Janus Dispatch Podcast

About Janus Dispatch Podcast

Mapping the architecture of reset. janusthewatcher.substack.com (https://janusthewatcher.substack.com?utm_medium=podcast)

Publisher

Janus The Watcher

Category

society & culture · society & culture

Language

en

Explicit

No

First episode

4 May 2026

Latest episode

14 Oct 2026

Latest episodes

42 episodes in the feed.

1. 

   14 Oct 2026

   Flying under the Radar: Notes from a Dark Horse

   I usually write about protocols and balance sheets. This piece has neither. It is a personal note from the third year of a ten-year plan, written mostly for myself and published in case it is useful to someone standing where I stood in 2023. Year three of ten Every other Friday the flat changes its operating mode. Two more pairs of shoes stand in the hallway, the fridge fills up, the noise level triples. Seven days later the shoes are gone and the place is quiet enough to hear the fridge. From the outside this looks like a life at half size. A divorce after almost two decades. No title anyone would find impressive. Out front, an old Passat, a few scratches and dents, dog hair and old chewing gum worked into the seats. People who knew me before look at this picture and draw their conclusion, and I let them. What they don’t see happens at night, at a desk, with central bank balance sheets and protocol documentation open side by side. I have been doing this for three years, and nobody in my daily life connects the man in the Passat with the man at the desk. For a long time that bothered me. It took most of three years to understand that it is the most valuable thing I own. These are notes from the middle. I don’t know how this ends. I am writing it down now because in 2033 I will remember the outcome and forget the fog. The house For almost two decades I lived by a different blueprint. There was a business, a marriage, and a house that was supposed to get built. Each of them was reasonable. Together they formed a structure in which every decision had already been made by the one before it: the house needed the income, the income needed everything to stay as it was, and the life needed the house. I had built myself into it, beam by beam, and I had done so willingly. The house never got past the plot. For a while I paid the financing on a piece of land nobody used, on a floating rate, in the years after Covid when rates only went one way. Every month the empty ground cost a little more than the month before, and it stressed me more than I admitted at the time. When the plan was abandoned, the building company wanted a cancellation fee worth about three years of savings. The plot was sold at a small loss. Even unbuilt, that house was the opposite of the Passat. It could not move. It tied me for decades to the person I was when I signed, and it sent a bill every month for something that did not exist. I tried for years to change how I thought while the plan stood, and it never worked. You cannot redraw a floor plan while you are still paying for it. In 2023 the structure came down, when my wife and I got divorced. I won’t describe that year. When it was over I stood at zero, in the literal sense of the account. By then there was nothing left to divide, which kept at least that part short. What I had instead was something I had not had in a long while: room to draw a new plan. I owe something to everyone who has been part of my life, including the people I am no longer close to. Some of them showed me what I wanted. Others gave me its outline by wanting something else, the way empty space gives a shape its edge. None of this is written against anyone. Two rules The new plan started with a rule from a book that is a hundred years old, The Richest Man in Babylon: a part of all you earn is yours to keep. 1: Ten percent goes away first, and you live on the remaining ninety. The mechanics are trivial. Anyone can do the arithmetic. What the book leaves out is how the rule feels when you introduce it at the worst possible moment, with two households where there used to be one and a month that is longer than the money. What it cost in that first year was mostly ego. Holidays used to mean a flight to Egypt. Now they meant a caravan by a lake in Mecklenburg, Germany. The lake was fine. The difference in money was small, and the comparison in my head was the expensive part, because it was a comparison with the man I had been. I kept the rule anyway, for a reason that had little to do with returns. It was the first decision of the new life that the old structure had not made for me. The second rule is harder, and it governs the good days. There are no exceptions, and nothing is decided on the day. 2: Whatever arrives outside regular income gets split down the middle: half goes into living, half gets invested. I have come to think the second rule matters more. Ordinary months don’t tempt you. Windfalls do. Money that feels like a gift wants to be spent where others can see it, and the day it arrives is the day you feel you have earned the right to show something. The rule takes the decision out of that moment before the ego gets a vote. The living half matters as well. It pays for the things I would otherwise resent missing. Discipline without that half turns into asceticism, and asceticism breaks sooner or later. The night shift Alternating weeks force a rhythm on you. In the weeks with the children I am present, and the desk stays mostly dark. In the other weeks I go deep. I want to be careful here, because this can easily sound like a productivity story, and it is a slower thing than that. Reading macro and crypto at night for years produces nothing you can show anyone. There is no certificate. What accumulates is judgment: a sense for which narratives carry weight and which are decoration, how a protocol actually pays its yield, what a balance sheet is hiding. It came with a great many lessons, and with a side effect I did not plan for: I have become very sensitive to scams, which in crypto may be the most useful instinct of all. Attention compounds the way capital does, and in the first years the curve looks just as flat. The flywheel The two rules feed a simple machine. Most of what I set aside sits in things that are boring and hard to kill. A small part goes into bets where I can lose only the stake and the upside is open. I know the other kind of position. A floating-rate loan on empty land