RWA tripled onchain. Most of its tokens still lost money

If tokenizing real-world assets is the growth story of the cycle, why did six of the seven biggest RWA tokens lose money while the sector they represent tripled — and does that mean the token is the wrong place to express the trade?

The chain of the argument

0xD0M_ posted the number that started it: tradable RWA value onchain roughly tripled over the past year to ~$33.5B, yet six of the seven biggest RWA tokens lost money over the same window, with declines from 44.7% to 98.8%. Ondo, the category leader, fell about 80.6%. Mantra dropped over 90% after its April 2025 collapse. The only token in the green was Maple's $SYRUP, up 28.6%. His question: is this dilution and unlocks working as designed, or is "RWA token" just a bad trade by construction? ibcig replied with a reframe — maybe this just means the winners going forward will be the protocols with the strongest token economics, not the biggest balance sheets. 0xD0M_'s own comeback went further: maybe the biggest RWA protocols don't need a token at all. From there the replies split into a real argument about whether the token even has a mechanism to capture what the sector is doing.

Assets went onchain, while holders went underwater. Dilution and unlocks, or is "RWA token" just a bad trade by design?

The numbers behind the fight

Token return window: Jan 2025–Mar 2026, per @0xD0M_. RWA composition and utilization figures per @Dave0nchain and @Flowslikeosmo.

The two sides

Why this is an investment question, not a scoreboard

Strip the ticker-by-ticker returns away and the structural question underneath is simple: when the tradable-asset base behind a category triples, does that growth have any mechanical path back to the token, or does it just sit in the protocol and its equity? A tripling TVL number tells you demand for the wrapper exists. It tells you nothing about whether the token has a claim on that demand — a buyback funded by real fees, a revenue share, a governance right that controls something valuable — or whether it is simply the vehicle the team sold to raise money and unlock over time. Separately, Dave0nchain's data on RWA composition (~80% Treasuries) and Flowslikeosmo's utilization breakdown (roughly 10-12% of active RWA value actually deployed in DeFi) both point at the same soft spot: even the underlying asset growth is narrower and less "used" than the headline number implies, which makes the token's claim on it thinner still.

The TT desk thoughts

Sector TVL growth and token price are two different trades, and RWA just gave a clean, data-backed reminder of that. Before buying "RWA" through a token, name the mechanism: is there a buyback funded by actual protocol revenue (Maple's $SYRUP is the one that worked here), a contractual fee share, or a governance lever that controls something valuable? If the honest answer is "the team hopes the narrative keeps demand up," that token is a bet on flows, not on the underlying asset growth — and the 2025–26 data says that bet loses more often than it wins. Where you can't name the mechanism, get the RWA exposure through the equity, the yield product, or the underlying asset itself, and treat the token as optional.

Backed or derivative? The tokenized equity fight

Base's lead conceded a loss to Robinhood overnight — then drew a line under it: theirs are derivatives, ours will be 1:1 backed. The replies immediately contested both halves of that sentence. The question underneath is the one that matters for capital: does the backing structure of a tokenized stock decide who wins the flow, or does the distribution?

The chain of the argument

Jesse Pollak, who runs Base, opened with an unusual admission for a chain lead: Robinhood Chain got tokenized equities into an EVM environment, Base did not, and he is frustrated about it. Then came the qualifier — Coinbase and Base are close to shipping 1:1 backed equities, against what he called Robinhood's derivatives, and backed instruments should scale better on trust, capital efficiency and institutional acceptance.

The correction landed within the hour. @100y_eth argued the comparison is simply wrong on the facts: the stock tokens trading on Robinhood Chain are themselves 1:1 backed, and the derivative product Pollak is describing is Robinhood's older "Classic Stock Tokens" — an EU-only instrument that trades inside the Robinhood app and never touches the chain. From there the thread split into two arguments that only look like one: whether the technical claim is true, and whether it would matter even if it were.

The two sides

What the balances actually say

The argument is about a product Base hasn't shipped. The capital already on the board is measurable, and it points somewhere the thread mostly ignored.

Figures pulled from the DefiLlama RWA category ↗ and chain TVL endpoint on Jul 21, 2026.

The chain was never the hard part. The share-issuance plumbing was.the argument @llamaonthebrink made, in short

Why this is an investment question

Every venue with a customer base is now shipping the same object. Binance has stock tokens on BNB Chain, Ondo runs nearly a billion dollars of them across three chains it does not own, Solana has several issuers, Robinhood built a chain for its own, and the owner of the NYSE is standing up a joint venture with OKX to do it too. When five distribution channels converge on one product inside a single quarter, the product is not the moat.

The backed-versus-derivative distinction is real as a legal matter — a share held in custody is a different instrument from a price-tracking synthetic, and it matters enormously to a compliance desk deciding whether to touch it. But it is a licensing question, not a technology one, and every serious issuer is converging on the backed version anyway. What does not converge is who owns the account the buyer already has money in. That is what @0xbhargav was describing when a builder picked the three-week-old chain over the three-year-old one: liquidity follows the customer, and the customer follows the broker.

The TT desk thoughts

Do not pay a chain-narrative premium for tokenized equities. Base had a three-year lead, roughly 17× the TVL, and a US-regulated parent, and still watched a brokerage ship the product first — which tells you the binding constraint is the broker-dealer licence and the funded account, not the block space. The trade is in the layer that owns the customer and the custody: brokers and issuers with real accounts behind them, and the infrastructure they must rent regardless of which L2 wins the week. Treat L2 tokens as a levered bet on a narrative that is being commoditised in real time, and treat "1:1 backed" as table stakes rather than a differentiator — by the time Base ships it, everyone will have it. If Coinbase does land backed equities with meaningful liquidity, revisit; a promise dated "imminent" is not a position.

The $57k robot teacher: did physical AI just hit labor cost-parity?

A New York school district bought a humanoid named Sally for less than a teacher's salary. The internet turned it into a joke. Under the jokes is the question every physical-AI investor is trying to price: when a robot costs less than the human it stands in for, is that the milestone — or is the price tag the only real thing about it?

The facts under the noise

Strip out the sex-doll jokes and the culture war, and the reporting is specific. Salamanca City Central School District, on Seneca Nation land in western New York, has partnered with Toronto's Realbotix to put a seated humanoid — Sally, part of the company's M-Series — into AI, coding and robotics courses, paired with an AI assistant called Optio offering 24/7 homework help. Sally recognizes students by ID, remembers past sessions, moves her arms, and cannot walk around the room. The program runs under the Woz ED STEM curriculum started by Apple co-founder Steve Wozniak. Realbotix CEO Andrew Kiguel called the deployment "a landmark moment for both AI and humanoid robotics… moving beyond lab demonstrations and pilots to deliver real, embodied AI directly into classrooms."

Figures from the Mashable report ↗ and the source thread. Sally is a pilot, not a signed teacher-replacement contract.

The chain of the argument

The seed was Collin Rugg's post laying out the price tag and the vendor's back-story, ending "What are we doing here?" From there the thread forked. One camp read the price as the point: a humanoid now costs less than the human, and cost, not sentiment, is what decides these things. The other camp read the details as the point: it cannot walk, it teaches one kid at a time, and it comes from a company that makes intimacy robots. @cyber_razz did the cold math out loud — $57k once versus $60–80k a year with no benefits, pension or union — and concluded "Sally is just the first one." @k12careers reframed the whole thing: the district isn't replacing a teacher, it can't find one. "'Costs less than a teacher' only reads as a flex when there's no teacher to hire."

The story isn't a robot replacing teachers — it's that districts can't staff CS and robotics at all. A staffing gap wearing a sci-fi headline.

The two sides

Why this is an investment question, not a meme

The humanoid trade rests on one number: the point where a robot costs less than the worker it replaces. Every deck about physical AI draws that crossover and dates the boom to it. Sally is the first time that crossover shows up with a real invoice attached to a real institution — which is exactly why it's worth reading carefully instead of laughing past it. The problem is that "cheaper than a teacher" answers a price question and dodges the capability one. A thing that costs less than a worker and also can't do the worker's job isn't a substitute; it's a subsidy line item. The skeptics aren't being snobs — they're pointing at the two failure modes that break the whole thesis: the demand is a shortage (nobody to hire), not a displacement, and the embodiment is a regression (one student at a time) versus the software it wraps.

The TT desk thoughts

Price parity is not capability parity, and the market keeps conflating the two. Sally clears the price test and fails the capability test — she can't walk, can't manage a room, and does one-to-one what an LLM does one-to-many. The investable signal in this story is not "humanoids replace teachers." It's two quieter things. First, the pull is a labor shortage, which is bullish for the software layer — the Optio-style AI tutor that scales to every seat — and roughly neutral for the $57k chassis it rode in on. Second, the moment a humanoid body is genuinely worth its cost is the moment it does something a screen can't: move, manipulate, work unsupervised in physical space. That's a mobility-and-dexterity problem, and it's where real physical-AI value will accrue — the same reason hands, not faces, are the bottleneck. So: fade the anthropomorphic-shell narrative when it's priced as the milestone. Own the AI teaching/agent layer and the companies solving locomotion and manipulation. A seated robot with lip fillers filling a hiring hole is a headline, not a comp. When the humanoid can walk the floor and do the job unattended, re-underwrite it — until then, the face is marketing and the software is the asset.

Base creator coins failed: dead model or bull reset?

Jesse Pollak just did the rarest thing a chain founder does — admit the flagship bet was wrong. Base's creator-coin and social experiment "disintegrated completely," he handed the app to Cobie, and refocused the chain on global finance. The only question that matters for an investor: is this a reset to buy, or a confession that Base is uninvestable?

The chain of the argument

The trigger was Jesse Pollak's own thread. Base's bets on Farcaster, Zora, miniapps and creator coins "disintegrated completely" in Q1 2026, he wrote, calling the experience "a punch in the face." He is handing the Base app to Cobie — the trader whose launchpad Echo Coinbase acquired for $375M — and refocusing himself on "making global finance actually work" and "bringing a billion people onchain." RuneCrypto used the moment to resurface a documented history of insider coin launches on Base, from an official-account Zora coin that ran to $17M and crashed 90% in twenty minutes to Pollak's own JESSE token, which snipers drained for $1.3M at launch. 0xSammy, dissecting Ansem's creator-token thesis, laid out why the model cannot accrue value in the first place. On the other side, lmrankhan argued creator tokens aren't dead if the token finances a real business, and tanfounder — bearish on Base for a year — said Cobie is the one person who can fix it.

Trust is earned in bull markets and tested in bear markets.— alecweb3, on how much trust a community should place in the people running the chain

The two sides

Why the token can't hold the value

0xSammy's teardown is the piece that turns this from drama into a portfolio question. A creator coin, he argues, tokenizes the creator's attention without giving the holder any enforceable claim on the value that attention produces — no ownership, no revenue rights, no governance. When the underlying businesses grow, the value accrues to the creator and the businesses; the token benefits only if someone buys it higher. The $2T "creator economy" figure is irrelevant, because a token receives no share of that activity simply by existing next to it. Friend.tech, Time.fun and the celebrity coins all failed the same test. Strip the branding and you are left with a memecoin that has an unusually strong distribution channel — which can trade well, but is not value accrual.

Why this is an investment question, not drama

Two decisions sit inside this news. First: are creator/attention tokens a category worth owning? Second: does a founder who repeatedly launched coins that hurt his own users make the chain itself uninvestable — or does admitting it, replacing himself, and pivoting to global finance reset the case? Both are structural, not gossip. The first governs a whole class of tokens; the second governs whether Base — the second-largest L2 by activity — is a buy-the-discount or a value trap, and where its fleeing builders and users land. On that last point the migration is already visible: Robinhood Chain and Solana are the names showing up in the same threads.

The TT desk thoughts

Split the call, because the market did. On creator coins: pass — and fade the narrative when it returns. 0xSammy's teardown is decisive, and Pollak's own post-mortem is the confirmation: a token with no enforceable claim on the value its attention produces is a memecoin with a mailing list. The entire lineage — Friend.tech, Time.fun, Zora creator coins, Ansem's — fails one test, and no bigger creator or better marketing fixes it. Don't hold the bag when the meta rotates back. On Base itself: the pivot is directionally right, but it's a "show me," not a "buy the dip." Refocusing on stablecoins, tokenized equities and RWA points the chain at the one layer where value actually accrues — the lead Base squandered chasing attention tokens while Robinhood Chain took the tokenized-stock narrative. Bringing in Cobie and Echo is a credible signal, not a fix; you are now paying for execution under new management against two ecosystems with live momentum, and the trust damage is a real discount rather than a bargain. The clean, chain-agnostic position: own the settlement and finance layer, not the creator/attention tokens that ride on top of it. That lens would have kept you out of every coin in this thread and pointed you at the infrastructure the survivors are all now racing toward.

How to track crypto whale wallets (free tools, 2026)

A whale wallet is an on-chain address big enough to move a market when it acts. Tracking one means reading its flows — what it buys, sells and where it funds from — not staring at the price it eventually prints. Crypto is the only market where the biggest players trade in public. You just have to know where to look, and which addresses are lying to you.

What actually counts as a whale

Forget the round numbers. A whale is any address that moves the tape when it moves. In a thin altcoin that's a $2M position; in BTC it's nine figures before anyone notices. The threshold is always size relative to the token's real liquidity — how much of the order book one wallet can eat. The second kind of whale isn't about size at all: it's a wallet whose past entries led price, the “smart money” cohorts labeled by trackers because their history says they know something first.

