Alex Tang

I opened a decent-sized position in Google after the market rout. Folks are completely over-indexing on large companies’ ability to take advantage of these market opportunities and are fully underestimating that Google has a full-stack solution, unlike most companies that are trying to build one...from scratch Not sure how to balance the option strat... yet...

4th year supporting @bhorowitz & @FeliciaHorowitz and it still hits. Watching @GrandmasterCaz and @nilerodgers finally run the real version of Rapper’s Delight on stage is such a point. The architects of hip-hop finally getting paid in full (financially + culturally) is one of the clearest corrections of the record in music history. Culture shapes how we see the world and making sure the people who built it get their due has been an amazing mission @paidinfullorg

A few thoughts after staring at this chart… i believe the biggest risk in investing in a new era is getting very good at underwriting the last one. And when I look at this chart from a16z it’s a useful picture of where AI infrastructure spending goes. It also raises a question for me…how much of our investment thinking is shaped by what we already know how to value? And as a consequence of that placing everything in previous world frameworks. Chips, power, networking, data centers are all obviously very essential. All relatively familiar and easy to underwrite…capacity, contracts, utilization, margins. I think so what if you really dig into it is that similar to the early days of the Internet and even the current stage we’ve seen businesses that emerge when intelligence becomes dramatically cheaper and easier to access are harder to model. Their customers may behave differently. Their business models may not have obvious comparables. we’re seeing these pressures with token economics in AI. Funny enough Crypto has been modeling this for a decade at this point The reason I think this matters is because capital follows perceived risk-adjusted returns. What feels lower-risk is often what fits an existing framework. To go a little bit deeper- we’ve seen versions of this through the internet, SaaS, and mobile transitions. Each created new ways to distribute products, charge customers, and build habits. Investors who carried forward the wrong assumptions could recognize the importance of the technology and still miss where the value would accrue. The venture firms that were willing to think differently and underwrite in completely ridiculous ways won big. I think consumer AI is one of those places today. Think about why people wanted a BlackBerry…access to people, information, and decisions without returning to a desk. Phones expanded that access enormously, but we still spend a lot of time translating what we want into what software requires. Open an app. Search. Type. Do a thing… and the switch apps. Repeat. I do the same thing in Google Chrome tabs far more efficiently. I just have to carry a clunky laptop around. Well what happens when we can simply express an intention and trust something to carry it through? Devices like Meta’s announced Muse Charm make that question tangible. The opportunity extends well beyond any particular device- intelligence that fits naturally into daily life, through conversation, context, and less effort. And I think their glasses were old take things to another level- especially if they continue on their current trajectory. For Apple, the threat begins before someone gives up their iPhone. It begins when another company becomes the first place they turn with a question, a decision, or something they need done- and they get tired of looking at their device all the time. And I think with the dumb device sales it’s cleared that there is a screen fatigue. Adding a second screen in my opinion is not the solution. The thing is that easier access actually creates another problem. If every founder, executive, and management team consults similar systems, we may get more polished versions of the same … group think. If you’ve listened to a recent cuddly earnings call, it’s pretty sad to hear the politicians executive. Discernment obviously becomes way more valuable. knowing which perspective matters, what deserves skepticism, and when the apparent consensus is wrong. The most important will be who is brave enough to actually think this was an action on it. Partners have to earn TRUST through judgment—including the willingness to disagree. My bet is that we are underinvested in discovering these consumer behaviors and building around them, even as enormous capital flows toward supplying the underlying intelligence. This is a question i keep asking myself which assumptions from the last era are becoming less useful, and what opportunities become interesting when we let them go?