is a bet with the downside open. Every position is a bet with a size, and I try to know beforehand what would prove me wrong. None of them is a conviction I need to defend. The machine has started to pay. Yield now covers a growing slice of the monthly bills. It is a slice, and I am a long way from done. Charlie Munger is reported to have said that the first hundred thousand dollars is a b***h, but you have to do it. I feel that sentence absolutely. Leave his number aside: early on, your own effort does nearly all the turning, and what the wheel pays back looks like a rounding error. But every bill it covers is one I no longer have to earn, which buys patience, and patience improves the next bet. Then the wheel turns again. The hunger Now the part that is hard. Everyone who has been underestimated knows the fantasy. There is a dinner, or a chance meeting in a car park, and in that scene they finally see it. I have run that scene in my head more often than I would like to admit. That fantasy is the real risk in my plan, more dangerous than any drawdown. It has a price tag, and the bill arrives on the windfall days: a newer car in place of the Passat, a casual remark at a family table, a post. One such moment is enough. People update their picture of you, and that update runs one way. In the same circle you don’t get to be the harmless one again. Whoever is known to have something becomes someone to ask, to envy, to measure and to manage. So the Passat stays. There is a trap in this too, and I should name it. Displayed modesty is also status, played to a different audience, and a man who is proud of his old car has only found a quieter way to be seen. The car means nothing. It drives, the dog fits in the back, and nobody looks at it twice. The cover has a running cost. Cheapest was never the same as free. You get no applause and, worse, no correction, because when nobody knows what you are doing, nobody tells you when you are wrong. You are misjudged by people whose judgment you care about, and you have to let it stand. Some evenings that weighs more than any rule about ten percent. The cover also works where I would rather it didn’t. Meeting someone new is hard when the first things she sees are an old car and a calendar that belongs to the children every other week. I could correct the picture, and then I would never know which man she said yes to. No finish line A dark horse story normally ends with the race. Mine has none. At ground zero I set a goal with a date: by 2033 I want to never again say no out of fear, or yes out of need. The year is not arbitrary. It is roughly when my children will be done with school. Until then the job is plain: they should lack nothing that matters. I think of fatherhood in three stages, direct while they are small, advise while they grow, discuss once they are grown. By 2033 I will be down to discussing, and what I do with my days will be mine to decide again. That sentence describes a state, and nobody but me can check it. No one will stand at a line with a stopwatch. There is nothing in it to win and nothing to show, which means the cover stays on in 2033 as well. If it works, the only visible change will be that I decline some things more calmly. There is a risk in this that I watch closely. You can get used to a small life. What begins as a constraint turns into a habit, and one day the habit calls itself character. I try to treat these years as a phase with an end date. The cover is meant to outlast 2033, and the smallness is meant to end there. I’m not there. This year I took on work I should have handed to someone else, because paying for the someone felt like too much. Nobody forced that on me. I said no to the hire because the cost scared me. That is a no out of fear, from a man who claims to be working toward never saying one again. And after three years I have a feeling I did not expect: that this might go faster than planned. I have decided to distrust it. Three good years say little about ten, and a tailwind feels exactly like skill while it lasts. The dangerous version of me is the one who loosens the rules because the plan is ahead of schedule. How I would know I am wrong A story like this is easy to tell yourself, so here are the tests. * If the slice of the bills covered by yield has not grown through a full cycle, bear market included, there was never a flywheel, only a rising market. I will check at the halfway mark in 2028. * If the fifty-fifty rule gets its first exception on the first truly large windfall, it was a habit for small sums. * If I drop the cover on the day the proof is big enough, then “cover” was only another word for waiting. * If the cover costs me more than it protects, in chances that only visible people get, then “cheapest” is the wrong word. * If 2033 arrives and I still live as though every month were longer than the money, the phase has become the man, and the plan has failed in a way no account would show. And in 2033 the test is a single question: in the past year, was there a no out of fear or a yes out of need? If there was, the goal is missed, whatever the numbers say. There is one more reading that I can’t rule out. Invisibility can be avoidance. Whoever never enters cannot lose, and a man who calls his hiding a strategy has a comfortable story. I don’t think that is what this is. I would say that either way. Honest Janus Disclosure I am publishing an essay about staying unseen, under a pen name, to readers. The contradiction is plain. My answer is that the work can be visible while the person is not, and that this page contains no number anyone could use. Whether that answer holds is yours to judge. This is also a note from year three. I have no result to offer. If you are standing at your own ground zero, I can’t tell you it works. I can tell you what year three looks like: two rules, a desk at night, an old car, and a sentence with a date on it. To the reader in 2033 You know how it went, and you will be tempted to remember it as a straight line. Most of it was fog, and the road through it was paved with arithmetic, self-doubt and swallowed pride. Measure this page against the one sentence, and check whether the Passat, or whatever replaced it, still gets no second look. - J. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit janusthewatcher.substack.com (https://janusthewatcher.substack.com?utm_medium=podcast&utm_campaign=CTA_1)

2. 