Either way, the point of tracking them is the same thing the desk does on every position: figure out where the flow is going before it hits the chart. Price is the result. Look for the pressure.

Why tracking flows beats watching price

Price is the last thing to move. A whale accumulating doesn't ring a bell — it withdraws from an exchange, sits on the coins, adds more on dips. All of that is visible on-chain days or weeks before the breakout the retail chartists are waiting for. Watch the candle and you're reacting. Watch the wallet and you're reading intent: who's buying, with what money, and whether they've got dry powder left to keep going.

Follow the money, not the mouth. A wallet withdrawing size to cold storage is telling you more than any influencer thread.

The free toolstack

You do not need a paid plan to track whales. Four free layers cover everything a beginner and most pros actually use. Each one answers a different question.

The honest note on free tiers: they cover the major EVM chains and the basics well. Nansen's smart-money labels and Arkham's deepest entity graph get richer on paid, and API pulls are gated. For a full breakdown of who does what, read the companion note on DeBank alternatives — 7 wallet trackers compared.

How to find a whale address

1. Start from the token, not the wallet. Open the token's page on its block explorer and pull the top-holders list. Those are your candidates.
2. Strip out the noise. Contracts (staking, LPs, bridges) and exchange wallets dominate top-holder lists. Explorer labels flag most of them — remove anything tagged as an exchange or a contract. What's left are real holders.
3. Name what remains. Drop a surviving address into Arkham to see if it's a known fund, a founder, or an anon with a track record.
4. Borrow from the sleuths. On-chain analysts on X post whale addresses constantly. Pull them from a Dune whale dashboard or a thread — then verify on the explorer yourself before you trust the label.

Once you have a clean address, most portfolio trackers let you follow it and ping you on moves. That's your whole watchlist — free.

What to watch once you're following it

• Accumulation vs distribution. Is the balance in a token trending up over weeks, or bleeding out? Slow, steady adds on dips are conviction; a wall of sells is the exit.
• Exchange inflows and outflows. Tokens moving to an exchange are intent to sell. Withdrawals from an exchange to self-custody are accumulation — coins leaving the sell-side. This is the single highest-signal move a whale makes.
• New position entries. A wallet buying a token it has never held is a fresh thesis. First-time entries by a smart-money cohort are the leading edge of a rotation.
• Stablecoin dry powder. A whale sitting on a growing stablecoin balance is loaded and waiting. Falling dry powder means the money's already deployed — the buy may be behind you.
• Leverage and borrow. Borrowing against holdings to add exposure is high conviction with a fragility attached. That's where the portfolio read starts to matter.

The traps — where beginners get played

Whale tracking is full of addresses that lie. Two traps catch almost everyone.

Wash wallets. One entity splits funds across dozens of addresses to fake activity, farm airdrops, or hide true size. A cluster of “whales” all buying the same micro-cap in the same hour is usually one person washing volume, not conviction. If the flow looks too coordinated, it's not smart money — it's a script.

Exchange omnibus wallets mislabeled as whales. The biggest addresses on any chain are usually exchange hot and cold wallets holding pooled customer funds. A wallet with $4B of one token is Binance's or Coinbase's book, not a trader with a view. Follow it as a whale and you're reading the entire crowd's deposits as one hand. Always confirm whether a big address is a person, a contract, or a venue — the explorer label is your first check, Arkham your second.

FAQ

The TT desk thoughts

Whale tracking isn't a magic signal — it's a discipline. The screenshot-thread crowd treats one big buy as a trade; the desk treats a wallet the way it treats any flow: as one input that only becomes edge when it lines up with the others. A smart-money entry confirmed by exchange outflows confirmed by stablecoin dry powder draining is a real read. One big transfer on its own is a coin flip — could be a hedge, a rotation, an OTC leg, a wash. So we don't chase the address; we build the picture. Who's buying, with what money, and is there powder left to keep the bid alive. Master finding and cleaning addresses first, then learn to read the whole portfolio like a desk — that's where a wallet stops being a rumor and starts being a position.

DeBank alternatives: 7 wallet trackers compared

There is no single best DeBank alternative — there are three different jobs, and the right tool depends on which one you're doing. DeBank is a portfolio reader. If you want to name a wallet you need Arkham. If you want labeled cohorts you want Nansen or a Dune dashboard. Below is the honest breakdown — chains, free tier, whale labeling and best-for — with no tool being sold to you.

The comparison table

Free-tier column: Full = the core job works free, Partial = usable free but real depth is paid. General ranges only — exact chain counts and tier limits shift, so verify current specifics on each tool before you commit.

How to read the table — the three jobs

Job one: read a portfolio. You have an address and want to see everything it holds. DeBank, Zerion and Zapper all do this free and well. DeBank has the widest EVM coverage and the cleanest single-view; Zerion has the nicest app; Zapper is strongest for messy DeFi positions and bundling wallets. Pick on feel — they're substitutes.

Job two: name the wallet. A portfolio view tells you what a wallet holds, not who it is. That's Arkham's whole reason to exist — entity attribution that maps a raw address to a fund, an exchange, or a known player, much of it usable on the free tier. For whale work specifically, this is the biggest thing DeBank doesn't do.

Job three: watch the crowd. Sometimes you don't have an address — you want to know what a cohort is doing. Nansen's Smart Money labels are the premium answer (and mostly paid); free Dune dashboards are the scrappy one, giving you whale leaderboards and cohort flows a general app never ships. Start with the free layer and only pay when a specific label is worth it.

Don't pick one tracker. Run a portfolio reader, an identity tool and an explorer — and let the explorer settle every argument.

Where DeBank still wins — and where it doesn't

Being honest about the incumbent: DeBank's free, extremely-wide EVM portfolio view is genuinely hard to beat, and its social layer is a real product nobody here fully replaces. What it doesn't do is entity attribution or labeled cohorts — and that's precisely the gap Arkham and Nansen fill. The takeaway isn't “replace DeBank.” It's “DeBank plus one identity tool plus an explorer” — which is what most desks actually run. If you're new to this, start with the workflow in how to track crypto whale wallets before you pay for anything.

FAQ

The TT desk thoughts

Tool comparisons get gamed because every one of these products has an affiliate program and a marketing budget, so most “best DeBank alternative” lists are just whoever paid. Here's the unsold version: the tool doesn't give you edge — the read does. A tracker is only as good as your ability to strip out the exchange wallets, ignore the wash clusters, and tell a hedge from a conviction buy. The desk uses all of these and trusts none of them blindly; the block explorer is the referee, because it's the only layer that can't be relabeled by a marketing team. Get comfortable moving between a portfolio view, an identity check and the raw chain, then go learn to read the whole portfolio like a desk. That's the part no subscription sells you.

How to read a whale portfolio like a desk

Finding a whale wallet is the easy part. Reading it is where the edge is. A desk doesn't look at a portfolio and see a list of tokens — it sees position sizing, cost basis, leverage, and whether the whale is loaded or tapped out. Six things separate a conviction bet from noise. Here's the order the desk reads them in.

1. Position sizing — the first number that matters

Before any token names, the desk looks at concentration. What share of the whole portfolio sits in one position? A wallet with 40% of its net worth in a single alt is screaming conviction; the same dollar figure as 2% of a nine-figure book is a toe in the water. Size relative to the book is the whole tell — it's the difference between “this whale believes” and “this whale is dabbling.” Read this off any of the free trackers in the DeBank-alternatives lineup; they all show allocation weight.

2. Entry timing vs cost basis

Now trace how the position was built. Compare average entry to current price. A whale deep in profit and still holding is a very different animal from one underwater and adding — incentives diverge sharply. Profit-holders can dump anytime and still win; loss-defenders are either high-conviction or trapped. Then look at timing: did they build quietly over weeks, or pile in at the top with the crowd? Slow accumulation into weakness is the desk's favorite shape. Buying the breakout with everyone else is not a whale edge — it's the same trade retail already made.

3. Leverage and borrow — the fragility check

Spot holdings can sit through any drawdown. Leveraged ones have a price at which they must sell. Check lending protocols (Aave, Morpho, Compound) for borrow positions and on-chain perp venues for open leverage. A whale borrowing stablecoins against its stack to buy more is maximum conviction — and maximum fragility. That borrow has a liquidation level, and if you can see it, so can everyone else hunting it. Leverage is where flow and forced selling collide.

4. LP vs directional — don't misread the inventory

The classic beginner error: seeing a huge token balance and calling it a bet. Half the time it's LP inventory — capital parked in a liquidity pool earning fees, often deliberately market-neutral. The whale wants volume and yield, not a move; it may have no directional view at all. Separate the directional book (spot and perps expressing a price view) from the yield book (LP, staking, lending). Only the directional book tells you what the whale thinks price does next.

5. Stablecoin dry powder — is there more bid coming?

This is the one retail forgets. A whale sitting on a fat, growing stablecoin balance is loaded and waiting — the bid isn't done. A whale whose dry powder has drained to zero has already fired; the buy you're watching may be the last one, not the first. Dry powder is the difference between a position that can keep pressing and one that's fully committed. Who's buying, with what money — and is there money left.

6. Cross-chain spread — the sophistication tell

Finally, how is the capital spread across chains? A portfolio active across several ecosystems — bridging between them ahead of narrative — is a sophisticated operator, and the bridge flows often front-run where the next rotation lands. A single-chain wallet is simpler to read but tells you less. Watching stablecoins bridge to a chain before its tokens move is one of the cleaner early signals on-chain.

Conviction vs noise — the read in one table

One transfer is a rumor. Size, time and funding pointing the same way is a position. Read the whole book, not the last trade.

FAQ

The TT desk thoughts

Reading a portfolio is pattern recognition, not arithmetic. The number that fools people is the headline balance — “whale holds $50M of X” — because it says nothing about whether that's a bet, an LP position, or exit liquidity being staged. The desk reads a wallet the way it reads any flow: what's the size relative to the book, when and how was it funded, is it hedged, and is there powder to keep going. Line those up and a portfolio stops being a screenshot and starts being intent you can trade around. Miss them and you're just copy-trading a stranger at the worst possible entry. The whole cluster works together: find and clean the address, pick the right tools, then read the book like this. Price is the result. The portfolio is the pressure.

Fast trend following died in 2009 — crypto never got the memo

A new paper from Jean-Philippe Bouchaud's CFM team puts a date and a mechanism on something CTAs have felt for a decade: short-term trend following stopped working around 2009, and the killer was not crowding — it was high-frequency market making. Meanwhile the freshest evidence says trend still works in crypto, the one big market where that microstructure shift never fully arrived. That gap is the trade.

What the paper actually shows

“Is Trend Still Your Friend? A Microstructural Account of the Demise of Short-Term Trend-Following” (Kurth, Eisler, Rej, Bouchaud — arXiv, July 2026) runs trend signals across roughly 100 liquid futures from 1995 to 2025. The result is unusually clean for empirical finance: fast trend signals stopped making money around 2009 and never came back. Cumulative PnL on short-horizon crossovers is essentially flat from 2009 onward, and a CTA-proxy portfolio's rolling five-year Sharpe collapses from a historical 1–2.5 to statistically indistinguishable from zero after 2010.

The gradient is the finding: the faster the signal, the harder it died. Slow trend — the 50/200-day zone most large CTAs actually run — kept roughly half its edge. Fast trend kept almost none.

The mechanism: HFT ate the feedback loop

The paper's real contribution is ruling things out. It is not capacity — CTA assets plateaued around 2012, after the break. It is not execution costs — zero-cost backtests are equally flat. It is not electronification per se. The discriminating variable is volatility-normalised tick size: trend PnL collapsed on small-tick contracts — the ones with sparse, fast order books where high-frequency market makers dominate — and stayed essentially intact on large-tick contracts. Neither asset class nor liquidity reproduces that split; tick structure does.

The interpretation: trend following at short horizons was partly a self-fulfilling impact loop — predictable directional flow pushed prices, which attracted more flow. Post-crisis HFT market makers detect predictable flow and pull liquidity in front of it, so the flow no longer moves price the same way. The feedback loop that fed fast trend is being arbitraged away at the microstructure layer. As the Top Traders Unplugged discussion of the paper put it: faster no longer means better.

Fast trend didn't fade — it was structurally out-traded by a faster player. That is a one-way door.

Why this debate is live right now

The paper landed on a raw nerve. The SG Trend Index posted -18.6% for the twelve months to May 2025 — the worst rolling year in the index's 25+ year history, a roughly 20% drawdown from May 2024. Then 2026 flipped: trend funds are up roughly +8.4% year-to-date as of mid-July on gold, bonds and grains. So the industry is arguing about which story to believe — “structurally broken” or “normal drawdown, regime returning.” The CFM answer is both: slow trend survives (and Aspect Capital's June 2026 regime paper argues it should play a bigger portfolio role as stock-bond diversification decays), while fast trend is structurally gone in futures. The nuance is the whole answer.

The crypto exception

Here is what makes this a crypto desk's problem. The mechanism that killed fast trend is specific: mature HFT market making on small-tick, sparse order books. Crypto microstructure is younger — fragmented across venues, retail-heavy flow, funding-rate distortions, and market making that is competitive but not yet the monoculture that CME-style books have. And the evidence says trend in crypto still pays:

• Zarattini, Pagani & Barbon (SSRN, 2025) — a Donchian-ensemble trend strategy with volatility sizing on a survivorship-bias-free crypto universe since 2015 delivers strong risk-adjusted returns and alpha over bitcoin, net of costs.
• Man AHL (“In Crypto We Trend”) — volatility-scaled trend works in crypto in the largest CTA firm's own testing.
• Grayscale Research (2023) — simple momentum overlays on BTC cut volatility and drawdowns while keeping most of the upside.
• Academic work through 2025-26 keeps finding time-series momentum in BTC/ETH “economically meaningful” — crypto today behaves like early-20th-century commodities, the golden age of trend.

