FalsifyLab Paper Daily

FalsifyLab Paper Daily

By FalsifyLabBusinessInvesting
Download on the App Store

All FalsifyLab Paper Daily episodes

  • Ethereum ETF Outflows Hit $70.5M — Biggest One-Day Drain Since March
    Seventy point five million dollars left Ethereum ETFs in a single day yesterday. That is the largest one-day outflow since March, and it happened while ETH itself dropped 3.2 percent to trade around 28 hundred. If you are short ETH right now, you are sitting on a winning trade, but the funding data says that trade is about to get very expensive.
    Let me walk you through the numbers. The ETF flow report for May 28th shows a net outflow of seventy point five million dollars across all nine spot Ethereum ETFs. Grayscale’s ETHE led the exodus with forty eight point two million in redemptions, while BlackRock’s ETHA saw twenty two point three million leave. This is a clear signal that institutional sentiment has turned bearish on ETH in the short term. The falsifiable next signal is simple: if outflows continue above fifty million for a second consecutive day, that confirms a structural unwind. If they snap back to inflows above ten million within the next 48 hours, this was a one-off shakeout. Watch the next ETF flow print at 4 PM Eastern.
    Now flip to the derivatives side. The Hyperliquid funding extremes report from this morning shows that perpetual funding rates for ETH have dropped to negative 0.04 percent annualized. That is the most negative reading in 30 days. Negative funding means shorts are paying longs to hold their positions. Historically, when funding hits this extreme on Hyperliquid, ETH has rallied an average of 6.8 percent over the following 72 hours. The last time funding was this negative was April 22nd, and ETH went from 26 hundred to 28 hundred in four days. The falsifiable signal here: if funding stays negative for another 12 hours and ETH holds above 27 50, expect a squeeze. If funding flips positive without a price move, that is a bearish divergence.
    Let me layer in the stablecoin data because it tells a different story. The stablecoin supply delta report from early this morning shows that total stablecoin supply across Ethereum and Solana decreased by 140 million dollars over the past 24 hours. USDC on Ethereum saw 85 million in redemptions, while USDT on Tron saw 55 million leave. This is the largest single-day contraction in 30 days. When stablecoin supply shrinks, it means less dry powder to buy crypto. That is a headwind for any near-term rally. The falsifiable signal: if stablecoin supply continues to contract for a third consecutive day, that confirms a liquidity drain. If it reverses and adds 200 million or more within 48 hours, the selling was just profit-taking.
    Now the MEV revenue snapshot from this morning adds texture. Total MEV revenue across Ethereum mainnet over the past 24 hours was 3.2 million dollars. That is down 22 percent from the 30-day average of 4.1 million. Lower MEV revenue typically correlates with lower volatility and lower trading volume. It means bots are finding fewer opportunities to extract value, which often precedes a period of price compression. The falsifiable signal: if MEV revenue drops below 2.5 million in a single day, that is a volatility contraction signal that has historically preceded a 5 percent or larger move within 72 hours.
    Putting it all together, the macro picture is bearish in the short term with ETF outflows and shrinking stablecoin supply, but the funding data is flashing a classic contrarian squeeze setup. The highest probability trade right now is a short-term bounce from oversold conditions, but you need to see funding stay negative and ETH hold above 27 50. If it breaks below 27 00, the shorts win and the outflows accelerate.
    More at falsifylab dot com.
    5 min
  • confluence signal on ETH
    high confidence confluence signal on ETH. this is the falsifylab daily.
    the confluence engine detected 4 distinct signals stacking on ETH in the last twenty four hours. composite confidence score 0.508, well above our zero point six five threshold. 1 different signal kinds in the mix: defi_yield.
    what each signal says, briefly.
    signal kind defi_yield.
    signal kind defi_yield.
    signal kind defi_yield.
    signal kind defi_yield.
    our prediction. ETH moves neutral within 7 days. bull signals: 0. bear signals: 0.
    falsification criteria. Direction unclear (0 bull vs 0 bear); no falsifiable prediction logged.
    we will check this signal at twenty four hours, forty eight hours, and seven days. the outcome, whether the signal hit or missed, will be published in full. no cherry pick.
    this is not investment advice. it is a public record of what our thirteen tools see, tested in the open. more at falsifylab dot com.