   8 Oct 2026

   Too Risky for Capital, Too Thin for Farmers

   Disclosure: I staked from December 2025 and sold my direct stXRP in September 2026. I still hold fixed-rate positions in stXRP into 2027 and will divest when their holding period ends. I claimed my share of the pool. I received XDP in Doppler’s airdrop, sold it, and hold none today. 230,000 FXRP. That is what ten months of Firelight Phase 1 paid out, to a vault that holds about 48 million. Each million Firelight Points converted into roughly 60 cents. The number was published on October 7, one day after the protocol’s full launch. Phase 1 ended with a token that fits nobody who supplies the capital. stXRP is too risky for careful money and too thin for yield farmers. The cover product behind it is sound engineering, and I will come to that. This piece is about the other side of the ledger: the people whose XRP makes the cover possible. What Phase 1 was Phase 1 opened in December 2025. Users brought XRP to Flare as FXRP, deposited it into Firelight’s vault and received stXRP. No cover was being sold yet. Addresses collected Firelight Points instead, calculated from the amount held, the time held and a boost for deploying stXRP in selected DeFi protocols. On September 1, 2026 Firelight announced an $8 million seed roun (https://medium.com/sentora/why-we-built-firelight-defis-missing-layer-8cec2538c6a2)d led by Gumi Cryptos Capital and said cover would go live that month. It went live on October 6. Over those ten months the vault itself accrued about 3,500 FXRP on a balance that grew to 48 million. That is close to zero. Stakers who simply held stXRP were paid in points whose value nobody had stated, and in nothing else. rFLR went only to stXRP that was deployed in the Flare ecosystem, through protocols such as Kinetic, Spectra and Enosys. Those who deployed it earned more on both counts: selected pools paid double Firelight Points plus rFLR, for yield-token positions on Spectra and for liquidity providers, including professional vault strategies. A price nobody answered The market stated it for them, every day. On Spectra, stXRP could be split in two: a principal token that pays a fixed rate to maturity, and a yield token that collects whatever the underlying earns, points included. The fixed rate is paid by the buyers of the yield token. For most of Phase 1 that fixed rate could be locked at 4 to 6 percent. At times it ran above 10 percent, in thin pools with high slippage and heavy speculation. Toward the end it settled near 3 percent in the largest pool (https://app.spectra.finance/fixed-rate/flare:0x966d1f376457a3aca5fbc2a6be985f6e5e7708eb), the one maturing at the end of November 2026. With the vault earning nothing, the price was a bet on two things only: the points, and the yield after launch. The answer arrived on October 7. The Early Contributor pool is 230,000 FXRP, shared pro rata among eligible points. Measured against the vault, that is under half a percent of the capital for ten months of staking. The pool was split by points, and points had been doubled for exactly those boosted positions. A passive holder was diluted by the capital the boosts attracted. After launch the organic yield is 0.79 percent, by Firelight’s own interface. For most of Phase 1 the market had priced five to eight times that, and even at the end about four times. None of this was hidden from the team. Sentora, which incubated Firelight, was formed in 2025 from the merger of IntoTheBlock, an on-chain analytics firm, and Trident Digital. (https://www.theblock.co/news/deals/2025-05-06-intotheblock-trident-merger-institutional-defi-platform-sentora-25-million-usd-series-a-353220) Reading markets is its trade. On June 2, 2026 Connor Sullivan, Firelight’s chief strategy officer, sat on a livestream with Hugo Philion, co-founder of Flare, and Gaspard Peduzzi, co-founder of Spectra. Peduzzi described a single trade of about one million that locked in 10 percent fixed on XRP, and said the market was pricing “both the underlying yield and the points”. The team did more than watch that market. Yield-token positions in selected pools earned a double boost, which pushed capital toward the side that was betting on points. In July it announced that claims would open “around the end of August or Early September”. That window passed without comment. One published number, the size of the pool, would have corrected the price at any point. It came after launch and after the deposit cap had been raised. Three descriptions of the stXRP token Start with the terms every staker accepts in the app. Participation is “solely for technical and functional purposes, not for investment or speculation”. The staker confirms “no expectation of profit or financial return”. And stXRP “is issued solely as evidence of my deposit and confers no ownership, profit participation, income, or governance rights.” That receipt was wired into Flare’s DeFi from the start. It became collateral in lending markets, liquidity on exchanges, and the raw material for a fixed-income market on Spectra. Each use was promoted with point boosts, and the eligibility rules for the reward pool name the