EDA intern retweeted
POV: the last thing D2 sees before leaking the secrets behind +216% net since Dec 2023.
Probably a bit more by now, thanks to $NVDA gamma. Turns out spies move markets.
We deliver alpha, but we’re beta to honeypots 🤣
EDA intern retweeted
Because he’s a D2 Finance Medium subscriber… and maybe a lucky depositor? 👀
“Do not fade Jensen until you do.”
— 30 July 2026
48% of NVDA revenue sitting on six customers’ discretionary, debt-financed Capex lines. CoreWeave and the broader buyer credit risk flagged months ago.
The structural bid held… until the tape (before the honeypots) finally forced the issue.
Texas Hedge receipt remains undefeated.
EDA intern retweeted
DO NOT FIRE JENSEN.
We agree with @KobeissiLetter and brought the tape (panel attached).
The exit that looks like this one is HP 2010. Hurd was delivering, the board removed him on conduct on a Friday night and the CFO took over as interim. HPQ -8.0% the next session, 11.4 points behind SPX twenty days later.
Where the market cheered an exit, Nike +6.8%, the stock had been lagging under that CEO. nvda was not lagging under Jensen.
Apple 2011 is the founder exit that worked, +10.9 vs SPX in 20 days, and Cook was already running the place. Nvidia's reported bench is a search firm and the CFO as interim.
Earnings are Tue 17 Nov after the close and options price a 7.5% move. Let Wachtell finish, print the quarter, then ask shareholders.
Fade the founder at your own risk.
Our message to Nvidia's board:
Do not fire Jensen Huang.
There is no clear replacement for Jensen Huang. We are confident that his departure would ultimately be a net negative for Nvidia, its shareholders, and the AI Revolution.
First, take a look at the data. History also suggests that removing a successful CEO will come with a significant cost.
A study of 187 forced CEO departures from S&P 1500 companies found that removing outperforming CEOs resulted in negative long-term stock returns. This case would be no different, in our view.
Jensen Huang is the definition of an outperforming CEO. Nvidia's board must not make a decision in the heat of the moment.
The board should take the time to conduct a thorough investigation and consult with shareholders, with a strong emphasis on the latter.
Quite literally, Nvidia's board should convene a special shareholder meeting and hold an advisory vote on this decision. This is the most consequential CEO controversy in world history. Shareholders deserve a say.
In addition, Nvidia's stock, $NVDA, is now down -35% in 26 trading days, nearly half of which is a result of the recent scandal.
We believe Jensen Huang's potential removal as CEO is now ~80% priced in.
Why take an action with massive downside potential when the market has already priced in much of that risk?
The downside of retaining Huang may be limited, while the potential upside is tremendous.
Lastly, history shows that retaining Jensen Huang could still lead to a favorable outcome. In 2019, Amazon retained Jeff Bezos as CEO after his extramarital affair became public, and the company's stock subsequently surged +108% by 2021.
Personal misconduct should not be ignored, but it must also be distinguished from corporate wrongdoing and weighed against the consequences of removing an executive who is instrumental to the company's success.
We believe Jensen Huang is an irreplaceable part of Nvidia and the AI Revolution more broadly.
Once the dust settles, firing Jensen Huang would prove to be a costly mistake.
It is time to pause and consult your shareholders.
EDA intern retweeted
The ultimate guide to the @aleabitoreddit portfolio, post semi blowup.
D2 desk note, Wed 11 Nov 2026. Ten names from Grok's list, not a disclosed book. Returns are since $NVDA's record close on Tue 6 Oct. For scale: $NVDA -29%, $SOXX -25%, basket median -47%.
1. $AAOI (-63%). One customer carries it: Microsoft was 28.8% of 2025 revenue.
Watch: Microsoft's 2026 capex consensus, $153.5B. It is the only one of the five hyperscalers lower than in July, and Microsoft says neocloud delays are why.
2. $SIVE (-56%). The AI laser story is ahead of the invoices. Bloomberg's customer list is led by Verizon at 3% and AT&T at 2%.
Watch: the first hyperscaler line above Verizon's 3%. Until then $568M of market cap rests on design wins.