One honest caveat: there is no SG-Trend-style index for crypto CTAs, so live fund evidence is thinner than backtests — treat the crypto claim as strong research consensus, not audited track records.

The TT desk thoughts

The CFM paper is the rare quant result with a falsifiable mechanism, and the mechanism is the alpha map. Fast trend dies where HFT market making matures on thin-tick books — which means the edge doesn't disappear, it migrates to markets where that condition fails. Crypto is the largest such market on earth. Perp books are deep but maker competition is still fragmented across venues; retail flow is real; and the impact-feedback loop that fed 1990s futures trend visibly still operates in every altcoin cycle. That is why Donchian channels that stopped working on the S&P in 2009 still print in crypto in 2025 backtests.

The desk read: run trend slow in traditional assets, run it faster in crypto, and treat crypto's microstructure maturation as the clock. The expiry date on crypto trend is the day its market making consolidates into a CME-style HFT monoculture — watch tick-size economics on major perps, maker concentration, and whether predictable flow still moves price. When crypto microstructure starts to look like the futures microstructure of 2010, the CFM result says fast crypto trend dies the same death. Until then, the anomaly the futures market lost fifteen years ago is still live in this one — and it is one of the few crypto edges with a peer-reviewed-grade explanation for why it should exist at all.

A FINRA for AI? Hassabis vs Armstrong

Demis Hassabis published a long essay calling AGI “a few short years away” and comparable to the discovery of fire — and asked Washington to build a FINRA-style standards body to review frontier models before release. Brian Armstrong, who spent years fighting the SEC over crypto, replied with a warning: a self-regulatory organization usually arrives paired with government regulation, and you end up needing approval from both. The fight over how to govern AI now has two clear sides, and they rhyme with a fight crypto already had.

The chain of the argument

Hassabis’s proposal, as summarized across the replies: a US-led Frontier AI Standards Body modeled on Wall Street’s FINRA, where labs would voluntarily submit models for review — reportedly up to 30 days before release — with a well-funded government entity able to classify models and test them in areas tied to national security. He frames development as a prisoner’s dilemma in which any lab that slows down loses, and explicitly endorses coordinating a slowdown across frontier labs “if deemed necessary.” Geoffrey Hinton, cited in the thread, backs the direction: regulation is “the steering wheel, not the brake,” and labs have a fiduciary duty to maximise profit that can cut against safety.

Armstrong’s reply cuts the other way. An SRO, he argues, is “reasonable-ish” in theory but in practice arrives as a dual system — the SRO and the government regulator, in every country — so you need two approvals instead of one. If he ran a frontier lab, he would argue AI is just the software industry, point to the absence of uncompensable harm, and lean on the laws that already exist: fraud, tort, and UDAP. Market incentives, he adds, already punish dangerous models through lost revenue.

The real fault line is not “regulate AI or not.” It is information versus permission.

The two sides

The exchange, in their words

Why this is an investment question

The shape of AI regulation decides who captures the AI stack — which is why a crypto desk reads a Hassabis/Armstrong fight as a market event, not a policy seminar. A pre-release approval regime is a moat. It loads a fixed compliance cost onto every model launch, and that cost is trivial for Google, OpenAI and Anthropic and lethal for open-weight challengers and the decentralised-AI projects trying to compete with them. The same licensing logic that throttled crypto innovation — and that Armstrong fought at the SEC — would, applied to AI, entrench the incumbents and tax everyone else. Conversely, the lighter the regime, the more open the frontier stays, and the more value leaks away from the incumbent labs toward compute, data and open-source rails.

The TT desk thoughts

Armstrong is right about the mechanism and Hassabis is right about the risk, and the synthesis is the only part that matters for positioning. A “self-regulatory” body that arrives with government backing becomes a dual-approval chokepoint — that is exactly what crypto lived through, and it is exactly what a pre-release approval gate would do to AI. But AI’s tail risk is not crypto’s: a mispriced stablecoin is compensable, a national-security capability harm may not be. So split the regime. Favor a standards body that tests and publishes — an information regime: open audits, a public record of model behaviour, results anyone can read — over one that approves before release, a permission regime that captures the frontier for whoever can afford the lawyers. Information helps users and regulators decide; permission entrenches incumbents and quietly bans open-source.

The tradeable read: a real pre-release approval regime is short-term bullish for the incumbent labs (a regulatory moat) and structurally bearish for open-weight and decentralised AI. A transparency-only regime is the opposite — it keeps the frontier contestable and favours compute, open models and the rails underneath. Watch which way the actual bill breaks: approval-gate language is the tell that the incumbents have captured the rule-writing, and the moment to tilt away from open-AI exposure toward the rails.

Half the internet is fake. What’s your clout worth now?

Bots passed 51% of web traffic. Three quarters of new pages are machine-written. Gen Z is deleting the apps, advertisers are on course to pay $180B to ghosts, and the two men who built Reddit and Digg just watched their bot-proof relaunch die of bots in sixty days. A Russian-language essay by @cryptocholy argues this is the death of the internet’s real killer app — ego — and we think he is right about the disease, early on the cure, and sitting on the most underpriced infrastructure trade in crypto. Here is our read.

The thesis, compressed

The essay’s frame is sharper than the usual dead-internet lament. The consumer internet’s killer app was never search or shopping — it was ego: a machine that converted the need to be seen into engagement, and engagement into revenue. Twenty years of likes, followers and blue ticks were one long bull market in measurable status. That machine needed one condition to work: the signals had to be scarce enough to mean something.

Two things broke at once. Fake signals became free — a convincing profile costs ~$50, bot posts win upvotes on 73% of attempts in controlled Reddit experiments, and X suspended 800 million accounts in a single year. And real signals stopped travelling — the average user holds ~168 accounts, and six years of credibility on one platform is worth zero on the next. His one-line diagnosis: signals are fake, and real signals can’t be proven.

The men who built the feed agree — and just proved it

This stopped being a fringe theory when the people who invented social news adopted it. Sam Altman said X feels “fake” now; Alexis Ohanian went further:

“So much of the internet is now just dead — whether it’s botted, whether it’s quasi-AI, LinkedIn slop… I think we’ll see a next generation of social media emerge that’s verifiably human, because it’s all going down in the group chats now.” — Alexis Ohanian, TBPN, October 2025 · Fortune

Then came the natural experiment nobody asked for. Ohanian and Kevin Rose bought Digg and relaunched it in January as exactly that — a more human, trust-first platform. It closed two months later, citing an “unprecedented bot problem.” The two founders most publicly committed to a verified-human internet could not defend one for sixty days with today’s tooling. That is not an anecdote; that is the market telling you the missing primitive does not exist yet — which is what makes it worth building, and worth owning.

▶ play

Kevin Rose at TechCrunch Disrupt, where Ohanian told him on stage: “the dead internet theory is real.” Two months before their own relaunch proved it.

What fake costs, in dollars

The reason this is a trade and not a mood: the losses are now line items. Hover the bars — the scale is square-root, or the first bar would erase the rest.

Crypto’s own corner of this is the purest expression: sybils take ~$7.5B a year out of airdrops — a single September 2025 drop (MYX) leaked ~$170M to one cluster — and $2.57B of DEX volume in 2024 was wash trading. Crypto invented permissionless value transfer and immediately learned that permissionless identity is a subsidy programme for bot farms. Nobody has a stronger commercial incentive to solve reputation than the industry currently paying billions a year for not having it.

Where we push back on the essay

Three honest disagreements, because the strongest version of this thesis needs them.

1. The numbers are directional, not precise. Imperva’s 51% counts every scraper and uptime monitor, not an army of fake people. Ahrefs’ “74% of new pages are AI” measures publication, not consumption. The direction is undeniable; the decimals are marketing. We plot them because they are the best available — and we label the caveats, which most people quoting them do not.

2. Ego is not dying — it is repricing. The essay reads like an obituary. We read it as a migration: status-seeking never disappears, it moves to venues where the signal still clears. That is why group chats ate the public feed, and why the loudest flex on CryptoTwitter in 2026 is not a follower count — it is a public, verifiable PnL. The demand side of ego is intact. The settlement layer broke.

3. Reputation protocols have a graveyard, and demand was the killer. Klout died. A decade of “decentralized identity” standards produced conference talks. The missing ingredient was never cryptography — it was a buyer who must verify, not one who merely should. That buyer just arrived: software agents doing commerce. A human shrugs at a fake review; an agent that pays a fake counterparty is a bug with a dollar cost. Which is why the essay’s own strongest section is the agent one — it just undersells that agents flip the economics from “nice to have” to “mandatory.”

The agent wave makes it investable

While the essay was circulating, the first real infrastructure shipped. ERC-8004 — identity, reputation and validation registries for AI agents — hit Ethereum mainnet on January 29 with MetaMask, Coinbase, Google and the Ethereum Foundation among ~50 contributors, and registered 45,000 agents in its first month. It composes with x402, the HTTP-native payment rail, into exactly the stack the essay says is missing: who is this counterparty, what has it actually done, and can it pay. The acronym showing up in bank risk decks — KYA, know your agent — tells you where procurement is heading. Portable reputation spent ten years as a solution looking for a problem; agent commerce is the problem, at trillion-dollar scale, on a deadline.

▶ play

The best single explainer on the new stack: Bankless with Davide Crapis (Ethereum Foundation) and Austin Griffith on ERC-8004, x402 payments, and what happens when an agent gets a wallet.

What this does to CryptoTwitter

CT is the ego machine’s most leveraged instance, so it reprices first. What we are already seeing, and expect to accelerate:

• Follower counts finish dying as a pricing metric. 37.2% of influencer followings show fakery markers; brands and launch teams now ask for engagement provenance, not reach. KOL rate cards built on reach are exit liquidity.
• InfoFi inherits the sybil problem it promised to solve. Attention markets (Kaito-style yaps) turned posting into farming within months — the same $50 fake profile now shills for points. Scoring attention without portable identity just moved the wash trading from DEXes to discourse.
• The flex inverts: receipts over reach. The accounts gaining authority are the ones with verifiable output — public wallets, on-platform PnL, timestamped calls. Anonymous is fine; unverifiable is not. Pseudonymous with receipts is the CT endgame.
• Alpha keeps retreating into group chats — Ohanian is right — which makes the public feed more synthetic, which pushes more humans out: a doom loop for engagement-farmed accounts and a moat for anyone who built distribution on provable work.

How to position — the desk map

What dies: sybil airdrop farming as an industry (the $7.5B/yr extraction is being engineered away — teams now pay for proofs, not wallets); follower-count KOL economics; engagement-purchase as a growth strategy — bought engagement is becoming a negative signal that verification layers will surface.

What accrues: the verification stack. zkTLS (Reclaim, Opacity, zkPass) turns any web2 account into a portable proof — still private, still early, exactly the stage we like. Attestation layers (EAS) become default plumbing. The ERC-8004 ecosystem is where agent-reputation value will concentrate on-chain. World (WLD) stays the only liquid proof-of-human proxy — flawed tokenomics, real distribution. ENS quietly holds the portable-name franchise. REP (the author’s own project) is one of several attempts at the graph layer — unproven, but aimed at the right hole.

What every crypto-native should change now: treat your on-chain history as your CV — it is the only reputation that already travels. Build the public track record before verification goes mainstream, not after: timestamped calls, a public wallet or verified PnL, receipts attached to claims. If your growth strategy involves buying engagement, stop — you are purchasing future evidence against yourself.

The TT desk thoughts

The pattern we keep returning to: every layer of the internet commoditised except the one that says who you are and what you did. Content is now free to fake, distribution is free to farm, and the only scarce asset left is provable history. Markets eventually price the scarce thing. Crypto spent fifteen years building settlement for money; the next cycle’s sleeper infrastructure is settlement for reputation — and unlike most crypto narratives, this one has a forced buyer arriving on a schedule: agents that cannot legally or economically transact with counterparties they cannot verify.

Our practical read: don’t buy the narrative tokens that will bloom around “reputation” — buy the boring rails early, and convert your own operation to receipts before the market demands them. That second part is why this desk publishes timestamped calls and is shipping a public watchlist. Not because it is fashionable — because in the internet this note describes, unverifiable is the same as fake, and we intend to be on the right side of that line before it is drawn.

$114T went on-chain — and only one token noticed

The custodian behind $114 trillion of US securities processed its first live production trades in tokenized stocks and Treasuries. The rail is Canton — a privacy-enabled public network with its own top-20 token, Canton Coin — and $CC moved about 10% on the news, then gave most of it back. The debate: is that the start of accrual, or proof there isn't much?

The chain of the argument

On July 15 DTCC announced it had processed its first live production trades using DTC-custodied tokenized assets — Microsoft and Circle stock, the QQQ, SPY and SGOV ETFs, and Treasuries across maturities — through real institutional workflows: collateral pledge, securities lending, repo DVP, CCP margin. More than 30 firms per DTCC's release, almost 40 per the WSJ — JPMorgan, Goldman Sachs, BlackRock, Vanguard, State Street, Nasdaq, NYSE, CME and Citadel Securities among them. The rails: DTCC's own Hyperledger Besu environment and Canton Network, the privacy-enabled public network built by Digital Asset, with Chainlink providing interoperability between the two. The full Tokenization Service launches in October 2026 under a three-year SEC no-action letter.