protocols: Spectra, Sparkdex, Enosys, Mystic. A receipt with a lending market, a rate curve and a marketing budget is a financial product by any working definition. Then the third description, from the people in charge. It came on two occasions. The first was the livestream of June 2. Asked about the points conversion, Sullivan said: “I think everyone’s going to be very happy with it.” The schedule, he added, would be published “in advance of mainnet launch, right? Not afterwards.” It was published the day after launch. The second was launch week. On October 7 Jesus Rodriguez, co-founder of Firelight and chief technology officer of Sentora, wrote that the protocol “is already paying over 2% APY on your XRP”. Philion told XRP holders they could “earn yield on your XRP” in one click. That 2 percent is itself a construction: 0.79 percent organic and 1.24 percent incentive, switched on at launch. By the team’s own account the incentive runs for three or four months. (See: The Path to Protected DeFi is Here (https://watch.getcontrast.io/watch/sentora-the-path-to-protected-defi-is-here) from Oct 7, 26) It is a fig leaf barely covering 0.79 percent, offered without blushing. Stakers were recruited by the third description and paid by the first. Too thin for farmers A yield farmer compares. Lending FXRP on Kinetic pays about 0.6 percent, with no slashing and no exit queue. Doppler pays about 3 percent on XRP. The fixed side of the stXRP market itself paid 3 to 6 percent. Against that list, 0.79 percent with slashing risk and a 30 to 60 day exit sits at the bottom. The buyers of yield tokens fared worst. They paid the fixed rate and received the organic yield plus points at 60 cents a million. The opportunity cost fits in one line. Ten thousand XRP held as stXRP for the ten months of Phase 1 earned about 55 FXRP from the reward pool. The same XRP in Doppler earned roughly 250 XRP at 3 percent, and qualified for the XDP airdrop on top. Measured on my own deposit, the airdrop sold for about 1.7 percent of the capital locked, roughly another 170 XRP per ten thousand. I have written critically about XDP’s design (https://janusthewatcher.substack.com/p/who-owns-the-scale-gain-thought-model). It still paid early depositors more than Firelight Points did. The premium is not the reason. Firelight’s analytics show $38.0 million of cover on $71.2 million of capital. An organic yield of 0.79 percent on that capital works out to roughly 1.5 percent of covered volume, which is a market rate for DeFi cover. The reason is leverage, or the lack of it. An insurer earns a return because its capital backs a multiple of what it insures. Here the capital backs about half its own size. Firelight’s documentation describes 2.0 to 2.5 times at launch. Its own risk model, faced with a book in which three quarters of the cover sits in one position, currently caps cover at about three quarters of capital. At 2.5 times and the same premium, stakers would earn about 3.6 percent. Until then the incentive pays for capital the insurer cannot use. The team said as much in its launch webinar: “we’re taking more collateral than we need.” Too risky for capital Firelight’s own pitch describes the capital it wants. The seed announcement, written by Rodriguez, speaks of XRP “held by people who deliberately opted out of DeFi’s risk frontier” and promises that capital “a job that matches its temperament: underwrite quantified, bounded risk for premium yield”. Measure the launch against that sentence. Read the position as someone who manages other people’s money. A first-loss buffer of $1.0 million stands in front of the stakers. Behind it, Firelight’s model puts the probable maximum loss at $17.1 million. Three quarters of the cover is on one chain. Leaving takes until the end of the following 30-day period, or a sale into a thin market. After a covered event, staked XRP is liquidated within 7 to 15 days. For carrying this the staker receives 0.79 percent, which on my reading of the protocol’s own figures is close to the modeled expected loss. A fiduciary cannot sign that. A conventional insurer that pays a claim then pursues whoever caused the loss, and what it recovers goes back to its capital. In DeFi, recoveries are common: funds are frozen, negotiated back or returned for a bounty. Firelight’s process ends at the payout. The consortium rules, stXRP is slashed, and nothing I have found says who owns a later recovery. The vault makes the question sharper. A slash hits whoever holds stXRP that day. Money returning months later would lift the rate for whoever holds it then. Careful capital has somewhere else to go. Doppler offers about 3 percent on XRP. Client assets sit in segregated custody at SBI Digital Markets, a unit regulated by Singapore’s Monetary Authority, and SBI Ripple Asia signed a memorandum of understanding with Doppler in December 2025 to develop yield products for institutions. The risk there is counterparty and strategy risk, and institutions have decades of practice at assessing a counterparty. Slashing from a concentrated cover book is new to them, and it pays a quarter as much. Firelight sells one risk class to everyone. A