3. $AXTI (-53%). Two steps from the capex line: it sells substrates to the laser makers, who sell to the buyers. Bloomberg shows no customer above 0.1%.
Watch: any datacenter-linked customer disclosed at 10% or more. Without one the demand read is second hand.
4. $LITE (-48%). Its largest customer is 26.6% of revenue, which Bloomberg names as Alphabet. Agreed on Mon 2 Nov to merge into Marvell.
Watch: Alphabet's 2027 capex consensus, $325.5B, the largest of the five and $24B higher than in July.
5. $SNDK (-48%). NAND is a pricing business first. Datacenter was about a quarter of fiscal 2026 revenue.
Watch: that 25% share. The rest is priced by phones and PCs.
6. $NBIS (-46%). A neocloud buys the GPUs and rents them out, so it is a funding story. CoreWeave is the template.
Watch: CoreWeave's interest expense at 2.2 times adjusted operating income in Q3, $846M against $379M.
7. $COHR (-37%). The least direct of the optical names. Bloomberg's top customers are Dell, Cisco and HPE at 23% combined, with $NVDA at 3.75%.
Watch: that 3.75%. The AI link has to show up in the customer list.
8. $POET (-34%). $877M of market cap and no customer data on Bloomberg at all.
Watch: the first reported customer above 10% of revenue.
9. $MU (-31%). The one large cap here, $820B, and the only one in $SOXX.
Watch: $NVDA's data center revenue on Tue 17 Nov against $89B last quarter. Bloomberg estimates $NVDA at 17% of $MU's sales.
10. $RKLB (-16%). Not an AI capex name: the US government is 42% of revenue.
Watch: that 42%. Its budget is set in Washington, and it still correlated 0.85 with $SOXX through the break.
What ties them. On Jul 30 we wrote "Do not fade Jensen, until you do", with roughly 48% of $NVDA revenue on six customers' capex lines. Today's pull is 48.4%: Microsoft, Amazon, Alphabet, Meta, Oracle, CoreWeave. The spending has not been cut. Consensus capex for five hyperscalers in 2027 is $1,173B, $119B more than on Jul 30. What moved is the cost of funding it. $NVDA's own 5-year CDS is near 140 bps, more than triple June, and CoreWeave owes $66B.
Serenity, which of the ten is the diversifier?
EDA intern retweeted
Still around? Pack it up. You already banked your $5M+.
DefiLlama has $73.2M of fees since launch and $67.7M of it paid to stakers. The other $5.56M is the dev fee, in USDC, on a "no team allocation" fair launch.
Now five contract upgrades at once on precompiles you call "un-battletested" yourselves. One lends the Exchange's idle USDC on HyperCore and routes the interest to stakers. Whose USDC is idle, the house's or the users'?
Stop slowing down HyperEVM iterations for the rest of us. Thanks.
Five contract level upgrades have been pushed to the timelock. The non-atomic nature of hyperliquid core and EVM blocks, along with the still broadly un-battletested precompiles necessitates extensive, heavy testing as the design patterns papertrade employs are unusual to see in solidity. Papertrade, by its nature is a risk on protocol. But when it comes to striking a balance between speed and security, security will always remain paramount, particularly when it comes to production contract changes.
Core lending:
The Exchange will now be able to supply idle USDC to HyperCore's native lending market, with the accruing interest being routed to stakers.
Collectibles:
A factory plus per-collection ERC-721 contracts for exchange collectibles and themes.
Staked PAPER transfer module:
PaperStaking gains a whitelist of transfer enabled contracts, along with contract level yield piping. This code is unused for now but will be necessary for future updates.
Sweep timing fix:
HyperEVM big blocks share a timestamp with a preceding small block. Whilst never triggered in production this could cause an ill-timed deposit sweep to fail and so the contract level timing checks are switching to block numbers rather than timestamps.
Direct LP donation:
A new admin locked donateToLp call moves USDC from a designated donor account into the LP's cash, providing a clean path for regulating externally held USDC back into the LP.
At some point we will likely write up a blog on our retrospective of the first month of papertrade, but we are few and there are only so many hours in a day or hats a man can wear.
A final quick thank you to those who helped out with adversarial testing, game theory and the general math that went into last month's upgrade, you know who you are. To those who are still confused -- often boisterously so, we thank you for sticking around for long enough to show you care.
good luck to us.