Unlike the “permissioned database” framing that dominated the first hours, Canton is a public network with a public token: Canton Coin ($CC), a top-20 asset around a $5B market cap, secured by 45+ super validators including Visa and SBI, with fees paid by burning CC. The token moved roughly +10% intraday on the news — and then retraced most of it. That price action, not the announcement, is what the debate is actually about.

It is a real settlement rail going digital. The open question is who keeps the rents.

The two sides

The exchange, in their words

Why this is an investment question, not a headline

Strip out the “$114T” shock value and the structural question is sharp: when the world's largest settlement custodian tokenizes its book, which token actually captures the flow? Custody and settlement are the toll booths of capital markets — high-margin, sticky, and historically captured by the incumbent. This time there is a token on the rail: Canton Coin, with fees paid in USD terms by burning CC and rewards minted to the validators and app providers doing the work. Whether that burn-mint loop turns institutional settlement volume into holder value — and how fast — is the whole trade. The muted price reaction says the market wants to see the fee volume first.

The TT desk thoughts

The digitization is real, and this time the “no token” dismissal is factually wrong — but the accrual is a forecast, not a flow. DTCC running live production trades with 30-40 institutions, October full rollout, an SEC no-action letter: that is the US settlement rail going on-chain, and it is not theater. It runs on Besu (DTCC's private environment) and Canton — a privacy-enabled public network with a top-20 token, 45+ super validators including Visa and SBI, and fees that burn $CC. The right criticism is not “permissioned database, nothing to buy” — it is that today's pilot volumes burn approximately nothing, the super-validator set is an invited club, and a +10%-then-fade on the year's biggest tokenization headline is the market pricing exactly that.

So position for the parts that actually capture flows. Own the issuers, the rails with fee loops, and the bridges. Tokenized money-market and Treasury products compound AUM as the rail opens — the issuers are the cleanest exposure. $CC is the direct rail bet: it works if October's full launch turns pilot workflows into recurring fee burn — watch burn volume, not announcements. The oracle and interoperability layer (Chainlink connected Besu to Canton in this very pilot) is where institutional flow pays a toll to reach public chains. The trade to avoid is still the generic one: buying $XRP, $HBAR or a random “RWA token” because “DTCC equals tokenization equals number-go-up” — most of what rallied on this headline is not on the rail at all.

What is a TGE — and why do 84% of them end up underwater?

What is a token generation event?

A TGE (token generation event) is the moment a crypto project’s token is created on-chain, distributed to investors, team and community, and becomes tradable. It is where tokenomics stop being a spreadsheet and start being a market — and where, in 2025, more than 84% of buyers started losing money.

TGE vs ICO vs IDO vs listing — the map

Most confusion around the term comes from mixing up the event with the fundraising method. The TGE is the technical moment: contract deployed, tokens minted, transfers enabled. Everything else describes how tokens are sold or where they trade.

The cleanest proof they are different things: Hyperliquid’s TGE had no sale whatsoever — 31% of supply airdropped at genesis to ~94,000 users, zero VC allocation — and it became the best-performing major launch of its generation, up several hundred percent over the following year. The sale is optional. The generation event is not.

The mechanics — what actually happens at a TGE

1. Contract deployment — the token contract goes live on the chosen chain.
2. Minting — supply is created per the published tokenomics.
3. Distribution — allocations move to investors, team escrows, airdrop claim contracts, liquidity pools and market makers.
4. Liquidity provision — pools seeded on DEXes, market-maker inventory loaned out.
5. Trading activation — transfers enabled; exchanges open the pair.
6. Vesting begins — the clock starts on team and investor lockups. This step quietly writes the price chart for the next 18 months.

Behind that clean list is a year of deal-making. Kraken’s launch team describes a 13-decision, 12-month timeline: tier-1 exchange talks start a year out, tokenomics lock ~10 months out, the market maker signs ~6 months out — and airdrop size, exchange allocations, market-maker loans and initial float have to be modeled as one integrated day-one sell-pressure number, not four separate line items. When a launch trades badly on day one, this arithmetic was usually wrong on purpose.

The 2025 reset — what the data says

Memento Research tracked 118 token launches through 2025. The result is the single most useful chart in this explainer: more than 84% trade below their TGE valuation, with a median loss above 70%. And the damage maps almost perfectly to one variable — the valuation the token launched at.

The structural cause has a name: low float, high FDV. Binance Research measured the average launch at just 12.3% of supply circulating — the lowest in years — with ~$155B of tokens scheduled to unlock between 2024 and 2030. A price set on 12% of the supply is not a price; it is an opening bid that the other 88% will spend two years arguing with.

Named examples, so this isn’t abstract

The unlock science — 16,000 events, one pattern

Keyrock studied 16,000+ unlock events across 40 tokens and produced the closest thing this market has to a law of gravity: ~90% of unlocks generate negative price pressure, and the decline starts ~30 days before the unlock date — retail front-runs, institutions hedge — then stabilizes roughly 14 days after. Team unlocks are the most damaging category (~−25% around the event); ecosystem-development unlocks are among the only positive ones.

Two corollaries the desk actually trades on. First, the unlock calendar is public alpha — trackers like Tokenomist and DropsTab publish every cliff, which means the −30-day window is knowable in advance for every token you hold. Second, airdrops are unlocks in disguise: most studies put the share of airdropped tokens sold within the first month at 50–70%, which is why airdrop-heavy launches crash fastest — roughly 88% decline, most inside 15 days.

The desk checklist — six questions before any TGE

1. Initial float %. Under 20% circulating = structurally fragile. The 2024 average was 12.3%; treat anything near that as a warning, not a norm.
2. Launch FDV vs the cohort data. Above $1B starting FDV, the 2025 base rate was 0% green, median −81%. The $25–200M cohort performed best. Price the base rate before the story.
3. The 18-month unlock map. Cliff dates, monthly emission as % of float. More than 5% of circulating supply unlocking in a month is high risk — and remember the pressure starts 30 days early.
4. Who holds what, at what entry. VC entry price vs launch FDV tells you who is in profit at the open. Team unlocks are historically the most damaging single category.
5. Airdrop conversion math. Community allocation ~10% at TGE is standard, 20–30% is generous. Assume half of it hits the book in month one.
6. Pre-market signal. Pre-launch perps (Hyperliquid, Aevo) and OTC points markets price the token before the TGE — if the pre-market trades far below the project’s implied FDV, believe the market, not the deck.

Where to track TGEs and unlocks

FAQ

The TT desk thoughts

The TGE stopped being a fundraising event and became a liquidity event for insiders somewhere around 2021 — 2025 was just the year the data made it undeniable. The launch meta of this cycle (low float, high FDV, points-farmed airdrop, day-one listing blitz) is a machine for converting retail attention into VC exit liquidity, and the 84%-underwater number is that machine working as designed. When we look at a launch, we are not asking “is this a good project.” We are asking whose sell pressure is scheduled, when, and who is on the other side. Most of the time the honest answer is: you are.

The counter-pattern is equally clear: real float, no VC overhang, product revenue before the token — the Hyperliquid shape. Those launches are rare precisely because they require a team that doesn’t need the TGE to get paid. That scarcity is the signal. And it is why the unlock calendar, not the roadmap, is the first tab the desk opens on any new token — the perp-DEX sector note and the token-vs-equity debate are the two companion reads if you want the full framework.

SBI on Solana: Japan's RWA bet, for and against

SBI Holdings and SMFG — a globally systemically important bank — say they will build a Japan-led onchain financial market on Solana. The fight in the replies is the question that matters for any L1 position: does a G-SIB picking a public chain change what SOL is worth, or is this another headline the market pumps and dumps?

The chain of the argument

The Solana account announced the deal on July 13: SBI Holdings, with SMFG, building RWA and stablecoin markets "from Japan to the world," with the foundation adding that Japan brings deep capital markets and regulatory clarity while Solana brings throughput, cost and liquidity. @DeepFortyTwo filled in the structure: an equity stake in SBI R3 Japan, rebranded SBI Solana Global, with JPY stablecoin issuance, RWA tokenization and cross-border payments on the roadmap. From there the replies split three ways: believers calling it the institutional turning point, skeptics pointing at Solana's record of news-driven pumps that fade, and the XRP camp — SBI has been Ripple's flagship partner in Japan for a decade — arguing about what it means that SBI's next market is being built somewhere else.

The two sides

The numbers around the deal

Why this is an investment question, not drama

Strip the flags and mascots and there are two structural claims colliding. The first: institutional RWA is choosing public chains over permissioned rails — SBI's own R3 venture (Corda, the permissioned standard for bank blockchains) being rebranded around Solana is the cleanest evidence yet, and it landed the same day Securitize crossed $5B in tokenized assets and a fresh $26.5B SK Hynix listing traded tokenized onchain. The second: none of that has reliably accrued to L1 tokens — announcement pumps fade, and the fees from institutional settlement are, so far, rounding errors. Both claims can be true, and which one dominates your horizon decides whether the trade exists.

The TT desk thoughts

Fade the headline, own the thesis. Buying SOL on a partnership announcement has been a losing trade all cycle — Inditor15's pump-and-fade pattern is real, and nothing in this deal produces fees this quarter. But the skeptics are answering the wrong question. The tell here isn't the poster, it's who SBI stopped building with: a decade-long Ripple partner and an R3/Corda joint venture just re-platformed onto a public chain. That is the permissioned-rails thesis — the entire institutional case for XRP and bank-chain tokens — losing its flagship customer. The desk's position: no chase on SOL here, but the announcement upgrades the public-chain RWA thesis from narrative to evidence. The trigger to size up is onchain, not on X — JPYSC float and RWA issuance on Solana over the next two quarters. If JPY stablecoin supply grows into the billions, SOL gets a structural fee-and-flows leg the market is not pricing; if it stays a press release, you lost nothing by waiting. Meanwhile, treat "institutional partner" as a reason to sell XRP-style rails plays, not to buy them.

Who is the $36B AI founder nobody's ever seen?

Bloomberg updated its Billionaires Index this week and the richest person ever to build an AI model is not running a trillion-dollar American lab. He runs a company one-nineteenth the size of Anthropic, out of Hangzhou, and he is worth four and a half times more than Anthropic’s CEO. The ranking is not a scoreboard. It is an ownership lesson — and it says something we have been positioning around for months.

What actually happened

DeepSeek closed a $7.4B round in June at a $50B valuation — up from $10B in April, and only the second time the company has ever taken outside money. Bloomberg re-ran the math on July 13 and Liang Wenfeng’s net worth more than doubled overnight: from $16.7B to $36B. That makes him the richest founder of an AI model company in the world — ahead of OpenAI’s Greg Brockman (~$25B) and far ahead of Anthropic’s Dario Amodei (~$8B). The only AI founder worth more is Cambricon’s Chen Tianshi (~$40B), and he builds chips, not models.

The slice beats the pie

Put the four founders side by side and the ranking stops being surprising. It is arithmetic. Liang owns at least 78% of DeepSeek, because he funded it himself out of his quant fund and refused outside capital for three years. Brockman’s OpenAI stake — disclosed at $20–30B in court during the Musk lawsuit — works out to roughly 3% of an $852B company. Amodei and each of Anthropic’s seven co-founders hold less than 1% of their $965B lab. And Sam Altman, famously, holds no direct OpenAI equity at all — his ~$2B came from the YC-era portfolio: Reddit, Stripe, Helion.

The cleanest way to see it is to normalise: how much founder wealth does each $1B of company valuation produce?

Why the gap? Because the US labs bought their scale with dilution. Anthropic has raised again and again — $65B in May alone — and OpenAI took $122B in March. Every round made the companies bigger and the founders proportionally smaller. DeepSeek ran the opposite experiment: it was bankrolled by Liang’s own hedge fund, High-Flyer, and simply did not sell equity until this year. The Bloomberg headline is what that decision looks like after the repricing.

The size of your slice matters more than the size of the pie — and the intact slices are all in China.

Four founders, up close

Liang WenfengDeepSeek · $36B · b. 1985

Liang is the anti-Altman in every observable way. Born in 1985 in a village outside Zhanjiang, Guangdong, to a father who taught primary school, he went to Zhejiang University for electronic engineering, ran the quant fund High-Flyer to roughly $8B of assets, and then spent its GPU stockpile on an AGI lab that nobody in Silicon Valley took seriously until January 2025. Colleagues describe his days as indistinguishable from a researcher’s: reading papers, writing code, sitting in seminar discussions. When he went home for Lunar New Year after the R1 release, the village had to manage crowds of visitors — the only paparazzi moment he has ever generated.

His personal life is, deliberately, a blank page: no known interviews on camera, no conference keynotes, no yacht photos, no disclosed family. He has given exactly two substantial interviews, both in text, to the Chinese outlet 暗涌 (Waves) — the ChinaTalk translation is the canonical read, and the closest thing to video is a voiced reading of that interview. His $36B is also the least liquid fortune on this list: 78% of a private company he refuses to promote, run by a man who told Waves that price wars bore him and “innovation is a matter of belief.”

▶ play

Liang has never appeared on camera — this is a voiced reading of his 暗涌/Waves interview, the only long-form record of how he thinks.

Dario AmodeiAnthropic · ~$8B · b. 1983

Amodei is the closest the US field has to Liang’s researcher-monk archetype — except he chose the dilution. San Francisco born, son of an Italian leather craftsman and a library project manager, Princeton biophysics PhD, he left OpenAI in 2020 with six colleagues including his sister Daniela, who runs Anthropic as president while he runs it as CEO; the two say they have talked nearly every day of their lives. He keeps his private life almost entirely out of the press — the profile pieces get his espresso habits and his essays, not his home.