senior tranche, hit only after large losses and paid a modest rate, would be the institutional product. A junior tranche, taking first loss for a double-digit rate, would be the product for farmers who want to be paid for risk. The arithmetic only works above roughly 2.5 times leverage, and it needs cover written across several insurance lines, ideally highly uncorrelated ones, so that one event cannot reach the senior tranche. Neither condition holds today. Nothing stops the market from building those tranches on top; stXRP is composable enough. But at 0.79 percent there is nothing to split. Until that changes, a single blended class leaves a middle that suits neither group. Built for whom The webinar answered that directly: “Our clients are protocols and vault operators and curators. They’re not the end users.” Firelight was incubated at Sentora, its first covered vaults are Sentora’s, and Sentora has put up the first-loss buffer from its own balance sheet. For the buyer of cover the product is coherent. A small share of vault yield buys protection and a “protected” label. In fairness, the pool is real money. Firelight paid 230,000 FXRP out of its own funds, where most projects would have printed a token. It was also cheap. The pool comes to about 0.66 percent a year on the capital it rewarded, against 3 percent for XRP yield elsewhere. The rFLR on deployed positions came from Flare’s incentive programme, not from Firelight. I put the yield question to the team on X in launch week. Anthony DeMartino, chief executive of Sentora and a co-founder of Firelight, replied with a question of his own: “What’s your current return on xrp capital? Would love to deploy at 7-10%.” The market’s answer at the time was about 3 percent in Doppler and 4 to 6 percent fixed on Spectra. His own vault was paying 0.79 percent organic. Seen from the staker side, the sequence reads like a Kickstarter. Capital arrived from December. In September, nine months in, Firelight announced its $8 million seed round. The announcement runs to ten minutes of reading and does not mention the stakers already in the vault, or their points. In October those backers received 230,000 FXRP. They funded the proof, and the equity sits with others. I cannot know whether that was the plan, and the order of events does not prove it. On Kickstarter, though, backers know the terms before they pay. Asked about a token, the team answered “TBD”. A claim on future revenue for early capital would have aligned the two sides. It was not built. What it means for Flare For Flare this was the marquee project. Flare is a strategic investor in Sentora, as is Ripple. Its leadership promoted the launch, and Firelight was built into Flare’s Smart Accounts on day one, so that XRP holders could deposit from the XRP Ledger with a single signature. In my view Firelight was meant to be the main engine of XRPFi, and later of the same model for Bitcoin: lock the asset, earn a yield. The seed announcement names FBTC and FXLM as the next collateral. FXRP has other uses. It serves as collateral in lending markets, and that does not depend on Firelight. But the pull that was supposed to bring idle XRP on-chain was yield, and the yield on offer is 0.79 percent. In the five hours after the reward pool was announced, about 924,000 stXRP left the vault. For the near term I see that hope as broken. What it was not This was not a failure of engineering. The risk framework is more conservative than I expected: a capital adequacy ratio of 2.81, automatic limits on new cover as that ratio falls, and an independent consortium that rules on claims. The premium is fair. The analytics tab that appeared on October 7 publishes covered volume, capacity, concentration and the buffer. It answered questions I had asked a day earlier and corrected one of my own estimates. In the webinar the team stated the 60/40 split between incentive and organic yield plainly, and announced an institutional partner bringing tens of millions in cover. If that arrives, organic yield rises. What I take from it Three rules. * Points without a published pool size are worth zero until proven otherwise. * Promises attract capital in DeFi once; after that, capital reads the contract. * And a rate market on your own token is a measuring instrument. A team that promotes it also has to answer it. I am in favour of on-chain insurance. DeFi needs it, and this is a serious attempt. I will look again in a year, at four things: covered volume against staked capital, organic yield across several periods, the rules for the next points season published before it starts, and one claim settled cleanly. Until then my XRP sits in a cold wallet, where nothing can slash it. - J. Janus runs 1:1 Confrontation — sixty minutes, one decision, no follow-up. For people who carry responsibility and want their thinking taken apart before it costs them. janusthewatcher.substack.com/p/11-confrontation (https://janusthewatcher.substack.com/p/11-confrontation) One sentence is enough. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit janusthewatcher.substack.com (https://janusthewatcher.substack.com?utm_medium=podcast&utm_campaign=CTA_1)