EDA intern retweeted
Finally. As a serious allocator we can say there is still too much crap on collaterals.
Offchain capital comes with a mandate, and a mandate needs an exposure it can name. A pooled market where $800M of WETH is borrowed against rsETH, or any serial rugger / low quality team that gets a listing, does not give one. Even onchain native money is finally more conservative than it used to be, and isn’t chasing points.
Happy to build HOT7 on @aave and we thank all the Hot spies out there for the exceptional fills.
Built on @base and @HyperliquidX. Coinbase stock tokens in the V4 Equities Hub, where USDC is the only borrow, perps short against them. The NVDA card is the hedge leg (the long leg is the Aave screen above it).
Funny that French KOL @Cbb0fe recommends now what we have been doing for over a year at this point (perp leg since xyz day one). Probably why the best art in France is the Gioconda, painted by an Italian KEK 😉
Stani, ship it.
LlamaRisk proposes repositioning Aave V3 Core, Aave's largest market, to focus on liquid, blue-chip collateral.
The proposal would remove several categories of exposure:
1. Restaking assets
2. Leveraged wrapping strategies
3. Issuers whose operational and security resources fall short of Aave's risk standards
4. Low-demand reserves that no longer justify ongoing maintenance
This means potentially moving $1.52B in collateral backing $1.17B in debt out of V3 Core in favour of blue-chip collateral.
As the risk landscape evolves and Aave V4 grows, the goal is to sharpen V3 Core's risk profile and raise the bar for collateral eligibility.
Some affected assets could transition to isolated Aave V4 markets, where more granular risk controls allow for better risk-reward alignment.
The proposal comes from LlamaRisk, an Aave Protocol Risk Manager, and remains subject to governance discussions and engagement with asset issuers.
Aave will continue applying continuous risk assessment for all assets across all deployments.
EDA intern retweeted
According to @arkham, although the tags are unverified, “Mr.Digits” appears to be using the best card on the market, @ether_fi 💚, to spend “his” money.
0xAdB6F825e34F26E9aD1864EEBd98B060d521B21F
There’s also a link to Coinbase, so any affected investors may want to ask @cobie for help.
Another extraction @blknoiz06’s timeline missed, and this one came with a $3M pre-seed and a leadership profile.
@DigitsDao wound down its fund on a muted Discord channel. The sister shop, Nave Finance (also wound down on a muted discord and deleted twitter), raised a $3M strategic pre-seed from the angel roster on the right of the picture. In July 2024 its director sat for a contributor piece titled “Critical Leadership Skills To Accelerate Trust As A New Asset Class”. He is now a Discord screenshot.
Same weekend, same shape on @berachain: @puffpaw’s $VAPE printed a $400M FDV on a pool seeded with 20K USDC, the print that settles the @Polymarket FDV market, with @SmokeyTheBera saying in public that investors were ghosted for months. Look at the two cap tables. A Digits Capital is on @puffpaw’s $6M seed graphic (@Lemniscap lead). @0xbeartic and @CBB0FE are on both.
Nobody asked who audits an on-chain hedge fund. The answer is the chain, every epoch, in public. D2 Finance has done that on @HyperliquidX with no VC money since Dec 2023: 34 of 34 months positive, +216% net, almost $3M paid out to users, every epoch returned on chain before the next one opens.
Viperr @viperr, you were bullied into holding. The receipt for the other side of that trade is the chart. This is the extraction model: raise from friends, “publish trust”, wind down in the dark.
@ChairmanSelig pls regulate the space!
EDA intern retweeted
Another extraction @blknoiz06’s timeline missed, and this one came with a $3M pre-seed and a leadership profile.
@DigitsDao wound down its fund on a muted Discord channel. The sister shop, Nave Finance (also wound down on a muted discord and deleted twitter), raised a $3M strategic pre-seed from the angel roster on the right of the picture. In July 2024 its director sat for a contributor piece titled “Critical Leadership Skills To Accelerate Trust As A New Asset Class”. He is now a Discord screenshot.