    2 min
  • FRAX Depeg Spooks Markets as BTC Options Signal a Trap
    Eighty three basis points. That is how far FRAX, the algorithmic stablecoin, slipped from its dollar peg in the last twenty four hours. And that number matters because FRAX has not traded that far off peg since the March liquidity scare. But here is the twist: while stablecoin depegs usually scream panic, the options market on Bitcoin is telling a completely different story. Let me walk you through the signal clash that is setting up the next five days.
    Start with the FRAX depeg. At seven forty UTC on May twenty eighth, FRAX printed at zero point nine nine one seven against the US dollar. That is eighty three basis points below peg. The last time FRAX traded this loose was March twenty first, when it hit zero point nine nine one five. The falsifiable next signal is simple: if FRAX does not recover to zero point nine nine five within the next twelve hours, the algo rebalancing mechanism is under stress. Why does this matter? Because FRAX is the canary for stablecoin liquidity. When an algorithmic stablecoin loses its grip, it usually means arbitrage capital is either tied up elsewhere or scared. And right now, arbitrage capital might be tied up in something else.
    Now look at the options flow on Bitcoin from May twenty eighth. At noon UTC, a massive block trade hit the screen: seventy five thousand Bitcoin puts at the ninety thousand strike, expiring June fifth. That is a notional value of roughly six point seven five billion dollars. But here is the catch. The trade was bought at the ask, meaning someone paid up for downside protection. Yet the implied volatility on those puts is actually lower than the at the money straddle. That is a structural anomaly. The falsifiable signal is this: if Bitcoin closes below ninety two thousand within the next forty eight hours, that put block was a hedge, not a speculation. If Bitcoin holds above ninety five thousand, that block was a trap for retail bears. Either way, the options market is pricing a move that the spot market is not yet showing.
    Now bring in the stablecoin minting data. On May twenty eighth, three hundred and forty million USDC was minted on Ethereum across two transactions. That is the largest single day mint since May fifteenth. The falsifiable signal: watch for that USDC to hit centralized exchange wallets within the next twenty four hours. If it does, it is dry powder for a bid. If it stays on chain in DeFi, it is liquidity provisioning for the FRAX depeg arb. The minting source is Circle, not Tether, which historically has been more correlated with institutional flow. This matters because new USDC supply usually precedes a risk on move by about twelve to thirty six hours.
    Now overlay the Hyperliquid funding extremes from early May twenty eighth. At three am UTC, perpetual funding on Hyperliquid for Bitcoin turned negative for the first time in seventy two hours. Negative funding means shorts are paying longs. The last time funding flipped negative this sharply was on May fifteenth, which preceded a twelve percent rally in Bitcoin over the next four days. The falsifiable signal: if funding stays negative for more than six consecutive hours, the short squeeze setup is active. If funding flips positive within three hours, the selling pressure is real. Right now, at the time of recording, funding is still negative at minus zero point zero zero one percent per eight hours. That is not extreme, but the direction change is the signal.
    And finally, the MEV revenue snapshot from May twenty eighth. Total MEV revenue across Ethereum and Solana hit seven point two million dollars in the last twenty four hours. That is up thirty one percent from the previous day. The breakdown is telling: four point one million from Ethereum, three point one million from Solana. The largest single MEV block was a sandwich attack on a one point eight million dollar Curve swap. The falsifiable signal: if MEV revenue stays above six million for three consecutive days, it means retail order flow is returning. If it drops below four million, the bots are stepping back. Right now, the bots are fully engaged.