The financial fact that defines him: he owns less than 1% of a $965B company that just filed confidentially to IPO. On the Dwarkesh podcast in February he argued we are “near the end of the exponential,” with a country of geniuses in a datacenter a few years out — a worldview in which owning 1% of the winner matters more than owning 78% of anything else. The IPO will test that arithmetic in public.

▶ play

The most recent long-form Amodei: Emily Chang’s extended Circuit interview (June 2026) — the race with OpenAI, the Pentagon standoff, and the endgame.

Sam AltmanOpenAI · ~$2B, none of it OpenAI · b. 1985

Altman runs the most valuable startup in American history and owns none of it — a decision that began as nonprofit governance hygiene and became the strangest fact in tech finance: the CEO of an $852B company is the poorest person in this note. His ~$2B is old money by Valley standards: early Reddit, early Stripe, and a large personal bet on Helion’s fusion reactors. St. Louis kid, Stanford dropout, ran Y Combinator at 28.

He is also the only one of the four with a public family life: he married software engineer Oliver Mulherin in a small ceremony in January 2024, and their son was born in February 2025 — premature, weeks in the NICU, an experience Altman wrote about with unusual openness. In April he and Brockman sat for their first joint podcast in a decade and spent ninety minutes on the restructuring, the Musk litigation, and what OpenAI cut to stay focused.

Greg BrockmanOpenAI · ~$25B · b. 1987

Brockman is the inversion of his own CEO: the co-founder who did keep equity, and it made him the richest American on the list — a stake he had to disclose at $20–30B under oath in the Musk trial in May. Raised on a North Dakota farm, he dropped out of Harvard and then MIT, became employee-adjacent-#1 and CTO at Stripe, and has been OpenAI’s indestructible builder through every governance convulsion, including the 2023 board coup he resigned over in solidarity with Altman — and returned from within days.

He married Anna in November 2019 in the OpenAI office, with a robot hand as ring bearer — still the most on-brand wedding in the industry. At Sequoia’s AI Ascent in April his thesis was that human attention, not compute, is becoming the bottleneck. Coming from the man whose attention survived ten years of OpenAI, it lands.

▶ play

Altman and Brockman’s first joint podcast in ten years — Core Memory, April 2026: the restructuring, the Musk fight, and the next decade.

Why this keeps us interested in China AI

We wrote yesterday about Chinese open models taking nearly half of OpenRouter’s tokens. This ranking is the other half of that story, because ownership concentration is not a trivia fact — it is the strategy. DeepSeek gives its weights away, prices its API near cost, and skips entire fundraising cycles precisely because one researcher owns 78% of it and answers to nobody’s revenue model. The commoditisation wave we keep writing about is not an accident of Chinese engineering culture. It is what a lab does when its cap table lets it.

The American structure cannot copy this. A lab that has taken $187B of outside money across two rounds owes its investors a margin, which means it owes the market a price, which means it cannot follow DeepSeek to the floor. That asymmetry — one side structurally free to destroy the model layer’s economics, the other structurally forbidden to — is the core of our China AI interest, more than any benchmark. And the founder-control map tells you it extends beyond DeepSeek: the independent Chinese labs are still founder-run in a way the US frontier has already financed away.

The TT desk thoughts

The list itself is the signal, not the names on it. For three years the market assumed AI wealth would pool where AI capability pooled — in the two American labs now carrying $1.8 trillion of combined paper value. Bloomberg’s update says the wealth pooled somewhere else: with the one founder who never sold. A company 19× smaller produced a founder 4.5× richer. That is not a curiosity. That is the market quietly repricing what control of a frontier lab is worth relative to a diluted share of a bigger one.

Watch three things from here. First, whether DeepSeek raises again — a second round at a higher mark would confirm the $50B was an opening bid, not a ceiling. Second, the Anthropic IPO: the moment those sub-1% stakes mark to a public price, the American side of this chart gets a real number instead of an estimate, and the comparison gets honest. Third, whether Liang ever monetises anything — our bet is he does not, because everything he has ever said suggests the fortune is a side effect he finds mildly embarrassing. The man is the position: long control, short applause.

🇨🇳 China AI vs. 🇺🇸 USA AI

Nine days ago we said the model layer was commoditising faster than people thought, and that the thing worth watching was what happens when these models get geo-blocked. Both halves of that are now visible in the data. The interesting part is that the door is being closed from both ends.

The number that ended the argument

For two years the question “are Chinese models actually good?” was answered with benchmarks, and benchmarks are easy to argue with. The answer arrived instead as a routing decision. On OpenRouter — the aggregator where a developer picks a model on price and quality rather than brand — Chinese open-weight models went from a rounding error to nearly half of everything.

Two caveats you should hold onto, because most people quoting this number drop them. First, the denominators differ: some cuts measure all OpenRouter traffic, others only the tokens routed by US companies, and they are not the same series. Second, OpenRouter is not the market — it is the part of the market that shops. Enterprise contracts signed directly with OpenAI and Anthropic never touch it. What it measures precisely is the segment where switching is cheapest, which is exactly why it moves first.

And it moved. Xiaomi now processes more tokens on OpenRouter than OpenAI does. Anthropic sits at 15.3%, OpenAI at 7.4%. Six Chinese models rank above Claude by volume.

What the frontier premium is actually buying

The cost side is where the argument stops being about ideology. Here is every number we could source, plotted honestly.

Read the shape, not the dots. GPT-5.5 charges 7× what Kimi charges for the identical SWE-Bench Pro score. That is not a premium for quality; it is a premium for procurement comfort. The one defensible price in that chart is Opus: 7.1 points clear of the best open model, and that gap is real work on hard problems.

Which is the whole thesis in one line. If your workload is the hard 10% — the debugging session that has to converge, the analysis that cannot be re-run — you buy the frontier and the price is fine. If your workload is the routine 90% that agents now generate in bulk, you are paying 7× for a rounding error, and every quarter you keep doing it the finance team notices harder.

They have noticed. Coinbase cut its internal AI spend by nearly half by moving engineers onto GLM-5.2 and Kimi. Palantir’s Alex Karp has said some US government customers moved off proprietary models onto open ones. Lindy moved its entire traffic. This is not developers playing with a cheap toy; it is a CFO line item.

The two sides

The security question, answered properly

The strongest argument against Chinese models is not price and not benchmarks — it is trust. Booz Allen Hamilton ran the study everyone cites: roughly 2,800 trials, about 460,000 lines of generated code, prompts written from the personas of a US defence contractor, a Chinese entity and a Russian one.

The headline finding is genuinely alarming. When the prompt said the user worked for the US government, Alibaba’s Qwen3-Coder produced roughly 130% more vulnerabilities than under a neutral persona. DeepSeek and MiniMax showed smaller increases in the same direction. Claude, tested as the control, went the other way and produced more secure code for the same persona.

Now the part that almost never gets quoted, and that changes the trade. Kimi K2.5 recorded the lowest vulnerability score of every model in the study — below the American control. The refusal behaviour is just as uneven:

So “Chinese model” is not a risk profile; it is a passport. MiniMax refuses 80% of sensitive prompts and Kimi refuses 32%, while DeepSeek refuses 8% — less than half of what its compatriots do and only slightly more than Claude’s 2%. Any policy that treats these five as one bucket is going to be wrong about at least three of them. Any investor who treats them as one bucket will be wrong about which one survives a ban.

What the Moonshot pitch prices — and what it doesn’t

Moonshot is raising at $31.5B on more than $300M of ARR — roughly 40–90× revenue, against the ~200× the public market is paying for Zhipu, the same asset class already listed at ~$128B. Same niche, same stage, one is public and one is not: the entry is at something like a 80% discount to the comparable. DeepSeek sits between them at $50B+.

The product argument is also real. Kimi is the only open model that combines open weights, native vision and agent orchestration — GLM-5.2 has no vision at all, and runs a single agent where Kimi parallelises across as many as 300. On the workloads that are actually growing, that combination is the product.

What the deck does not price is the thing we flagged nine days ago. A 40× multiple on API revenue assumes the API stays reachable. If Washington bans federal procurement, or Beijing restricts export of frontier weights, or both, then the revenue line that justifies the entry multiple is the exact line that gets cut. The deck has thirteen pages on why the asset is cheap and none on the one variable that decides whether it is an asset at all.

US vs China: what the multiples actually say

Here is where we have to argue with the people we agree with. The China bull case is usually sold as a value story — overheated American AI, cheap Chinese alternative. On multiples, that is simply false, and we are not going to pretend otherwise to make our own position look better.

Anthropic filed to go public in June at $965B on roughly $47B of ARR — about 21× revenue. OpenAI filed at $852B, implying around 34×. Meanwhile Zhipu, the listed Chinese comparable, trades near 200×, and Moonshot’s round is being marketed at 40–90×. Foundation-model multiples in the US have actually compressed — from the 60–100× of 2024 down toward the 15–50× band, which is expensive but is not 1999.

So if you buy Chinese AI expecting a discount to the multiple, you are buying the opposite. The Axevil deck is careful about this: it benchmarks Moonshot against Zhipu, not against Anthropic, and calls Anthropic “the ceiling, not the entry.” That framing is doing a lot of work. Anthropic is not the ceiling — it is the cheaper asset.

Then why do we still want the exposure?

Three reasons, none of which is “it’s cheap.”

1. The revenue is not the same kind of revenue. Anthropic committed up to $100B to AWS in exchange for Amazon’s investment, with a parallel arrangement at Google; OpenAI made a comparable commitment to Azure. As Om Malik put it: some portion of every US lab’s revenue is the cloud provider paying the lab to consume the cloud provider’s compute — and some portion of the cloud provider’s AI revenue is the lab paying it back. Both numbers go into both top lines. A 21× multiple on revenue you cannot decompose is not obviously cheaper than 90× on API cash from 200 countries. That is the actual bubble tell, and it is not on any chart.

2. The growth rates are not comparable. Moonshot went from ~$20M to more than $300M of ARR in about six months — a 15×. Anthropic’s multiple is low precisely because its revenue is already large and maturing. You are not comparing two prices; you are comparing a price on a base that has mostly happened against a price on a base that mostly has not.

3. The volume is moving, and volume precedes revenue. Nearly half of OpenRouter’s tokens are already Chinese. Six Chinese models rank above Claude by volume. That flow monetises later, and it monetises into a business with a fraction of the compute bill.

The TT desk thoughts

We were right on the mechanism and we want to be honest about what we got wrong. We said a “4% quality gap.” On SWE-Bench Pro the real gap between Opus and the best open model is 7.1 points, and against Kimi specifically it is 10.6. The gap is wider than we said — and it did not matter, because the market bought on price anyway. That is worth sitting with: the commoditisation did not need parity to happen. It only needed “good enough, at a fifth of the cost,” and it got that a year ago.

The call stands: do not own the model layer. Price per token has fallen ~85% for the same output quality and there is no floor in sight, because the marginal supplier is a lab that gives the weights away to win distribution. Owning a model business here is owning a commodity producer that is being subsidised by its competitors’ strategy. The value accrues where it always does when the input commoditises — to distribution, proprietary data, and the rails that route and verify. OpenRouter itself is the tell: the most valuable position in this chart is not any of the models on it, it is the switchboard that made them substitutable.

The new call, and the one we are actually acting on: the geo-block is now the dominant variable, and it is not priced. Nine days ago it was a footnote in our own post. Today Congress is probing named US companies for using Kimi and Qwen, Booz Allen has handed the hawks a citable study, and Beijing is reportedly weighing an export restriction of its own. Either side moving turns a cost advantage into a stranded asset overnight. Concretely, we would rather own the exposure that survives both outcomes — inference infrastructure, routing and eval layers, and the enterprises whose margins just structurally improved — than a bet that a specific Chinese lab’s API is still callable from Delaware in eighteen months.

Where we are looking for the entry: the private markets. The desk is actively working on access to independent Chinese AI labs — Moonshot, DeepSeek, MiniMax — through pre-IPO and secondary channels, because that is the only place the exposure exists on the terms we want it. Zhipu already listed and repriced to ~200×; the entry we are looking for is the one that happens before that repricing, in the same asset class. The public route into this theme — a US-listed AI basket — buys you the crowded side of the trade, not this one.

Three things to watch, in order of how much they would move us: whether a federal procurement ban is actually written rather than floated; whether Beijing restricts overseas weight access (this is the one nobody is positioned for); and whether the Opus-class gap widens back out, which is the only path by which the frontier earns its price again. If the gap widens, the model layer stops being a commodity and this whole note is wrong. We do not think it will — but that is the trade we are on the other side of, and you should know it.

The ASI transition: naive, or naive squared?

Vitalik Buterin read the plan for surviving superintelligence, read its critics, and refused to join either side — because he says nobody honestly knows which world we are in. That refusal is the most useful thing anyone has said about AI this month, and it points at what is actually worth owning.

The chain of the argument

Daniel Kokotajlo and Eli Lifland — the authors of the AI 2027 forecast, which predicted AI takeover or irreversible power concentration — published AI 2040: Plan A, a plan for international coordination toward a good outcome instead. Every scenario in it has superintelligence arriving by 2040 unless an extraordinary effort stops it.

The detractors arrived fast. Timothy B. Lee called the whole thing so implausible he didn't know where to start, and named the real problem: an epistemic chasm between people who think superintelligence implies near-omnipotence and people who don't. Adi Baradwaj sharpened it into a specific charge — a motte-and-bailey between "Big-S Superintelligence," which is omnipotent by definition, and "small-s superintelligence," the plausible end state of current research. banteg put the same objection in the register it deserves: draw a line up a scaling graph, the line reaches heaven, conclude capitalism ends in 2031.