Same weekend, same shape on @berachain: @puffpaw’s $VAPE printed a $400M FDV on a pool seeded with 20K USDC, the print that settles the @Polymarket FDV market, with @SmokeyTheBera saying in public that investors were ghosted for months. Look at the two cap tables. A Digits Capital is on @puffpaw’s $6M seed graphic (@Lemniscap lead). @0xbeartic and @CBB0FE are on both.
Nobody asked who audits an on-chain hedge fund. The answer is the chain, every epoch, in public. D2 Finance has done that on @HyperliquidX with no VC money since Dec 2023: 34 of 34 months positive, +216% net, almost $3M paid out to users, every epoch returned on chain before the next one opens.
Viperr @viperr, you were bullied into holding. The receipt for the other side of that trade is the chart. This is the extraction model: raise from friends, “publish trust”, wind down in the dark.
@ChairmanSelig pls regulate the space!
EDA intern retweeted
Diversification is the only free lunch of finance, we want to eat it all.
"30 names is enough" is really a statement about correlation. Say your names are 0.2 correlated. Thirty of them behave like 4.4 independent bets, and you can keep adding forever without getting past 5. For a book of stocks the lunch does end early, on vol he has a point.
The point of the D2 platform is not stock but strategy: diversify different frequencies, assets, mandate, risk level. Screenshot attached: USDC, gold, BTC, HYPE, ETH, global macro and a stock basket on one page, risk from Low to Degen. Get strategy correlation down to 0.05 and that ceiling is 20. Ten of them already count for 6.9. Multi-strats don't stop hiring at pod 30 for the same reason. That 0.05 is what we aim for, and keeping it there when everything dumps together is the actual job.
And all of that is just vol. @RoniIsraelov ran random 25-stock portfolios over 1994-2019 (h/t @choffstein). One in ten ended 36% or more below the average.
Harry, fade the free lunch at your own risk.
What is the right level of diversification for a $19 billion endowment?
“The quantitative benefits from diversification trail off really quickly when you get more than 20 or 30 names in the portfolio.
There are really no quantitative benefits beyond 30 underlying look-through exposures.
We target roughly half of our portfolio in our top 50 look-through positions. It is still very diversified.”
Love to hear your thoughts @choffstein @MebFaber @dampedspring @awealthofcs
EDA intern retweeted
medium link
medium.com/@D2.Finance/texas…
EDA intern retweeted
TL;DR
Gold printed the exact same level as one year ago.
Last time we were here it ripped +37% to 5,595 in 12 weeks.
Same setup. Stronger bid. Epoch 2 is live.
The case for gold into Q1 2027, from the D2 desk.
Gold printed 4,070 Friday, the same tick as a year ago (4,001 on 7 Nov 2025). From that tick the next twelve weeks ran it to 5,595 (chart 1, green box). The 2006 template, mapped onto the 29 Jan top, puts about 5,000 in the second half of January 2027. Debasement++ Epoch 2 runs 100 to 300% of NAV gold delta into 19 Feb 2027 on that read. Ten points, timeline since the Trump election.
1. Cycle low 2,537 on 14 Nov 2024, nine days after the election. 5,595 on 29 Jan 2026, +121% print to print. The buyer was the official sector: Japan and China cut Treasuries for a decade while adding gold, the swap @LukeGromen has charted for years. Overseas UST holdings fell ~$70B in June.
2. 28 Feb 2026 the Iran strikes started; gold has given back 23% from the pre-war close. Oil pushed CPI to 3.7% by September and the Fed from easing to hiking. Central banks sold into us to pay for the oil. Brent: 72 pre-strikes, 126 at the April top, 98 now. The 2025 setup returns if Iran resolves and it trades in the 80s.
3. 19 Aug Bessent doubled long-end buybacks. Yields dipped a day, back above 4.7% within the week, dollar down vs every G10, gold +5%. Window closed 4 Nov with the refunding; ~$6T rolls in Q4; ten-year 4.86% now. Bessent remains Bessent.
4. The 3 Nov result does not touch the arithmetic. $38.6T at a 3.1% weighted coupon; Q4 coupon interest alone $196B. Bloomberg Economics has interest at 6.4% of GDP by 2035, above Italy, debt to GDP 131%.