We were in over 15 cities in Q1…here’s what’s broken: The Non-Deployers vs. the Squirrley Operators Over the last few months, our MFV team sat down with sharp founders building in the Bay Area, NYC, Tokyo, Hong Kong, and Beijing - - plus some of the best venture funds in deep tech. The same refrain from most LPs and big funds: “We’re just not deploying right now.” Choppy markets. SaaS is getting eaten alive by agents. Hardware confusion. Endless physical AI hype cycles. On paper, it makes sense. In reality? It’s not caution. It’s a lack of conviction in what will actually move the needle next: AI-native infrastructure that agents and physical systems can rely on. It’s also strange to watch pundits and macro theologians opine loudly but rarely put capital behind their takes. We just met with a large fund of funds. Most of their managers are in pure observation mode - - rearticulating consensus rather than backing operators who ship. Meanwhile, the clearest signals come from builders in the trenches. No luxury of waiting. They’re solving real deployment problems today: settlement latency, compute tokenization, and getting physical systems to work in the wild. That’s been the MFV signal. So we’ve kept deploying - - across AI infra, agent economies, and physical systems. Boston to Beijing, and we’re only three months in. Here’s what the data actually shows right now (PitchBook + Crunchbase Q1 2026 signals): 1. Global dry powder remains at record levels. 2. Yet AI (infra, agents, physical/robotics) took over half of VC dollars in 2025 and continues accelerating. 3. Big funds (> $500M) are writing oversized checks into exactly these categories. 4. Everyone else? Mostly on the sidelines. The split is brutal: conviction in AI execution versus hesitation everywhere else. Founders - if your investor conversations feel like sophisticated information extraction with zero follow-through, you’re not imagining it. At MFV, we bet on Squirrley operators: nimble, high-conviction teams who treat tokenized compute and agent settlement like payroll software - - practical, repeatable, slightly eccentric in the best way. We’re always up for comparing notes. AI and agentic builders shipping infra, physical systems, or real agent economies right now - - what’s the #1 deployment or capital friction your VCs are missing? DM me. Let’s talk.

This letter was originally written for MFV’s limited partners and is shared in adapted form to give prospective partners a clearer view into how we think and operate. MFV Partners, As we’re writing this note, we’re sitting in East Asia. We’ve been spending time between New York and Beijing to start off our year and double-down on our East-meets-West approach. On the ground we smelled something in the air that reports are not picking up. Having experienced the rate of execution in the East first-hand for over a month, we can’t help but imagine the possibilities when paired with the vision of the West. We are convinced that Maximum Frequency Ventures, bridging the West and the East, is needed today more than ever before. Read more below 👇

The @andyyy pull up for a market download and clothing recommendations was a moment.

The way I think about this is there are two models people keep using to talk past each other. One is Haseeb’s model, which is basically revealed preference. The market already ran the experiment, and the only things that have worked at scale so far are financial. The other is Chris’s model, which is sequencing. Categories don’t appear all at once. Infrastructure and distribution come first, and only then do entirely new consumer behaviors become possible. Both models are directionally right. But the more correct model is the convergence of the two. Revealed preference is only revealed inside the constraint set of what can actually be shipped, distributed, and legally survived. Change the constraints and you reveal new preferences. That’s why I don’t buy the idea that “consumer crypto is dead.” I buy the idea that the first era of crypto was a narrow era, mostly finance, because finance was the only category that could survive early conditions and bootstrap the rest of the system. And I can say this firsthand. When we were building Libra/Diem inside Facebook, it wasn’t some pitch deck fantasy. It was the most serious attempt by the largest consumer distribution machine in the world to bring crypto rails into mainstream product surfaces. When regulators shut that down, it didn’t just stop one project. It sent a system-wide signal: touch this and you’ll get crushed. That second-order effect is everything. Less capital flows in. Less engineering talent stays. Fewer serious operators take the risk. Innovation slows. And the only things that keep compounding are the things that can survive in hostile conditions, which again are mostly financial. Now fast forward to today and the interesting part isn’t relitigating the last cycle. It’s understanding what the next constraint set unlocks. For the first time, we have high-performance protocols that can actually support consumer-grade UX. We have stablecoins becoming normalized. We have better wallets. We have clearer policy momentum. And we have the real wildcard ... AI commercialization accelerating the speed at which new product categories get built, iterated, and distributed. That convergence is where the next wave comes from. Not “decentralized Spotify.” The real frontier is things like IP tracking and rights systems in a world where AI is creating content at scale. New IP creation where ownership and provenance actually matter. Entirely new forms of transacting, not just person-to-person, but person-to-agent and agent-to-agent. New forms of social that look closer to Twitch economies and interactive media than the last era of social feeds. And a new kind of consumer ownership where the value isn’t a token for the sake of a token, but a market structure that makes the product itself work. This is exactly how mobile computing played out. In 1995, you couldn’t “revealed preference” your way into predicting the iPhone, app stores, creator economies, streaming, or the way people would live inside their phones. Because the sequencing was TCP/IP, telco infra, screens, batteries, browsers, distribution, then the consumer behaviors that felt obvious in hindsight. Crypto is on that curve. Finance was the first wave because it could bootstrap everything else. The next wave is where you find out what people actually want when the stack is finally fast, cheap, usable, and legally survivable.