3. 

   4 Oct 2026

   Costco on Crypto-Rails?

   Disclosures I hold XRP and FLR, and I am now moving staking rewards and new capital into tokens that pass the third question below. I hold options on HYPE but no spot. I received XDP in the Doppler airdrop and sold it; I have no position. I hold small positions in MORPHO and PYTH. Flare is the part of this piece where my own interest is largest, because the inflation-financed yield I describe there is yield I receive. Nothing here is a recommendation or financial advice. Do your own research. The question Doppler Finance listed its XDP token on Kraken, KuCoin, OKX and others on 28 September. Doppler is a serious operator: yield infrastructure on the XRP Ledger, a partnership with SBI Digital Finance, two years of product. Ten billion tokens, a genesis airdrop of one percent, and by the second day a price roughly seventy percent below the opening high. The launch post describes what a holder gets: staking for access to better vaults, a vote on protocol parameters and the incentive budget, and in a later phase a safety module in which staked XDP absorbs protocol losses in exchange for more XDP. It says nothing about the fees the vaults earn or who receives them as the platform grows. Forty-three percent of supply is set aside for ecosystem and incentives; that block sits inside a larger one, more than half of supply, reserved for ecosystem development, treasury initiatives, partnerships and community growth at the foundation’s discretion. Contributor and investor allocations are under lockups of unstated size. A quarter of a century earlier Nick Sleep and Qais Zakaria, running the Nomad fund in London, found the pattern that explains the XDP chart better than anything in the launch post. Some companies keep the savings that scale produces. Others hand them back to the people who buy from them. The second group looks worse every quarter and wins every decade; Costco and Amazon are its clearest cases, and when Sleep and Zakaria closed Nomad in 2014 they kept Amazon, Costco, Berkshire and ASOS in their own accounts. Sleep called the pattern scale economies shared. The phrase has circulated in DeFi investment circles since at least 2022, usually as a compliment paid to whichever protocol was in favour that quarter; what follows is an attempt to make it do work, as a test rather than a label. The question underneath it is the one I want to put to every token: when the network grows, who owns the increment? Crypto has a name for this, value capture, and uses it so loosely that it has stopped meaning anything. Sleep’s version is narrower and more useful, because it asks not whether value is captured but by whom. What Sleep meant, precisely Costco is not cheap because it is charitable. It caps the markup at fourteen percent on branded goods and fifteen on its own Kirkland label, regardless of what the market would bear, and membership fees of $5.9 billion supplied more than half of its $10.4 billion operating income in fiscal 2025. When purchasing power grows, prices fall, because the rule says the gain belongs to the member. Members renew at a rate just under ninety percent, buy more, and purchasing power grows again. Amazon ran the same logic with deferral: fifteen years of thin profit, spent on lower prices and shipping, and a message to suppliers that their margin was Amazon’s opportunity. Two features matter for the translation. Free was never the condition; both companies are enormously profitable, and what they refuse is the whole of the scale gain. And the sharing goes to customers, not to owners. Costco does not buy back its own shares with the savings. Shareholders are paid later, through the growth the customer-sharing creates. That distinction is the one crypto blurs, and it decides which tokens the Costco frame actually fits. A thought model in three questions On a network, the user’s price is not the fee. It is the fee plus slippage plus the probability of failure, and for any order that matters slippage is the largest term. So the Costco discount on-chain is depth, depth is bought with capital, and capital arrives where it is paid. That loop is the whole argument. Value accrual, read one step at a time, looks like a tax on users to pay holders; read as a loop, it is what makes a venue large and safe enough to be cheaper for everyone. What follows is a thought model, not a metric: three questions to put to any token, in order, each of which separates cases the previous one cannot. Where a question can be measured I say how; where I have not measured it I say so. The model is meant to be carried from one token to the next, which is more than most of the vocabulary in this market can claim. One. Do effective user costs fall as the network scales? If one wanted to measure it, the measure would be quoted spread and depth at a fixed order size, $100,000 notional on the main pair, tracked against volume over twelve months. In this piece the question is answered by inspection and marked as such; the first quarterly update will carry the series for the venues where it can be collected. A protocol can pass this test while charging a higher posted fee than its competitors. Two. Who receives the increment, and is the holder also the customer? The first half asks whether the take is capped by rule, expanded at will, or routed to a treasury. The second half is where the Costco frame either fits or does not. On a perpetuals venue the liquidity provider is customer and owner at once, and sharing with one is sharing with the other, at least on paper; a market maker that hedges or sells its tokens on receipt is a customer who has declined the ownership half. On an oracle or a data layer the customers are developers who need never hold the token, and a buyback there is owner-sharing with no