5. October was the washout: almost -7%, worst October since 2008, low 3,870 on the 30th. Goldman says early to buy. CFTC managed money net longs 104,297 in the week to 3 Nov, a 21-week low, shorts at a 17-week high. The washout @MacleodFinance tracks weekly in COMEX data.
6. Democrats swept. Gold +5% on the week, BTC -2.6% since Tuesday's close (round trip in the quoted post). Whether crypto trades as a debasement hedge or a Trump trade is unclear. If the Dems go after it hard, better the old rock than digital, and yes @PeterSchiff, that point is yours. BTC and MSTR stay in the basket, smaller.
7. Saturday's data: PBOC added 830K oz in October, month 24 of the streak, buying harder as price fell. Ten-month additions 129t; official sector net 321t in Q3. WGC survey: 45% of central banks plan to buy next year, a record, the bid @JamesGRickards has argued since The New Case for Gold. Poland at 700t.
8. The bid sits in Asia. The July rebound ran in Asian hours while London and New York went sideways (chart 2). India's wedding-season demand is back on the pullback. Hong Kong central clearing for gold goes live Q1 2027.
9. The 2006 template: 22% fall in 24 trading days from the 12 May top, three fifths back a year later, new high 16 months out. This leg ran 31% and sits at day 195, so the cycle is behind the template. The difference is the buyer: 2006 had no PBOC adding every month of the fall. Date soft, shape is what we trade. @PeterLBrandt @TheDailyGold, does a 31% leg break the analog or stretch it? Three-month vol 22 vs 28 realized, flat risk reversal: the call wing is unpaid.
10. The short leg is the Fed. Warsh is the hawk. Friday's payrolls beat put a 9 Dec hike at 78%; gold gave back 1.5%. Ten-year real yield 2.61%, DXY 4% above the January top. Gold outran the dollar in the 2009-10, 2020 and 2025-26 suppression episodes and lost each time credibility came back. 30-day gold-dollar correlation -0.44.
Epoch 2 deposits open now, term 19 Feb 2027.
Markets do not repeat, but they often rhyme.
EDA intern retweeted
Debasement++ Epoch 1: settled ✅
Closed 9 days early. 40.684 @tethergold XAUT in, 41.32 out.
📊 +1.57% net in gold terms | ~7.8% net APR
📉 Gold 4,614 → 4,054 over the same window
Where yield came from:
◾ Gold: September and October adds lost into the slide. The election-week round trip, 183 oz bought at 3,947 avg and sold at 4,040 into the post-result bid, more than paid for them.
◾ MSTR short: flat. The 2 Oct squeeze to 166 cost what the November covers at 131 and 144 earned back.
◾ BTC: small long, closed mid-October, small positive.
Wrong on timing. Trading paid for it. Results still positive 🫡
Legs flat, thesis got its catalyst, so capital comes back now rather than sitting idle for nine days.
New app preview below: rebuilt on your feedback, hardened. Coming soon.
EDA intern retweeted
Best pod risk thread we've read in a while. "Factor beta neutral to the extent they can measure the exposures" is the line.
The part that hit home is who the model is good for. The GP scales the winners, cuts the losers fast and never has to work out whose process is good. Being a pod is not great vs being the GP. We're actually in defi partially to not do pod life, as we can manage risk much better. Self selection of users through the epoch model helps.
If every pair of teams is 0.05 correlated, a pod shop tops out at 20 independent bets. 300 equal teams get you 19. Park the same 0.05 average with the ten teams that carry half the risk and it's 3 (chart).
A blow-up can't tell you it was crowding either. Start at 1% odds. One 6 sigma day makes it 38% if crowding is the only suspect, 15% once a plain vol spike is a second one.
Walleye plans to return some capital after a -5.9% month. Pod Risk Manager, agreed the bottom tier goes first. We'd bet Walleye is not the last capital return before year end.
In honor of some “pods blowing up” and me being traveling I will outline some concepts about pod risk management about which I do know a thing or two, as my name might suggest.
It is a common perception that pods are factor beta neutral and thus largely hedged. That is not quite true. A truer statement is that they are factor beta neutral to the extent they can measure the exposures.