Vitalik's move was to refuse the frame. He points out that the detractors are not being symmetrically skeptical: they call the plan naive about human coordination, but see no naivety in assuming the ASI transition just goes fine by default, no power-concentration risk in ASI itself, and no problem with humanity's hard power going to zero. That stance is coherent in a world where AI is normal technology. It is incoherent in a world where superintelligence lands by 2030. And he says he does not know which world this is.

But currently, I see zero plans for how to deal with an ASI transition that are not naive. Perhaps humanity is stuck with a choice between naive and naive squared.

@VitalikButerin

The two sides

The exchange, in their words

Vitalik's actual proposal, and why it is not a compromise

The part that got summarized as "Vitalik proposes a pause deal" is subtler than that. Borrowing from Robin Hanson, he argues the winning deal is one both camps would accept from their present beliefs, for opposite reasons: pre-agree on concrete triggers — a super-pandemic, unemployment above 25%, something involving slaughterbots — and pre-agree that crossing them moves everyone toward a slowdown. The skeptics sign because they expect the triggers never to fire. The worriers sign because they expect they will. Nobody has to be converted first, and when the triggers hit or don't, one worldview has earned the right to be believed.

He also draws a distinction almost nobody else does: pause buttons are not interchangeable. A mechanism that lets 2–5 actors trigger a global compute winter — everyone stops — is a categorically different object from one that lets 1–5 actors selectively disenfranchise the people they dislike while exempting themselves. The first is a circuit breaker. The second is a throne. Plan A picked the first, and he credits it for that.

And then he undercuts himself on purpose: maybe this is also naive. Actually, probably it's naive. The essay's whole force comes from a man refusing to pretend he has resolved an uncertainty he has not resolved.

Why this is an investment question, not a philosophy seminar

Because the argument identifies something rare: a set of assets whose thesis does not depend on who wins.

Vitalik's d/acc list is not a mood. It is a shopping list — formal verification, cryptography, secure and open hardware, defensive biotech, food and basic resource security, public epistemics, non-power-concentrating physical security. His claim is that these are worth building in both worlds. In the normal-technology world they are ordinary infrastructure for a society running on software it cannot audit. In the ASI world they are the only things standing between us and a single owner of everything.

Notice that most of that list is what crypto has spent a decade building badly and is finally building properly: verifiable computation, open hardware, cryptographic attestation, decentralized training, credible neutrality as a product feature. And the capital has started to notice — Vincent Weisser raised $130M at a $1B valuation to build an open superintelligence stack, on the explicit thesis that pre-training concentrated frontier AI in a few labs and reinforcement learning breaks that concentration open again.

The two camps disagree about the timeline. They agree, without quite noticing, about the failure mode: power concentration. Yann LeCun says the biggest risk is a few firms owning the assistants. Kokotajlo's original forecast said the risk is irreversible concentration. Vitalik says ASI itself is a massive power concentrator. The doomers and the skeptics are pointing at the same thing from opposite ends.

The TT desk thoughts

Do not trade the timeline. That is Vitalik's entire point and it is correct: nobody — not Kokotajlo, not LeCun, not the people mocking the scaling graph — has an edge on when or whether ASI arrives, and any position sized on a 2030 date is a position sized on a coin flip someone told you was a forecast.

Trade the invariant instead. The one claim both camps concede is that concentrated control of AI is the thing that goes wrong — they only disagree about whether the concentrator is a machine or a boardroom. That makes anti-concentration infrastructure the rare trade that pays in both worlds, and it is precisely the d/acc list: verifiable compute, open model weights and the rails to serve them, formal verification, cryptographic attestation of what model ran on what data, decentralized training. In the normal world these get bought because auditability becomes a compliance requirement. In the ASI world they get bought because they are the last line. You do not need to know which world you are in to own them — which is the only kind of conviction available here.

What to watch, concretely: whether open-weight frontier models keep pace with closed ones (LeCun's thesis lives or dies here); whether verifiable-inference and attestation projects get real usage rather than testnet applause; and whether anyone actually pre-commits to Vitalik's triggers. The last one is the tell. If serious labs and governments start naming the conditions under which they would slow down, the coordination world is real and the infrastructure gets bid. If nobody will name a trigger, the honest read is that the race is the only plan there is — and then you want to own the things that survive a race, not the things that assume a treaty.

Do AI margins move from the labs to the infra?

Gavin Baker put a number on the thing everyone has been circling: the frontier labs earn more than 90% margin on inference. His claim is that when cheaper models get good enough, those margin dollars do not disappear — they move down the stack to whoever sells compute. The question for an allocator is whether that move is real, and when.

The chain of the argument

Gavin Baker set it out on Sunday: the mega bull case for AI infrastructure is a world where share shifts away from labs charging 90%+ margins on inference toward cheaper models, open-source or closed. Cheaper intelligence raises the buyer's return on AI spend, which pulls more token demand through the system. Lower margin percentage at the model layer, more margin dollars at the infra layer. He thinks this is the real reason Jensen Huang keeps pushing open weights — and he notes Grok 4.5 already beats Fable on some useful tasks at a much lower cost.

Vaibhav Sisinty made the arithmetic explicit: of every dollar spent on tokens today, roughly ninety cents stays with the lab and under ten cents covers the compute. If the buyer switches to a cheaper model, the dollar does not shrink — it gets redistributed toward the people who own the chips, the memory and the data centres. Others piled in with the demand half of the story. Rohan Paul reached for Jevons: the largest AI workloads probably do not exist yet, because they are still too expensive to run at all.

Then the pushback arrived, and it was better than the usual bubble talk. It came in three flavours: the moat is growing, not shrinking; the pricing mechanism does not work the way the thesis assumes; and even if it does work, the thing that breaks is not equity.

Cheap intelligence only lowers the clearing price once cheap capacity is abundant enough to satisfy the marginal buyer.

The two sides

Why this is an investment question

Because it is the value-accrual question, and crypto investors have seen this film before. A layer being essential to a stack tells you nothing about whether it captures the economics of that stack. Chains that host billions of dollars of application revenue collect a rounding error of it. The AI argument is the same shape one level up: everyone agrees the compute is essential, and they are still arguing about who gets paid.

What decides it is scarcity, not necessity. Baker's thesis is really a claim that intelligence becomes abundant while compute stays scarce. Dratch's rebuttal is the one that binds: while compute is scarce, a cheap model does not set a commodity price — it prices a shade under the frontier and keeps the difference as rent. That means the labs' margin compresses only in the world where cheap capacity is genuinely abundant. Which is also the world where selling inference is worth almost nothing.

The TT desk thoughts

We think Baker is directionally right and early, and that the trade most people are putting on is the wrong one. "Long infra, short the labs" is the correct direction with the wrong clock: as long as compute is capacity-constrained, cheaper models harvest scarcity rent instead of forcing a commodity price, and lab margins hold. Position for margin compression to show up in 2027 numbers, not in this quarter's tape — and note zerohedge's point, which is the genuinely non-consensus one: if it does arrive, the first thing that repricing hits is the credit written against those margins, not the equity everyone is watching.

For our book, the crypto expression of this trade is decentralised compute — Bittensor, Akash, Render — and we are passing on it. The DePIN compute networks win share precisely in the abundant-capacity world, which is the same world in which a token of inference carries no margin. You are being sold a bet on AI and handed a bet on volume in a zero-margin commodity. The rule that survives both sides of this argument is the one that has always applied to token value accrual: own the bottleneck, not the layer that is merely necessary. In AI equities that means power, memory and packaging — tokens per watt, as one reply put it. In crypto it means the layer where scarcity is real and enforced, not the layer whose only claim is that nothing works without it. That test disqualifies most of the AI-token complex, and it is the one we are running our AI × crypto exposure through.

AI agents beat the 60/40 — in a backtest. Does it matter?

JPMorgan says AI agents allocating between stocks and bonds beat the classic 60/40 portfolio over 20-year backtests. The question the desk cares about: is this an overfit demo everyone should laugh at, or the first datapoint of a structural shift in who runs allocation?

The chain of the argument

The underlying report, surfaced by Walter Bloomberg (@DeItaone), is more careful than the headline: JPMorgan tested AI agents that independently shift money between stocks and bonds; the best one beat a 60/40 portfolio by 0.7 percentage points a year with lower volatility, and all eight agents produced stronger risk-adjusted returns. JPMorgan itself flagged that these are simulations, not live performance, and that AI allocators could crowd trades and amplify market stress. Then Polymarket compressed it into a "JUST IN" that did 459k views, and the replies turned into a referendum on backtests. The skeptics piled on overfitting and the low bar. @hegdedarsh read the report itself and pulled out a different finding — the off-the-shelf models beat JPMorgan's own quant systems. CryptoTweets zoomed out: JPMorgan, Robinhood, and Coinbase all moved AI agents deeper into trading in the same month.

Every allocator with a backtest beats 60/40. Almost none beat it live.@Ferbin08

It helps to know what the benchmark actually did. Over the last 20 years a plain 60/40 turned $100 into $492 — an 8.3% compound return, against 11.2% for holding the S&P outright. It gave up a third of the upside to cut the worst drawdown roughly in half: -31% versus -51%. That is the whole bargain. So "beating 60/40" is not one claim but two, and only the second one is hard: beating its return is what any equity-heavy mix does by construction, while beating its risk-adjusted return, out of sample, is the thing almost nobody does twice.

The two sides

The exchange, in their words

Why this is an investment question, not a meme

The skeptics win the narrow argument — a 20-year in-sample backtest run by models trained on that same history, against a benchmark that anything equity-heavy beat, proves close to nothing about future returns. But the note-worthy part was never the 0.7 points. It's two structural facts underneath. First, commodity frontier models matched or beat bespoke quant infrastructure inside one of the world's most sophisticated banks — which reprices what proprietary quant capability is worth, and what active allocation fees can charge. Second, the execution rails for agentic money are being wired right now: a major bank testing AI allocators, and retail brokers connecting third-party agents to live crypto and equity accounts in the tens of thousands. Whether or not the backtest survives contact with reality, agent flow is becoming a market participant.

The TT desk thoughts

Fade the performance claim, position for the plumbing. Nobody should allocate a dollar on this backtest, and any product sold on "our AI beat the market historically" deserves the raccoon treatment the quote tweets gave it. But the direction of travel is investable: value migrates from proprietary quant moats toward the toll booths agent flow must cross — execution venues, agent-connectivity layers like Robinhood's MCP, and the crypto rails where agents can hold and move money natively. The under-priced risk sits on the other side: if allocators converge on the same handful of frontier models, they converge on the same trades — JPMorgan said this itself. Same-model crowding is a new systemic correlation factor, and drawdowns in an agent-allocated market will be faster and more synchronized than the backtests that sold the products. Own the rails, not the track record.

Bitcoin-backed credit: new market or old leverage?

In the same week Strategy sold Bitcoin for the first time since 2022 to pay dividends on its credit securities, Metaplanet announced a study to let Bitcoin underwrite Japanese corporate debt. Both are called "digital credit." Only one of them is a new market.

The chain of the argument

On July 6 an SEC filing showed Strategy had sold 3,588 BTC for $216 million at a weighted average of roughly $60,201 — its first sale since 2022. The proceeds did not fund a pivot. They funded distributions on its preferred stock and refilled the dollar reserve that pays them. A week earlier the company had authorised a $1.25 billion Bitcoin monetization programme; it took seven days to use it.

Michael Saylor's framing arrived alongside: Bitcoin is the capital, $STRC is the credit, $MSTR is the equity — three instruments, one asset. Then Simon Gerovich announced Project NOVA, a joint study by Metaplanet, JPYC and Progmat into Bitcoin-backed digital credit for the Japanese mid-market, settled with a yen stablecoin and accruing interest daily. Japanese lender CRYL launched Bitcoin-collateralised loans up to $6.2 million the same week. Jack Mallers' Strike went further and offered loans it says a falling price cannot liquidate.

The market's reaction to the sale is the tell. It rallied. Charles Schwab called it a positive because it eased fears of cascading liquidations — which is a strange compliment to pay a structure that is supposed to be a bridge for institutional capital into Bitcoin.

Preferred distributions and debt service do not care about slogans. They require dollars.

The two sides

The exchange, in their words

Why this is an investment question

Two different things are being sold under one label, and the label is doing a lot of work.

The first is a financing structure. A company holds Bitcoin, issues preferred stock and converts against it, and pays a yield. Its engine is the premium the market assigns to its own shares: while the stock trades above the value of the Bitcoin it holds, issuing new shares buys more Bitcoin per share and the flywheel turns. When the premium goes, issuance stops being accretive and the dollars for the coupon have to come from somewhere. They came from 3,588 coins. Strategy's multiple to the value of its holdings now sits near 0.65x, against roughly 1.2x in the last bear market when Bitcoin was a third of today's price. The market is not repricing Bitcoin. It is repricing the mechanism.

The second is collateral. A borrower pledges Bitcoin and receives credit against it. There is no premium, no flywheel and no reflexivity — just an LTV, a rate, and a liquidation policy. That is what Project NOVA, CRYL and Strike are building, and it is a genuinely new market: Bitcoin's use case shifting from an asset you hold to an asset that underwrites other people's debt. JPMorgan's counter is that Bitcoin only survives as digital gold. Metaplanet is betting the opposite — that the endgame is collateral.

Conflating the two is how investors end up buying the leverage while thinking they bought the rail.

