Investment Partner @a16z | prev investing @dragonfly
New York, USA
Joined July 2020
- Tweets588
- Following1.7K
- Followers6.6K
- Likes1.6K
One prediction I have is we will see a media platform compete with its own supply-side creators by producing its own ai-generated content
This becomes especially relevant as ai content creation eclipses humans in eliciting max dopamine, which feels inevitable
The concept of an algorithm that selects from a pool of human-generated media will collapse into a new model
The platform itself will simply generate whatever next piece of content elicits max dopamine for you personally based off your training data they own
And as models improve, they will become very very good at this
To the point that the dopaminergic opportunity cost could become too high for purely human-generated platforms to compete on this
Obviously more nuanced but feels directionally plausible
There’s a simple framework for what consumer interfaces agents will replace:
Does this task induce dopamine or cortisol?
The interfaces for cortisol-inducing tasks (research, planning, scheduling, commodity shopping etc) will get replaced by agents
The interfaces for dopamine-inducing tasks (TikTok, instagram, X, discretionary shopping, gambling, dating apps etc) will be insulated
Scrolling is still the greatest form factor for delivering refined dopamine and agents won’t change that
The lion’s share of consumer gross profit gets paid out to whoever owns discovery. Today that’s Google, Meta and TikTok. Brands rent attention from platforms insulated by network effects
Personal agents fundamentally change this model
Google search infers your intent from a query. Social guesses from your likes and views. An agent gets your full intent (“I want to buy this, here’s my budget, help me”) and then makes the decision for you
Not only is the intent more legible, the agent also helps with (or entirely handles) executing it
It’s easy to understate how valuable this is. The hard part is monetizing it. There are four paths in my view:
(1) User pays: a monthly subscription for the consumer surplus the agent delivers, likely tiered by usage
(2) Payment for intent flow (PFIF): brands bid to win the sale or rank first. Like payment for order flow, it only works with a “best execution” guarantee that the user still gets the best outcome
(3) Brands pay to be considered, not chosen: brands buy a spot in the candidate set, and the ranking stays neutral
(4) Interchange: the agent issues its own card. Merchants do the subsidizing either way, so the recommendation can stay neutral
Agents will increasingly own discovery (especially for undifferentiated consumer products). The question is how they capture this value without corrupting the recommendation
The bull case for agentic commerce rests on the assumption that the inputs governing consumer decisions will be legible to an agent
Beyond price — which is rarely the only input — there are countless other qualitative inputs that aren’t just illegible, but dynamically revealed to consumers through the discovery process itself
In other words, I may have an idea of what I want to purchase but my conviction is refined through seeing multiple options
I’m incredibly bullish on agents augmenting consumer discovery and compressing time-to-purchase. But, for most consumer purchases, I expect humans will manually authorize the final decision and just pay with Apple Pay or a saved card
You could call this “agentic commerce”, semantically, but it doesn’t require new payment rails
Excited to announce that I'm joining @a16zcrypto as an Investment Partner
My thesis on crypto has always been the same: if you believe the arc of technological adoption bends toward efficiency, market forces will inevitably reveal that blockchains are the most efficient technology for global coordination
This is already evident across stablecoins, perps, prediction markets and tokenized assets. The only bottleneck to blockchains eating the rest of global markets – and creating net new markets – is time
The delta between sentiment and fundamentals has never been wider in crypto, and therefore I've never been more bullish
Couldn't be more fired up to work alongside @cdixon, @alive_, @guywuolletjr, @eddylazzarin and the entire @a16zcrypto team, backing the most ambitious founders pulling this future forward!