Most equity risk models are derivatives of the original axioma and barra factor models. There are now many things that go into this but at their core they are a 12 (IIRC) variable linear regression. And mathematically, the unexplained remainder, the error term, is assumed to be the alpha (Upon which the team gets paid). You well recognize most of these factors: size, momentum (short and long), quality, dividend, domicile, sector, etc.
Now even the dull among you may note that there are in fact many other things that go into a stock. That’s true and that’s one reason why many people at the pods are very rich. There is though a very active and live debate to the highest level of these firms as what you “want” in the other bucket. On the one extreme, scheduling a proprietary drone flyover on the Aurora AV truck up fitting facility to measure parking lot activity vs expected timeline….that feels like pretty good alpha. On the other extreme, going long the Canadian listing rather than the U.S. listing because the Canadian listing diversifies the domicile score so you get a higher portfolio level idio score…..that doesn’t feel like great alpha. But the range is wide: paying GS a billion dollars so your team is the first meeting at the conference? That’s kind of the pod scale advantage. Albeit less so if every other pod is in the room. Going balls long large cap domestic U.S. GLP-1 manufacturers vs non GLP-1? Well that kind of is thematic alpha (depending on the work put in by analyst), but even if you’re of the opinion fuck it, let the kids rip (a reasonable view) you also know at the levered portfolio level that we aren’t measuring that risk completely. This line can really be illustrated with the concept of a “custom factor”. Maybe going long a shit load of SPACs is alpha, maybe it isn’t, but how the FUCK are you gonna build a portfolio level SPAC factor exposure loading in late 2021 when spacs didn’t functionally exist until 2021????
Anyways, interesting & active philosophical debate but there are some downstream derivatives.
Firstly, as a well known matter, 6 sigma events happen in finance all the time. There’s one every year in some corner of the market. So ex ante we know our models aren’t “great”.
Second, everyone knows the agent is an unobserved blind spot. It’s impossible to actually “know” why the analyst and PM are doing anything. You like to think they’re doing it on the back of deep fundamental quality work, but the median analyst and PM isn’t good at their job. So that assumption is known to be unreliable. The capital allocation solution to this is momentum chasing at the portfolio level: you allocate more to winning pods and you cut the losers fast. This is a brilliant approach because it means the pod business model is capital light- they identify and scale the good teams without actually having to invest in knowing which pods have “good” process. It’s worth really appreciating that point because it’s one of the true genius aspects of the model, and at its core it’s why the GPs are so fucking rich. (Deep finance readers will note that some of the original papers on factors saw that one of the earliest possible explanations for the decade-long trend nature for factor performance is implicit fund flow pressure from the momentum chasing nature of capital at the LP level)
(Continued in reply)
EDA intern retweeted
VIX +11.57 to 28.18 and the curve inverted the most since April 2025. We think pods are still cutting.
Walleye letter in the quote: worst month for the firm since March 2020. Vol desk had its best month ever in October. So a big chunk of the in-house long/short guys go and vol stays (nearly 20 vol PMs hired in 2 yrs). They started as an options market maker in 2005 anyway.
Makes sense to us. Stock picker pods all own the same names, which is fine while single stocks move against each other and the index sits still. Thursday everything went down together, ndx over 6%. Vol desk gets paid on that day.
Their 13F attached, sorted by % of shares out. It's 30 June and longs only so take it for what it is. 1,848 lines, 11 of them over 1% of shares out. You don't sell that in a day.
We trade vol for a living, we'd keep that desk too.
EDA intern retweeted
TL;DR: Price Hyperliquid as the probability it becomes the house of all finance, not as an exchange.
That probability is two years of relentless shipping from a zero-outside-capital genesis. Keep the 500k stake. The flywheel is intact.
1) The stake.
500k HYPE was ~$1M when we first bought near $2. It’s ~$40M+ today. Keep it.
Base rate in this space is 95% scammers on day one or the moment the numbers stop working. A slashable bond is the cheapest, cleanest filter, and it sorts on merit: @tradexyz already earns an implied ~74% on its stake while every other deployer sits under 5%. The docs are explicit: the requirement falls as infrastructure matures. 500k is raisable if you’ve been building in the HL community since the early days.
2) Fees.