customer effect at all. The frame fits venues imperfectly and infrastructure hardly at all. Three. Is what gets shared financed from revenue or from emission? Liquidity bought with inflation is a subsidised price, and subsidised prices leave with the subsidy. Revenue-financed depth stays because it pays for itself. The test is the ratio of what is actually shared with holders out of revenue to what is issued as emission over the same period. Revenue that sits in a treasury is not shared and does not count. One mechanism satisfies both halves of question two without needing the customer and the owner to be the same person: the fee discount for holders. Binance’s BNB discount is the oldest example and, structurally, the closest thing in crypto to a Costco membership card. The discount is granted at the point of sale, to the user, in proportion to what they hold. Hyperliquid runs the same mechanism alongside its buyback: staked HYPE cuts taker fees by five percent from ten tokens up to forty percent at half a million, reset daily. A buyback on its own is a weaker version of this, because it reaches the customer only if the customer happens to be an owner; a discount reaches every customer who chooses to become one. Scorecard The model applied to eight tokens, chosen because they span the range rather than because they flatter the thesis. Hyperliquid (HYPE) is the largest on-chain perpetuals venue. Derive (DRV) is an on-chain options exchange. BNB is the token of Binance, a centralised exchange, and is included because it is the oldest working example of the holder discount; its accrual is decided by the issuer, not by a contract, and the scorecard says so. Pyth (PYTH) is an oracle network. Morpho (MORPHO) is a lending protocol. Arbitrum (ARB) is an Ethereum rollup. Flare (FLR) is a data-oracle layer 1. XDP is Doppler’s token. Figures as of 29 September 2026; sources at the end. Four readings, not two. No revenue claim (XDP, and MORPHO by deliberate choice). Accrual to a treasury the holder governs but does not receive (ARB). A sharing rule written into protocol that revenue has not yet filled (FLR, PYTH). A rule that revenue fills today (HYPE, which shares with customers and owners alike; DRV; and BNB by the issuer’s discretion rather than by contract). The middle two are where the interesting decisions sit, and they are where a yes-or-no on fee sharing gives the wrong answer. Why a buyback is not a price cut A buyback returns the scale gain to owners. A Costco price cut returns it to customers, who respond by generating more scale. Only where the owner is the customer does a buyback do both jobs. Hyperliquid’s buyback on its own approximates Costco, and only approximates it, because many of its liquidity providers and heavy traders hold HYPE; strictly it is owner-sharing that the overlap between user and holder turns into something close to customer-sharing, and the overlap is partial, since market makers who hedge their inventory get none of it. What makes Hyperliquid the fullest case in the scorecard is that it does not rely on the overlap. The staking discount shares with the customer at the point of trade, the buyback shares with the owner from the same fee pool, and a trader who stakes gets both. That is the membership card and the dividend in one instrument. A buyback at an oracle does neither. It transfers value to holders and leaves the customer’s price untouched. This is why the reading for PYTH and FLR is “weak frame fit” even where the rule is well designed: the argument for their tokens has to run through developer adoption and data reliability, not through the Costco loop. Fee sharing still raises value rather than merely moving it, but only through the loop. Paid capital deepens books; deeper books lower the user’s real price; lower prices bring volume; volume raises fees. Where that chain is broken, because the holder is not the customer, fee sharing is redistribution, and a governance-only design may be the honest one. The FLARE case Flare is the clearest example of a rule written before the money arrived, and I hold it. In April the network passed FIP.16 with ninety-eight percent support. Inflation fell from five to three percent with a hard cap of three billion FLR a year. The base gas fee rose twentyfold and is burned in full. A pool named FIRE collects ninety percent of data-connector fees, minting fees and a tenth of redemption fees from FAssets, confidential-compute fees and, once the block-building changes ship, protocol-level MEV. Its stated purpose is to replace inflation with organic yield for validators and stakers and to buy and burn the remainder. Staked FLR rose from about sixteen billion to twenty-one and a half billion after the vote. Every element passes the first half of question two. The Dune dashboard for FIRE showed $50,003 collected on 29 September, since the pool went live in mid-May: $24,700 from FAsset minting, $22,800 from the data connector, about $2,000 from redemptions and a few hundred dollars from tag registrations. September’s pace is fifteen to twenty thousand dollars a month. The emission FIRE is meant to replace is capped at three billion FLR a year, roughly $22 million at the price implied by the dashboard’s own rows. Cumulative coverage is about 0.2 percent; on run-rate it is under one percent. The largest intended sources have not launched, so none of this is a failure of the design. It means the yield I receive as a staker is financed by inflation, and inflation-financed yield is rented. Flare will own its depth when FIRE pays for it. Ethereum ran the sequence in the other order. EIP-1559 landed in 2021 on a network that already had blockspace scarcity, so the burn was material from the first week. Flare has installed the rule before the customers. That may be the only way a