Today is my last day @dragonflyvc
Very few places have fostered a culture that embraces independent thought like Dragonfly
Incredibly grateful for the opportunity to work alongside some of the sharpest investors in the space - @hosseeb @FengBo22588938 @HadickM @tomhschmidt @TheOneandOmsy @0xsudogm @kapursanat @0xelton - and the exceptional founders Dragonfly has had the privilege of partnering with
In terms of what's next, more to come
The biggest opportunity in fintech today is using stablecoin virtual accounts + cards to run the Ramp playbook globally
Ramp is valued at 32B and their entire business is US-based – a market where (1) dollars are the local currency and (2) most businesses already have some expense management software
Outside of the US, there is a massive delta between demand and penetration of dollar business banking wrapped in software (expense management, AP/AR, ERP)
It will be obvious in 5+ years that the killer use case for stablecoins was closing this delta at a speed and scale that wasn't possible before
Most importantly, these platforms make for exceptional businesses
Owning spend at the card layer is by far the most valuable layer of the stack as you clip interchange on every dollar spent. And once you're embedded into finance workflows, switching costs are incredibly high
Despite the recent DeFi exploits, two things will be true over the next 5 years:
(1) More yield will find its way into the pockets of end-users: Idle dollars earning nothing across banks and fintechs will be a competitive liability. Markets will naturally select for those that offer native yield
(2) The best risk-adjusted yields will exist on-chain: Blockchains make risk more legible. Atomic liquidations on a single state machine unlock a pareto improvement in risk management — making on-chain yields impossible to replicate off-chain
If you believe both of these things, you implicitly believe the on-chain economy grows by an order of magnitude over the coming years
Blockchains are the only venue in the world where one can issue net new assets without permission and distribute these assets globally to anyone with an internet connection
Every category-defining outcome in crypto has been downstream of permssionless issuance and global distribution – stablecoins, perps, prediction markets, synthetic dollars, memecoins, NFTs
This is the technological unlock for blockchains – enabling a new marketplace of supply and demand side participants, who would have otherwise been censored, to coordinate at scale
While it’s unclear what the next iterations of net new assets will be (RWA perps and collectibles are interesting front-runners), on-chain will remain the path of least resistance
Which payment rails (i.e. traditional card networks vs blockchains) end up underpinning the agentic economy can be reduced to a simple question:
Will legacy rails solve for micro-transaction costs / settlement speed faster than on-chain entrants solve for auth / risk scoring?
Robbie Petersen retweeted
Today we are incredibly excited to announce our oversubscribed $650mm fourth fund. It has been almost four years since @hosseeb and I initially started the discussions about me joining the @dragonflyvc partnership with him, Bo and @tomhschmidt yet it feels like different lifetimes altogether. In that period of time, we've had the complete deflation of the risk asset bubble that topped in 2021 and the blow ups of companies and protocols like Terra Luna, Genesis, and FTX. We've seen Solana at both $8 and $290. Larry Fink has become one of the industry's most prominent spokespersons. And the governments of many of the world's super powers have started to embrace the benefits of BTC, blockchains, stablecoins, tokenization and prediction markets.
Much has been written about the death of venture capital, or even more specifically, the death of crypto linked VC. As many would tell the story, VC is either extractive and short term oriented or it's long term oriented but unable to return capital to investors. That token designs are so broken that private equity is the only way to make realistically attractive returns. Platforms will eat the rest of the market except for small, niche asset managers with extremely narrow views of the world. And while there are truths embedded in all these points, there are also counterfactuals.
We feel strongly that our success over that time, and since inception, is due to the fact that we approach every day from first principles. What actually works in our space? Finance, payments, asset issuance, and markets. What is the appropriate instrument for investment? It depends and there is no one size fits all solution. How should we approach geographic focus? This is borderless, global technology that isn't defined to a single country or region. Are blockchain communities winner take all? Of course not, there are too many use cases that can be served by a variety of technologies. Day in and day out, we challenge our priors and each other to ensure we are not driven by the market, but that we are focused on what we expect to be true on the horizon.