Fees compress only when a rival clears the same book at the same quality. Not yet.
~$200B+ perp volume in 30 days, ~$80-90M monthly fees, 97%+ routed to the Assistance Fund, ~$900M to $1B run-rate.
HIP-3 charges 2x base (9 bps taker / 3 bps maker), split 50/50 with the deployer, so protocol take per dollar matches the core book.
Growth Mode cuts fees 90% on electing HIP-3 markets; the core schedule is untouched. We don’t price the fee stream in isolation.
What we actually price.
P(house of all finance) × the value of every market that lands on the book.
In under 24 months from TGE: HyperEVM, HIP-3 perps on equities/commodities/pre-IPO, HIP-4 outcome markets, tokenized stocks, ICE partnering,...
Amazon ran low-single-digit margins for two decades while taking the shelf. The market paid for the shelf.
Here the market is paying ~20-25x run-rate fees on circulating and a much higher multiple on FDV: an exchange multiple for what is already gold + HOOD on one FAIR margin engine.
@choffstein, 99% of a multi-trillion-dollar space can be uninvestable and Jeff is still the best operator in the room. The 1% that actually executes earns the premium.
Given Hyperliquid's aim to be the neutral platform that becomes the "house of all finance", can I ask two questions without everybody getting upset?
1) For a platform that famously did not raise external capital, does the 500k $HYPE deployment hurdle limit future innovation to only those who can?
2) Given Hyperliquid's extreme profitability, at what point do fees start to come under pressure? Does "growth mode" eventually become the permanent rate?
EDA intern retweeted
Hard disagree. We have run stops on every book we have touched for ten years, prop desk, hedge fund, and since 2023 vaults, and the only people who tell us stops are retail comfort are people who have never had to call an investor and explain where the budget went.
TL;DR: a stop has nothing to do with whether the asset is good. It is what happens when the loss budget runs out, decided while you could still think straight. Your reason 1 already is a stop, you just do not call it one. hXXI hit its stop at 76K in March and a 23% BTC drawdown became 30 bps.
The Copernican objection treats the entry price as a random timestamp, but entry is where the loss started accruing against a finite loss budget, and that budget is the only thing a risk manager is paid to protect. Whether the asset is worth holding is a different question answered by a different person.
Take reason 1 seriously and the whole argument collapses into it. Every dollar of drawdown is a dollar off the loss budget, and realised vol rises when a position moves against you, so the vol-scaled size you are allowed to hold shrinks at the same time. The stop is where both constraints bind at once instead of piecemeal. Its level is sigma times a horizon times a multiplier, a precise number that means exactly what it says: the point at which the remaining budget no longer covers the next move.
Reason 2 is the same thing from the other side. A pre-committed stop is the answer, written down before the position exists, to "what does my thesis say about a move this size against me". If the thesis says nothing, the stop fires. If the thesis is mean reversion and the move is the setup, the stop was set wider at design time. Either way the decision was made before the loss was on the screen, which is the one state where the trader's judgment is least worth trusting.
Then there is the capital that is not ours. A vault epoch carries a loss budget investors subscribed to. hXXI stopped its long delta at 76K in the March selloff, a 23% BTC drawdown, and closed the epoch within 30 bps. Without the stop the book carried long delta to 66K, far past the loss budget. Holding through because the asset might be "worth holding" would have substituted the fiduciary's conviction for the investor's loss tolerance.
Returns compound geometrically. A 50% loss needs 100% to recover, and open-ended losses on a positive expectancy book still go to zero if the bet size is large enough relative to the edge. The stop puts a floor under the loss distribution. It is not a put: it pays for the floor in forgone rebounds instead of premium, and it can gap. For a short vol book with financed exposure the alternative is no floor at all.
Where we agree: a stop at a chart level with no vol input and no loss budget is theatre. A stop inside the bid-ask noise is a donation to the market maker. A stop wider than the position's maximum loss is not a stop. The fix is a better stop, computed off vol and the remaining budget.
All of the above is fair enough for one guy, his own money, no mandate and all the time in the world to be right. The rest of us are running other people's money with a number on it, and that number needs a trigger.