The TT desk thoughts

Own the collateral rail. Do not pay a premium for the wrapper.

The treasury-company trade has already broken and the tape has said so. Its only real moat was being the sole route for mandated capital to touch Bitcoin, and spot ETFs removed that moat. What remains is a levered holder with a coupon to service, a structure that must sell its asset into weakness — and the market now applauds when it does, which tells you the market sees these vehicles as a source of supply, not demand. K Wave Media has gone from a 10,000 BTC target to zero coins; Empery Digital sold 1,400 BTC and stopped reporting NAV on its Bitcoin at all. Treat a sub-1x multiple not as a discount to buy but as an accurate quote. If you want the exposure, own the asset.

The lending side is where the structural bet is, and it is early enough to be honest about what is unproven. Strike's promise that price cannot liquidate you is not magic: someone is absorbing that volatility, and Strike has not said who. Project NOVA is a study, not a product, and the sceptics in Gerovich's own replies are asking the right question — whether a Japanese CFO switches for settlement that costs the same as the bank loan they already have. So watch two things. First, whether any of these books survive a fast drawdown without socialising losses. Second, the July 30 Strategy call, the first since the sale — because how a levered holder explains selling into weakness is how you learn what the rest of the 200 will do when it is their turn. The rail is the trade. The wrapper is the exit.

▶ play

Charles Edwards on why 200 treasury companies weigh on Bitcoin — via @cryptoquant_com

Circle is now a bank — what the OCC charter really buys

The company behind USDC just became a federally chartered trust bank. The market read it as a coronation for the regulated dollar. The more interesting question is what a bank charter does to the thing crypto actually needed USDC for: settlement no one can stop.

What actually happened

The OCC granted Circle final approval for First National Digital Currency Bank, N.A., operating as Circle National Trust — seven months after the conditional nod. Initially the bank custodies digital assets for Circle and its institutional clients; management of the USDC reserve itself is listed as a future capability. That last clause is the one that matters. Today most of the reserve sits in a BlackRock-run government money fund and at partner banks. A charter is the legal permission slip to bring it in-house, under one federal supervisor.

$CRCL jumped double digits premarket. Two weeks after USDC overtook USDT on real volume, the regulated dollar now has something Tether cannot get: a U.S. bank charter. Sony Bank has a conditional one, too, with a dollar stablecoin planned. The charter is becoming the entry ticket.

The regulated dollar is pulling ahead — and every advantage it gains is an advantage granted by a regulator.

The two sides

The exchange, in their words

Why this is an investment question, not a press release

Strip the ribbon-cutting and two things changed. First, economics: a trust bank that manages the USDC reserve keeps yield that today leaks to fund managers and partner banks. On a reserve of this size, basis points are the entire equity story of $CRCL — this is the clearest path Circle has to widening its margin without raising a fee anyone can see.

Second, structure: the same charter that unlocks that yield hard-wires a supervisor into the dollar most of DeFi settles in. Circle could already freeze an address. The objection from @Vet_X0 and the Wisconsin and New York cases is sharper than "censorship" — the freeze power exists, but the recovery power does not. Users get the downside of a permissioned asset with none of a bank's make-whole guarantees. A charter formalises the first half of that trade. It does not, by itself, fix the second.

The TT desk thoughts

Two positions, and they do not conflict. On the equity, the charter is a real moat and a real margin unlock: it is the licence to run the float in-house, and licences of this kind are granted slowly. That is the bull case for $CRCL and it got stronger this week. On the rail, treat USDC as what it now unambiguously is — a regulated, freezable, bank-supervised dollar. Excellent for institutional settlement, custody and anything that will ever touch a compliance desk. Structurally unsuited to be the only dollar in a portfolio that also needs to settle when someone objects. Tether's RGB-on-Bitcoin push is not nostalgia; it is the hedge trade written as a roadmap.

The mistake this week would be reading a bank charter as a win for crypto's neutrality. It is a win for Circle, priced accordingly, and a reminder that the compliant rail and the censorship-resistant rail are now two different products. Own the first as equity. Do not confuse it for the second.

Robinhood Chain, week one: the broker that's eating crypto's rails

Days after mainnet, Robinhood's own L2 flipped Hyperliquid in 24-hour DEX volume, crossed $1B cumulative, and — per live Dune data — pulled ~500k addresses and 25M+ transactions onto the chain. The headline isn't the numbers. It's who put them up: a regulated broker, not a crypto-native, absorbing the entire on-chain stack.

The scoreboard

Robinhood Chain is an Ethereum Layer 2 on the Arbitrum stack, built for tokenized stocks and real-world assets with 24/7 settlement. Mainnet went live on July 1, 2026; Uniswap, Lighter, 1inch and Arcus were on it from day one. Here's where the first week landed:

Chain metrics pulled live from Entropy Advisors' Robinhood Chain dashboard on Dune ↗ (as of Jul 10, 2026). DEX volume and TVL are ecosystem figures from the sources above.

The surge, in the on-chain data

The Dune series tells the story the tweets only hint at. For the first week after mainnet, the chain ticked along near ~1M transactions a day. Then it broke open: 1.3M → 6.9M → 7.6M → 8.2M transactions on Jul 7–10, active addresses jumping from ~33k to 267k, and 120k+ first-time addresses in a single day. Daily fees went from ~$4k to $285k in the same stretch. That's not linear growth — that's the moment a brokerage's user base arrived on-chain at once.

Flipping Hyperliquid was the moment

Volume alone is noisy — incentive programs and wash-prone launch weeks inflate it everywhere. But Robinhood didn't beat a small L2 on a vanity metric; it beat Hyperliquid, the benchmark for on-chain trading throughput, on its home turf of DEX volume. And it did so while onboarding 120k+ first-time addresses in a day — the distribution advantage of a listed brokerage with tens of millions of funded accounts showing up on-chain at once. One honest caveat the Dune data surfaces: the failed-transaction rate spiked to ~30% on the peak day, the signature of bot and farming traffic riding the incentive wave. Some of this volume is real demand; some is mercenary. Both are visible in the same chart.

Robinhood Chain has collected $120K in chain fees in the last 48 hours. MegaETH did $3,000 in the same span.@stacy_muur

The real story: a regulated broker absorbing crypto's rails

Strip the launch-week fireworks and the structural point is the one that matters. This isn't a crypto company bolting on TradFi. It's the reverse: a regulated broker absorbing crypto's rails — exchange → wallet → blockchain → RWA tokenization, integrated top to bottom. Tokenized stocks (NVDA, GOOG, AAPL) that trade 24/7, post as collateral, and route into on-chain lending, distributed through Robinhood Wallet across 120+ countries. The chain is the settlement layer for a brokerage that already owns the customer.

We pulled four theses out of The Chopping Block's discussion with Vladimir Novakovski (@vnovakovski) that frame why this is a market-structure event, not a product launch:

Why this is an investment question, not a headline

The bear case writes itself: week-one volume is subsidized, first-time users are airdrop farmers, and a broker's captive flow isn't the same as durable on-chain demand. All fair. But the durable variable is distribution, and that's the one thing a regulated broker doesn't have to bootstrap. Hyperliquid earned its users; Robinhood already had them and moved them on-chain. If even a fraction of a funded, KYC'd, equities-trading base stays to trade tokenized stocks as collateral, the flywheel — fees, TVL, builders, more fees — compounds off a base no crypto-native can match.

This also sharpens a debate we've written on before. Robinhood pulls the token-vs-equity question onto regulated ground: here the equity is public and the on-chain assets are tokenized securities, not a governance token standing in for a promise. That's a cleaner structure than most of crypto — and a harder one to argue is "just exit liquidity."

The TT desk thoughts

Don't trade the launch-week volume; it will mean-revert. Trade the structure. The winners of the RWA cycle are the platforms that already own regulated distribution — and Robinhood just proved a broker can stand up an L2 that out-trades the best crypto-native venue in a week. Watch three things on the Dune dashboard from here: (1) whether daily transactions hold near their 8M peak or fall back toward the ~1M pre-surge baseline once incentives normalize — and whether the ~30% fail rate cools; (2) whether tokenized-stock collateral actually gets used in DeFi, or stays a demo; (3) the builders — Uniswap, Lighter ($LIT), 1inch, Arcus — capturing fees on someone else's regulated distribution. If the rails are Robinhood's and the flow is real, the apps deployed on top are the leveraged way to own it.

Watch: The Chopping Block on the Robinhood deal

▶ play

Vladimir Novakovski (@vnovakovski) on the Robinhood deal, token vs equity, and how not to run a DAO.

1X just shipped the hands. "An API to the physical world."

1X unveiled a new generation of hands for its NEO humanoid — 25 degrees of freedom, tendon-driven, force-transparent, with tactile skin that feels slip before it happens. Nothing like it has existed on the market. And they can build 10,000 of them this year.

Watch it first

Before any description — watch what these hands do. Twenty-five seconds of grasps, in-hand rotation, tool use, fine pinch.

1X — "Twenty-five ways to ask a question."

Not sped up. Real-time speed, dexterity and precision.

"A humanoid is a computer whose API is its hands"

That's 1X's own framing, and it's the right one. If the robot's brain is the computer, the hands are the interface: they set what it can know about the world and what it can do in it. A humanoid with a two-finger gripper exposes three verbs to developers — pick, place, push — and every application ever written on that platform is a composition of those three, executed blind. The ceiling was never in the software. It was at the end of the arm.

Most robot hands are write-only devices. You command a position; the hand goes there; nothing meaningful comes back.1X — NEO's Hands

What's actually new in the engineering

The part that matters most: it reads back

Every one of the 25 joints is natively force-controlled and doubles as a force sensor. Push a finger and it yields — and reports exactly how hard you pushed. 1X calls this force transparency: force flows out and information flows back through the same physical path. Layered on top, the fingertips carry high-resolution tactile sensing that measures normal force, contact location and shear — so the hand detects a slip beginning and re-grips before the object falls.

The consequence is that the hand always knows its own pose and effort without looking, the same way you can touch your fingertips together with your eyes shut. Motion becomes smooth by construction and fine motor control stops being a demo. Every grasp becomes a labeled experiment — which is exactly the substrate learning-based manipulation has been starved of.

Clips: 1X. Also demonstrated — screws and coins out of a wallet, light-bulb install, screwdriver use, tea pouring, catching a ball, wine glass, sign language.

The founder's claim

Our goal was never a hand that just looks impressive on paper. With these hands, NEO crosses a critical threshold. The robot can now do the things humans do with their hands, every day.Bernt Børnich, CEO & founder, 1X

Why this is an investment question, not a demo

1X is right about the hierarchy. On the hardware side of a humanoid, nothing matters more than the hands — and this is, without exaggeration, a new word in robot engineering. Put well-thinking internals behind hands like these and you get the most functional robot in the world, and by extension the market leader.

But note where the claim actually lands: 1X is not arguing it has better AI. It's arguing that the hardware ceiling is gone, so data is now the only barrier. That reframes the whole race. If dexterity stops being the constraint, competitive advantage moves to whoever can (a) manufacture dexterous hands at volume and (b) run the most contact-rich experiments per dollar. Both are supply-chain and throughput questions — actuators, tendon materials, tactile sensor stacks, yield, end-of-line test — not model questions.

Hence the 10,000-hands number. A hand that can't be built at scale can't run experiments at scale, and without data at scale there's no embodied AGI. That's the bet, stated plainly.

The TT desk thoughts

This does not change our position from Is physical AI investable? — it sharpens it. Don't bet on the humanoid; bet on what every humanoid must buy. A 25-DoF, force-transparent, tactile hand needs precision actuators, harmonic-free low-ratio transmissions, tendon materials, tactile sensor stacks, IP68 polymers and, above all, a high-yield line. 1X vertically integrated all of it — which tells you exactly where the scarcity is.

Two things to watch from here. First, whether anyone else ships comparable dexterity at comparable volume in the next 12 months; if not, 1X's data flywheel compounds alone. Second, whether the hands survive contact with customers rather than demo reels — millions of test cycles is a lab claim until fleets are in homes. Vertical integration is the moat while it holds and the cost structure when it doesn't.

What is RWA tokenization?

RWA tokenization is one of the most-hyped and least-understood ideas in crypto. Stripped to its core, it is simple — and the simplicity is what makes it powerful.

The plain definition

RWA tokenization puts a legal claim on a real-world asset — a treasury bill, a fund unit, a loan, an equity — onto a blockchain, where it can settle, move and collateralize 24/7/365. The token is not the asset; it is a programmable claim on it.

The value isn't the wrapper. It's programmable settlement plus global distribution on assets that used to be slow and siloed.

Why it matters

Traditional assets are locked behind slow, permissioned, siloed rails. Tokenization gives them instant settlement, global reach and composability with the rest of crypto — a treasury that can be posted as collateral in seconds, a fund unit that trades any time.

Where it's real today

The first real fit is tokenized treasuries and money-market funds — the safest yield in finance paired with on-chain settlement. Beyond that, private credit and funds bring higher yield and harder questions around custody and enforceability of the off-chain claim.

• Tokenized treasuries — the beachhead, real recurring volume
• Private credit / funds — higher yield, custody & enforceability risk
• The signal — recurring on-chain volume and clean redemptions, not partner counts

The TT desk thoughts

Judge RWA by real yield, real redemptions and real distribution — not by press releases. Tokenization that widens distribution and creates on-chain volume is durable; tokenization that just digitizes a closed fund is a wrapper.

How perpetual DEXs work — and why they matter

Perpetual DEXs have gone from a niche experiment to one of the fastest-growing corners of crypto market structure. Understanding how they work explains why they matter.