Ultimately, we have the easy job as capital allocators who exist primarily to partner with the best founders building in this arena (and i'm lucky to call many of them friends). But at a time when both our space and many of our competitors are going through existential crisis', we wake up lucky to know that we've built a team and partnership that can handle any and all market cycles and form factors - and is ready for the next phase of financial evolution.
More below from @leomschwartz
fortune.com/2026/02/17/drago…
Robbie Petersen retweeted
We just closed Dragonfly Fund IV at $650M.
It's a big milestone, and yet, it’s a weird time to celebrate. Spirits are low, fear is extreme, and the gloom of a bear market has set in.
But here's the thing: we raised almost every single Dragonfly fund into bear markets.
Fund I we raised through the 2018 ICO winter, when almost nobody believed in this space anymore. Fund III we raised right before Luna collapsed. Those were brutal times to deploy capital. But they turned out to be our best vintages.
Last week I caught a lot of heat for arguing that non-financial crypto has failed. I meant that. But the flip side of that argument is: financial crypto is exploding. Stablecoins are eating the world. DeFi has grown so big it's rivaling CeFi. Financial institutions around the world are racing to build out their crypto strategies. And prediction markets are becoming the most trusted source of truth on the internet.
Fund IV is our biggest bet yet that the crypto revolution is still early in its exponential. If you look at our recent bets—Polymarket, Ethena, Rain, Mesh—the growth speaks for itself. Agentic payments, on-chain privacy, the tokenization of everything—crypto's surface area is about to explode, and we want to be backing the founders at the center of it.
We've always believed that the most important work gets done when the noise dies down. We believe moment is now. In fact, we’re putting money on it.
If you're building what comes next, @ us.
The nuance everyone is missing with prediction markets is that order-flow is inherently more toxic
While a random bettor in Pennsylvania is unlikely to have an edge buying APPL stock, they may actually know more than the market on the Taylor Swift or Trump election market
This fundamentally changes how value will accrue to front-ends vs exchanges (i.e., Polymarket, Kalshi)
In equity markets, 90%+ of volume settles off-exchange with wholesalers (Citadel, Virtu) internalizing flow by matching buyers/sellers from their own inventory
Intuitively, this micro-structure biases front-ends (Robinhood) who disproportionately capture value by selling their "non-toxic" retail order flow directly to MMs, completely circumventing exchanges
Market makers however cannot warehouse binary outcome risk like they can with stock inventory given (1) binary contracts are structurally imbalanced (especially close to resolution) and (2) there are no instruments to directly hedge exposure
This suggests nearly all prediction market will have to settle canonically on-exchange
While front-ends like Robinhood and Coinbase will still capture value through front-end fees and rebates, they won't be able to monetize via PFOF. There are also regulatory reasons why volume will have to settle on-exchange as PMs will be under the regulatory purview of the CFTC, which mandates centralized settlement
This micro-structure biases the exchanges themselves (i.e., Polymarket, Kalshi) - being the most liquid venue where price discovery happens canonically becomes the moat
My theory for why crypto has underperformed is counterintuitively because trump won the election and delivered regulatory clarity
Operating under a degree of optimistic uncertainty was perhaps the optimal recipe for positive price action. It gave people a reason to buy, and more importantly, to not sell as they patiently waited for these catalysts to materialize
However, as this "optimistic uncertainty" became realized by the market, there's now simply less of an incentive to buy or hold
Crypto prices are more or less proxies for social sentiment and sentiment is something that needs to be constantly fed with new catalysts
Ironically, the best catalysts are those that never definitively arrive (e.g. perpetually being on the cusp of "institutional adoption", "monetary debasement" etc)
In the absence of legible fundamentals, it's the anticipation - not realization - of these catalysts, that govern price action
The illegibility of crypto forces the marginal buyer to operate under “borrowed conviction”
Consequently price become a proxy for social sentiment
However social sentiment is itself a function of price
This is why crypto markets are so reflexive, both to the upside and downside
This is also why sentiment bottoms are the ultimate signal