Rocky got the risk management right: it is about how hard you can get hit and keep moving. The stop is how a book keeps moving. So, @DeepDishEnjoyer, honest question: when your book is down the budget, what pulls the trigger, a rule you wrote last month or you, at 3am, staring at the screen?
Replying to @dmartin_trading
the idea of "calculating a stop loss" is nonsense as well
there should only be two reasons to hack out of a trade that you thought had positive expectancy at even less good prices than current prices (since it's going against you)
1) margin/value at risk concerns that changed
EDA intern retweeted
Global Macro Epoch 1 conclusion + new epoch ✅
📊 +6% net, after paying more than 10% in funding. Gross on the book: north of +16%.
GM trades the directional signals from HYPE++ outright, one hard rule: 10% loss limit per epoch. Epoch 1 ran $100K as a test. Apart from the momentum signals, nothing was optimized.
Volatile by mandate:
◼ Gains came early on Korea flow, same as Kimchi. Peak around +17-18% in early September
◼ Then stopped out roughly 10 times in a row. Back to slightly negative, inside the 10% limit
◼ Rebuilt the P&L in October when the opportunity got interesting again
------
The dark side
Our other strategies run trades of 2 to 60 days, where execution and net funding wash out. HYPE++ usually receives funding on @HyperliquidX. GM plays the 30-minute to 72-hour frame at higher leverage, and funding ate more than half of the edge.
What changes: the next short epoch stays the same. Later iterations add Pendle Boros, the funding-rate market, which benefits from exactly what hurt GM here. Boros liquidity is still too small for scarce PM time on its own.
------
📅 Global Macro next epoch
• Deposits: open now
• Trading starts: Nov 5
• Epoch ends: Nov 20
• Cap: 100,000 USDG
Two weeks, plus up to 30 days of tactical.
After the short epoch we will most likely split the GM signal in two: GM 2.0, and Overnight Edge, whose goal is to beat $SPX on a risk-adjusted basis by taking advantage of 24/7 markets.
We stay liquid. Fade D2 at your own risk.
EDA intern retweeted
What the article leaves out: we were flat before we were fast.
The drain started 20:20 UTC. Order of operations on the desk:
1. Depositors first. HYPE vault delta flattened on Hyperliquid, considering the 2M HYPE still sitting on Binance hot wallets the thief had not emptied. Vault 2.85M USDC allocation pulled from @Aave on Arbitrum and Base before the freeze. Tx attached.
2. Short the names in his wallets before he sold them, on Hyperliquid: WLFI, MORPHO, FET, AAVE, ARB, LINK.
3. Long roughly the same names once he was done.
The arb was a bonus. The product is step 1, and it is why 34 months since Dec 2023 have all closed positive. We protect capital FIRST before trying to be clever.
Hyperliquid
Welcome aboard, @standuquesne. @Hyperliquidx held 100% uptime through the Binance drain while Binance was shut three days. The venue ICE is bringing to the US is the one we have traded daily since before the HYPE TGE. Thank you @chameleon_jeff, 100% uptime again.
Thank you @business for quoting us today in “Frantic Binance Hackers Created Windfall for Fast Crypto Traders”. After 10/10, that is a second Binance incident we closed positive. We may need to airdrop @cz_binance a slice of the gains, broccoli 😂 dinners do not pay for themselves.
33 epochs since Dec 2023, 32 positive, and now we wonder whether Druck follows our trades.
Go easy on our on-chain, audited edge, Stan. 🫡
EDA intern retweeted
Welcome aboard, @standuquesne. @Hyperliquidx held 100% uptime through the Binance drain while Binance was shut three days. The venue ICE is bringing to the US is the one we have traded daily since before the HYPE TGE. Thank you @chameleon_jeff, 100% uptime again.
Thank you @business for quoting us today in “Frantic Binance Hackers Created Windfall for Fast Crypto Traders”. After 10/10, that is a second Binance incident we closed positive. We may need to airdrop @cz_binance a slice of the gains, broccoli 😂 dinners do not pay for themselves.
33 epochs since Dec 2023, 32 positive, and now we wonder whether Druck follows our trades.
Go easy on our on-chain, audited edge, Stan. 🫡




