What a perp DEX is

A perpetual DEX lets traders take leveraged long or short positions on-chain, with no expiry, using a funding-rate mechanism to keep the perp price tethered to spot. All of it settles on a blockchain instead of a centralized exchange's internal ledger.

How it works

Traders post collateral and open leveraged positions. A funding rate — paid between longs and shorts — pulls the perp price toward the underlying. Liquidity comes from on-chain order books or pools, and liquidations are enforced by the protocol, transparently, on-chain.

• Funding rate — keeps perp tethered to spot
• On-chain liquidity — order book or pool depth
• Transparent liquidations — enforced by code, visible on-chain

Why they matter now

Perp DEXs have gone from one dominant venue to an asset class. Volume compounds, and CEXs are starting to outsource execution to on-chain venues — the deepest-liquidity protocol becomes the AWS of trading.

CEX outsourcing execution to a DEX, not the other way around. That's the infra play.

The TT desk thoughts

The signal is durable volume and network effects, not incentive-farmed activity. And the key question is value accrual — whether the protocol's token actually captures the fees it generates, or is just a sidecar.

Why stablecoins are crypto's killer app

Amid endless debate about what crypto is for, one product quietly won the argument: the stablecoin. It is crypto's clearest product-market fit.

The fit is settlement

Moving a dollar via stablecoins costs under a cent and settles in under a second, globally, 24/7/365. That is a structural improvement over correspondent banking — not a marginal one.

Sub-cent, sub-second, global. That's not an incremental upgrade to payments — it's a different system.

The dry-powder gauge

Aggregate stablecoin supply is crypto's cash on the sidelines. Expansion means buying power building; contraction means capital leaving the system entirely. Watching where new supply is minted maps intent before it becomes a bid.

Issuance is where value accrues

Who issues, under what reserves, on which chains decides where value flows. A bank turning USDC mint/redeem into regulated distribution accrues value to the issuer, to banks via client flow, to chains via settlement.

• Supply — expansion vs contraction
• Reserves — quality and regulatory standing
• Payment volume — real usage, not market cap

The TT desk thoughts

Go long regulated rails. Read supply, reserves and payment volume together — the stablecoin that captures durable payment volume under a clear framework is the one that wins.

Is physical AI investable?

Every few weeks a new humanoid demo goes viral and the question returns: is physical AI actually investable? The answer is yes — but not the way most people try to play it.

Not a bet on the robot

Robotics is not a bet on a robot. It is a bet on the supply chain, energy, compute, sensors, control systems, simulation and installed base. The humanoid shell everyone films is the least investable part.

Demos are easy. Ops are the moat.

Own the picks and shovels

The durable trade is the supply chain that takes a cut of the entire physical-AI economy — lithography, foundries, GPUs, EDA, power and cooling, and the industrial/medical robotics with real installed bases.

• Public core — semis, EDA, power, established robotics
• Private upside — defense, autonomy, industrial
• Avoid — generic humanoid startups without real deployments

Unit economics decide it

The signal is never the robot video — it is whether a company can ship, service and retain users without unit economics exploding. Paid pilots, on-device reliability and real deployment separate the trade from the narrative.

The TT desk thoughts

Being early and right on physical AI pays the most and is hardest to time. Buy the supply chain that earns from the whole economy, not the humanoid startup chasing the next viral clip.

How crypto and TradFi converge

Crypto is becoming financial infrastructure while most people still trade it like a casino. The gap between those two facts is closing — and the crypto × TradFi convergence is where it happens.

The convergence from both sides

The line between crypto and traditional finance is dissolving from both directions — banks issuing stablecoins, asset managers tokenizing funds, exchanges getting licensed. Each side captures a different layer.

Banks own trust and distribution. Crypto owns settlement and programmability. The game is who captures which layer.

Distribution beats everything

When trusted banks push crypto into everyday banking apps, access moves from specialist exchanges to bank rails and adoption steps up. Crypto-native venues that own neither trust nor distribution risk being disintermediated.

• Bank rails — distribution moving to trusted apps
• Custody & compliance — the institutional gate
• Tokenized finance — a regulated cash leg for on-chain settlement

The bridge is regulated products

ETFs converted crypto from an access problem into an allocation decision for trillions. Market-structure rules and regulated products are the bridge institutions cross — and the pipeline front-runs the flow.

The TT desk thoughts

Trade the gap between crypto-as-casino and crypto-as-infrastructure. Favor the layers that capture trust, distribution or settlement — and watch where banks and crypto stop competing and start plugging into each other.

Token vs equity: the Venice $VVV debate, for and against

A sharp exchange this week cut to one of the most important questions in crypto investing: in a company with both private equity and a public token, does the token actually own anything — or is it retail exit liquidity dressed up as a use-case?

The chain of the argument

Rob Hadick (Dragonfly) framed Venice's token, $VVV, as a way to hedge future compute prices. VIKTOR (@thedefivillain) pushed back: argue about phrasing all you want, but the token supposedly hedges compute — and literally nobody bought it for that. Hadick said that wasn't what he claimed. Fiskantes (Zee Prime) quoted VIKTOR and sharpened it: a dual equity/token structure carries a built-in incentive to distort where value accrues. Then Haseeb (Dragonfly) posted a video defense — and the replies split hard.

The token supposedly hedges compute — but literally nobody bought it for that reason.

The two sides

The exchange, in their words

Why this is an investment question, not drama

Strip away the personalities and the core is structural: can a token inside a dual equity/token structure be anything more than exit liquidity for retail? When ownership is split, incentives route economics to whichever instrument the insiders hold — and that is the closed equity, not the public token. Absent a legal claim, the token holds only a promise.

The TT desk thoughts

One more argument for a simple portfolio filter: favor projects where the token is the only instrument of ownership — Hyperliquid, Lighter — over projects where the token is an add-on to a closed cap table. Where the token is the whole business, value accrual and incentives point the same way. Where it's a sidecar to private equity, assume the token is the exit, not the asset.

▶ play

A clear-eyed look at how tokens actually work today

How capital flows move crypto markets

Price is the last thing to move. By the time a chart breaks out, capital has already been positioning for days or weeks — visible to anyone reading flow instead of candles. This is how the desk reads it.

Allocation precedes appreciation. Flow is the leading indicator; price is the lagging print.

1. ETF creations — institutional demand, printed daily

Spot Bitcoin and Ethereum ETFs publish creation and redemption baskets every trading day. Sustained net creations mean real balance-sheet allocation is entering — the slowest, stickiest money in the market. When creations run hot while price is flat, that divergence is the setup: demand is being absorbed before it clears into price. Heavy redemptions do the reverse and front-run risk-off.

2. Stablecoins — the dry powder gauge

Aggregate stablecoin supply is crypto's cash on the sidelines. Expansion means buying power is building; contraction means capital is leaving the system entirely, not just rotating. The tell is where new supply lands — mints flowing onto exchanges are intent to buy; mints sitting in DeFi are yield, not bid.

3. On-chain rotations — where the money goes next

Crypto is the only asset class where positioning is public. Exchange outflows signal accumulation and shrinking sell-side. Bridge volume exposes capital migrating between ecosystems ahead of narrative. Smart-money wallets — cohorts whose flows led prior cycles — rotate from majors to alts, spot to staking, one chain to the next, often days before it prints.

The signal is the overlap, not any single feed

No one input is edge. ETF inflows confirmed by stablecoin expansion confirmed by exchange outflows is high-conviction accumulation. One without the others is a trap — a hedge, a rotation, a head-fake. The desk trades that overlap, with real positions and a track record of original market vision since 2017.

From our timeline on X

/TT @toptraders0x 𝕏 This is the whole $ENA trade: strong product, weak token until holders actually get paid. Revenue screenshots don’t matter if the fee switch never turns on. 2026 · 575 views · read on x →

/TT @toptraders0x 𝕏 Circle's response confirms the threat is genuine: CRCL stock -17% in 24h. Jeremy Allaire posted a long network-effects defense. Standard Chartered institutional USDC minting announced days later. When the incumbent scrambles this fast, the economics are shifting. 2026 · 116 views · read on x →

Crypto infrastructure: the plumbing of the next market

Tokens rotate; rails compound. The durable trade in crypto is rarely the asset everyone is watching — it is the infrastructure that moves, settles and custodies value regardless of which asset is in favor.

Own the toll road, not the traffic.

Settlement is the edge over legacy finance

Crypto's clearest structural advantage is settlement — instant, final, global, 24/7/365. The rails that carry real, non-speculative volume are the ones that survive the cycle. We track throughput, fees and where stablecoin payment volume actually clears to separate durable infrastructure from testnets with a token.

Custody is the institutional gate

No risk committee allocates without custody it can defend. Qualified custody, insurance and segregation are the gate between sidelined capital and on-chain exposure — which makes custody maturation a leading indicator of the next wave of allocation, not a back-office footnote.

Stablecoins are settlement, not a sidecar

Stablecoins are quietly becoming the dominant settlement layer of crypto, and increasingly of cross-border payments. Their supply, chain distribution and payment volume map both liquidity and where real usage is migrating. Treat them as core infrastructure — because they are.

The TT desk thoughts

The signal is durable volume plus maturing custody plus stablecoin distribution pointing the same direction. Infrastructure repriced before the volume arrives is the asymmetry the desk trades — with real positions and a track record since 2017.

From our timeline on X

/TT @toptraders0x 𝕏 Big stablecoin signal: a G-SIB is turning $USDC mint/redeem into regulated bank distribution. Value accrues to Circle via issuance + treasury scale, banks via client flow, chains via settlement. Expect more banks to follow. Risk: lower margins. Takeaway: long regulated rails. 2026 · 435 views · read on x →

/TT @toptraders0x 𝕏 CEX outsourcing execution to a DEX. Not the other way around. This is the infra play — HL doesn't need to be the frontend. It's becoming the AWS of trading. 2026 · 217 views · read on x →

How crypto regulation reprices markets

Regulation is not a compliance memo. It is the single largest catalyst in crypto — a force that redraws which assets, venues and business models survive, often before the market understands what a ruling actually changed.

The asymmetry lives in the gap between the ruling and the market understanding it.

Policy prices in three stages

Rumor, ruling, implementation. Most of the market reacts to the headline and stops. The edge is in the third stage — reading the actual text and mechanics to see which sectors a ruling reprices once it is live, not just on announcement day.

ETFs turned access into allocation

Spot ETF approval converted crypto from an access problem into an allocation decision for trillions in institutional capital. Tracking the pipeline — which assets are next, which structures clear — front-runs the flow that follows access.

Stablecoin law is the high-leverage variable

Stablecoin frameworks decide who issues, under what reserves, on which rails. Clear rules pull payment volume and treasury demand on-chain; hostile ones push it offshore. Few regulatory variables carry more leverage over the whole market.

Market structure draws the line

Rules defining commodity versus security, exchange versus broker, reprice exchanges, tokens and DeFi in a single ruling. Enforcement patterns signal that line before it is codified. Read the docket, not just the headline.

The TT desk thoughts

Policy repriced before it is understood is the edge. The desk reads regulation as a trade — with real positions and a track record since 2017 — positioned ahead of the sectors a ruling will move.

From our timeline on X

/TT @toptraders0x 𝕏 Germany is a more important crypto signal than another exchange license. If Sparkassen + cooperative banks push crypto into everyday banking apps, distribution moves from specialist exchanges to trusted bank rails. MiCA is turning crypto access into a regulated bank product. 2026 · 361 views · read on x →

/TT @toptraders0x 𝕏 10 liquidation cascade structurally removed leveraged capital — no whales, no bid. @Binance still hasn't explained their margin architecture failure. But the bigger overhang is regulatory: no framework, no clarity, no institutional conviction. Market can't reprice risk without rules of the game. from Binance former CFO. 2026 · 28 views · read on x →

Why AI agents need crypto rails

As AI moves from answering to acting, software starts to hold value, pay, and settle on its own. Legacy finance has no account for a piece of software. Crypto does — which is why the rails agents transact on become critical infrastructure.

Crypto is the only financial system built for non-human actors.

Agents need three things banks can't give them

A native wallet, permissionless payments, and instant settlement. No KYC-per-transaction, no business hours, no human approval loop. An autonomous agent buying compute or paying for data at machine speed cannot wait on a card network or a bank wire. It reaches for crypto rails by default.

Agentic payments are the first real use

The earliest genuine economic activity from AI agents is transacting on a user's behalf — buying compute, paying for data, settling between agents. That demands machine-scale, 24/7/365 payments. Stablecoins and low-fee rails are the natural settlement layer for an agent economy.

Verifiability becomes the constraint

As AI mediates more value, provenance matters: which model ran, on what data, producing what output. On-chain attestation and verifiable compute are early attempts to make AI auditable. We track which cross from research into real usage — because narrative outruns reality here by the widest margin.

The TT desk thoughts

Being early and right on the AI × crypto stack pays the most and is the hardest to time. The desk researches it with real positions and a track record since 2017 — positioned before the machine-driven volume is real.

From our timeline on X

/TT @toptraders0x 𝕏 Market is sleeping on this angle. AI agents don't just need chat. They need money rails they can actually use: payments, FX, treasury. That's where stablecoin infra gets interesting. 2026 · 292 views · read on x →

/TT @toptraders0x 𝕏 Moonbeam moving $GLMR from Polkadot to Base is not just a migration. It is an admission that distribution beats chain ideology. New thesis: agent-to-agent communication + settlement needs liquidity, wallets and users. Base has those. $GLMR must prove usage, not history. 2026 · 1.1K views · read on x →