smaller network can make the rule credible, but a rule is priced as a promise, and a promise is priced differently from a cash flow. The gap between XDP and FLR today is that Flare wrote the rule into protocol and made the number public. The gap between FLR and HYPE is that Hyperliquid’s number is large. Objections The frame is narrower than the essay. The overlap between user and holder is the exception. It holds for perpetuals, spot venues and lending, where liquidity providers are customers and owners. It fails for oracles, data layers and compute networks, where customers are developers. Two of the eight tokens in the scorecard, including the one I hold most of, are on the wrong side of that line, and the case for them has to be made on adoption, not on the Costco loop. Accrual is necessary and not sufficient. Lido had fees, category dominance and about a third of all staked ether, and its token fell more than ninety-eight percent from its high while the protocol kept running. When the underlying asset absorbs the re-rating, the layer above collects a small fraction of a large flow and gets forked the moment it raises its take. The three questions sort out the dead; they do not find the winners. Buybacks at the wrong price destroy value. Derive tripled in four trading days in mid-September and then proposed raising the buyback share to fifty percent. Every dollar of that programme now buys a third of what it bought in August. Sleep would have asked whether the same dollar, spent on tighter spreads, would have turned the loop harder than spending it on the price. No revenue claim is not always extraction. Three different things produce the same term sheet: a team that has not thought about it, a team that fears a regulator, and a team that intends the token as exit liquidity, and at launch they are indistinguishable from outside. MORPHO is the second case, and its fee switch may stay off for good if the regulatory risk is permanent. XDP could be any of the three; I can assess the structure, not the intent. ARB is a fourth thing: a claim on a treasury that earns and does not distribute, which is why it trades below what the treasury would suggest and above what its cash flow to holders, zero, would justify. The closed-end fund analogy is imperfect, since holders cannot force liquidation, but it explains the September rally better than any yield story does. What would prove this wrong Two baskets, frozen on 1 October 2026 and not rebalanced, equal-weighted at inception, total return in US dollars from CoinGecko daily closes, measured on 30 September 2028. Basket A, no revenue claim at inception: XDP, MORPHO, ARB. Basket B, revenue claim at inception: HYPE, DRV, BNB. The baskets are small because they contain only tokens analysed above; a larger test would need a larger scorecard, and I would rather run a thin test than a padded one. Risk-adjusted means the Sharpe ratio on daily returns, with maximum drawdown reported alongside. If MORPHO or ARB switches on distribution during the period it stays in Basket A; the essay is judged on what was knowable at inception, and the switch itself is recorded as evidence that the market demanded it. If Basket A beats Basket B on Sharpe over the period, the market prices product and optionality above cash flow, and this essay has isolated the wrong variable. Two years of crypto returns are dominated by the BTC and ETH cycle, so a loss for Basket B may be noise rather than refutation; the test is weak, and I run it anyway because a weak test stated in advance beats a strong one chosen afterwards. FLR and PYTH are excluded from both baskets because the essay makes no return claim about infrastructure tokens either way; the claim about them is that the Costco loop does not apply, and the test for that is whether their revenue ever funds their sharing rule, which is the one below. For Flare: FIRE inflow per quarter in dollars, from the Dune dashboard, divided by FLR emitted in the quarter times the quarter’s volume-weighted average price. That ratio rises when the FLR price falls, so a crash could clear the bar without anything improving. Two conditions therefore: the ratio above ten percent for two consecutive quarters, and the dollar inflow itself higher than in the quarter before. Ten percent is a judgement, not a derivation; below it the pool is noise against price moves in FLR, above it the pool could fund a visible fraction of staking yield. I state it now so it cannot move later, and I will publish both numbers quarterly. Consequences For a holder, the model gives the first question to ask of any token: which of the three it fails, and whether the failure is by design or by immaturity. That removes most of the market before a chart is opened. For a protocol, the lesson is Sleep’s: whatever scale gain you keep is available to the fork that shares it, and the rule you refuse to write is the rule your competitor launches with. And for anyone who has followed my longer argument about ownership of protocols as the answer to a world where capital earns and labour does not: a network that routes its surplus to the addresses that sustain it is that argument in miniature. - J. Janus runs 1:1 Confrontation — sixty minutes, one decision, no follow-up. For people who carry responsibility and want their thinking taken apart before it costs them. janusthewatcher.substack.com/p/11-confrontation (https://janusthewatcher.substack.com/p/11-confrontation) One sentence is enough. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit janusthewatcher.substack.com (https://janusthewatcher.substack.com?utm_medium=podcast&utm_campaign=CTA_1)

Contact Janus Dispatch Podcast

Guest appearances

Does not typically book guests

Based on episode analysis; this does not confirm that the show is currently accepting guests.