manage edge power consumption becomes crucial. After all, if the grid shuts down, none of these systems matter. You can't run AI at the edge if you don't have electricity, so you need to balance the needs of AI against the needs of the grid. And by keeping the grid up, you can get rewarded for doing your part. The future of digital infrastructure hinges on reliable electricity. And in a world where energy demand only goes up, it's hard to imagine a future without tightly orchestrated VPPs powering the last mile. Lfg Daylight.

Mango lassis are the biggest bullshit in the restaurant industry. I know you guys pour that from a jug you got at Patel Brothers and yet you have the audacity to charge $7

The journey to finding the right GPU partner for the Black Box was an interesting one. I've long felt frustrated by how decentralized GPU computing seemed to be drifting away from retail users, focusing instead on $3k+ GPUs. Yet retail built this industry! Retail is where the power is. While I can't knock following where demand is, I believe that the true strength of DePIN lies in enabling broad participation. Massive consumer involvement is the most important force multiplier for crypto I'm concerned with atm. So when we asked ourselves, "What's the best way to reward our users using the BB GPUs?" we realized we had to move up the stack with our protocol partners and find protocols designed specifically to harness the retail's power. We wanted a partner whose technology was positioned to meet future industry demands over the next five years by uniquely leveraging retail. With this backdrop in mind for our values and the market of protocols available to us, it's hard not to fall in love with the elegance of Swarm Inference. @inikitin and @thisisfirsov 's vision at @fortytwonetwork of building an inference layer from many lightweight models working collectively, competing with expensive GPUs, is genuinely elegant. Think of the emergent properties of ant colonies, or bee hives. In the same way, Swarm Inference in AI feels almost magical. It gives me real hope that the future of AI can indeed be powered by average households. There are other demand signals that reinforce my belief that Swarm Inference will become an essential part of AI’s future. Supporting this vision through the Black Box is something I'm very stoked for. We've got an incredible future to build.

Multicast is one of those terms you'll probably only encounter if you're deep in learning about Internet engineering. It lets a single sender deliver data to multiple recipients at once. Given how much content we consume simultaneously (live sports, major events), you'd think multicast would be everywhere, right? Yet most of the Internet relies on unicast, which is only one sender and one recipient. So even if ten people in the same neighborhood are streaming the same event, ten copies of that event are being sent over the network. Multicast feels like the Internet’s silver bullet, but it's been limited mostly to core and private networks, while the last mile (where the consumption actually happens) is largely untouched. This is exactly why I'm so excited about what my good friend @oaliramadan is building at @BlockcastNet. Many know Omar as "the multicast guy," but his real vision is even bigger -- combining multicast with an edge-based CDN. This is exactly how tech moves from niche engineering to solving real problems at scale. Content delivery needs help scaling, and Blockcast brings new technology, in new locations, and a model that leverages existing infrastructure folks already have. And the protocol that governs this technology intersection between content and the orchestration of how it's delivered, has to exist. The opportunity is too meaningful -- networks today have a 10x efficiency upgrade just waiting to be tapped by the right orchestration layer. We're soon going to be live streaming historical moments like the Artemis moon landing with hundreds of millions, maybe billions watching simultaneously. The current Internet is simply not ready for events of this scale. Content and its delivery are deeply intertwined, and I'm thrilled to help make Omar’s vision a reality. Especially for wireless networks like DAWN, optimizing airwaves is critical. We've got some exciting research with Blockcast coming out soon on this and more. Stay tuned.

weebs work at @pumpdotfun confirmed

Before building wireless networks (actually less than six months before first building Andrena with Princeton) I got completely sucked into cryptography because of homomorphic encryption. Funny enough, I wanted to build houseless poker and blackjack. Thankfully @yrschrade and the @ArciumHQ team have a much better vision than I did. I'd like to think there's a moment in everyone's cryptography journey where public key cryptography just clicks. You see its elegance and realize why the Internet relies on it, how powerful it is for securely sharing data. For me, the work Arcium is doing with FHE, ZKPs, secure MPC + more, feels like that next elegant leap in cryptography. If public key gave us scalable encrypted data sharing, then these techniques are the logical next step -- securely working on encrypted data. This tech is so foundational for the future of a better, decentralized Internet that fewer things are as important. Arcium's technique towards building the encrypted supercomputer for the world finds itself at this interestinct intersection of computational demand, practicality, and necessity. It truly is an honor to help them build this vision with the Black Box. And I'm stoked to see a world where your HTTPS lock is upgraded to a purple Arcium lock that lets you know your data never left your device.

Very few folks genuinely understand how to make IBRL a reality. One of the things I love most about the @solana ecosystem is that it attracts the best and brightest who actually know how to make meaningful progress here. Saying I'm "excited" to partner with @pipenetwork doesn't quite capture it. @DavidRhodus and the Pipe team are more like kindred spirits, driven by the same deeply technical vision. When it comes to the content delivery stack for billions of users, Pipe operates literally one hop above us, right after the network layer. They're relentlessly obsessed with optimizing their layer, just as we are with ours. And because our layer directly hands off packets to theirs, this partnership is uniquely important. It means we have to execute flawlessly so Pipe's code can achieve peak performance. There's a specific piece of the DAWN Black Box purpose built exactly for this: the virtualization and network layer. Our system virtualizes workloads and passes data directly to applications using zero-copy networking. This ensures that the Pipe team can have confidence that on our end, bandwidth is maximized and latency minimized, down to the hardware level. I'm eager to dive deeper into this tech because as IBRL gains importance, every application serious about it will have to run something similar to what we've built. The Black Box isn't just a platform for convenience. It's a platform for performance.

I'm so excited for this. After reading AI 2027 over the weekend, I'm so long on user-owned AI. It's gotta happen or there will be some serious risks down the road. Honored to team up with @inference_net and help bring power back to users. 10 total partner announcements coming.