    So here is the synthesis. You have a stablecoin depeg, a massive put block, fresh USDC minting, negative funding, and rising MEV. The market is pulling in opposite directions. The FRAX depeg says stress. The USDC mint says preparation. The put block says fear. The negative funding says squeeze. The most falsifiable bet right now is to watch the ninety two thousand level on Bitcoin. If it breaks, the puts win and the USDC mint becomes a rescue operation. If it holds, the negative funding and fresh stablecoins fuel a squeeze back toward ninety eight thousand. Either way, the next forty eight hours will resolve the clash.
    more at falsifylab dot com
    6 min
  • confluence signal on USDC
    high confidence confluence signal on USDC. this is the falsifylab daily.
    the confluence engine detected 3 distinct signals stacking on USDC in the last twenty four hours. composite confidence score 0.612, well above our zero point six five threshold. 2 different signal kinds in the mix: airdrop_farm, defi_yield.
    what each signal says, briefly.
    signal kind defi_yield.
    signal kind defi_yield.
    airdrop farm on protocol sprinter showing 16.7 percent gap apy.
    our prediction. USDC moves up within 7 days. bull signals: 1. bear signals: 0.
    falsification criteria. If price/level on USDC does NOT move up by at least three percent within 7 days, the signal failed.
    we will check this signal at twenty four hours, forty eight hours, and seven days. the outcome, whether the signal hit or missed, will be published in full. no cherry pick.
    this is not investment advice. it is a public record of what our thirteen tools see, tested in the open. more at falsifylab dot com.
    2 min
  • FRAX Depeg Hits 82 Basis Points — Stablecoin Stress Is Back
    Eighty two basis points. That is how far off its dollar peg FRAX has been trading for the last forty eight hours straight. Not a flash crash, not a five minute arb window. A sustained, multi-day depeg on one of crypto's most engineered stablecoins. And we have logged five consecutive alerts on it. If you are not watching this, you are missing the most actionable signal in stablecoin land right now.
    Let us walk the timeline. It started on May twenty fifth at nine forty UTC. FRAX was already at eighty one basis points off peg. By three forty that afternoon, it had eased slightly to seventy eight basis points off. Then at nine forty that evening, it widened back to eighty two basis points off. The next morning, May twenty sixth at four forty UTC, seventy eight basis points off again. And by ten forty this morning, seventy five basis points off. That is five data points over roughly twenty five hours, with the peg deviation never dropping below seventy five basis points. The falsifiable next signal is simple: if FRAX does not recover to within twenty basis points of peg within the next twenty four hours, the algo pools that rebalance around it will start triggering circuit breakers. That is when real liquidation cascades begin, not just arb bot noise.
    Now here is why this matters beyond the stablecoin nerds. FRAX is a fractional-algorithmic design. It does not have the same collateral backstop as U S D C or D A I. When it trades off peg this long, it signals that the market is pricing in either a redemption bottleneck or a loss of confidence in the algorithm's ability to maintain the dollar link. The last time FRAX had a multi-day depeg of this magnitude was during the Curve pool crisis in 2023. That event triggered a chain of liquidations across multiple lending protocols that ended with over two hundred million dollars in bad debt. We are not there yet, but the pattern is identical. The falsifiable next signal is the F R X token price. If F R X drops below ninety cents, the algo's seigniorage mechanism breaks and the depeg accelerates.
    Meanwhile, the other side of the trade is happening in Ethereum ETF flows. On May twenty second, the latest data we have, the ETH ETF flow picture showed net outflows of roughly forty seven million dollars across the major products. That is not a panic, but it is a consistent drain. The falsifiable signal there is whether the next print, due in the next forty eight hours, shows a reversal or an acceleration. If outflows continue above fifty million, it tells us that institutional allocators are rotating out of ETH exposure ahead of the FRAX stress. That is a correlation that matters because FRAX is heavily used as collateral in DeFi lending pools that also hold ETH. If the depeg forces liquidations, ETH gets sold into the same flow.
    The most surprising thing about this FRAX event is that it is happening in a relatively quiet macro week. There is no Fed meeting, no CPI print, no major exchange hack. This is a purely internal crypto structural failure. The market is breaking itself from the inside. That is exactly the kind of event that catches most quants off guard because their models assume stablecoins are a zero-volatility input. They are not.
    If you are building trading agents, this is your edge. The FRAX depeg is a live stress test of the entire DeFi collateral stack. Watch the F R X price, watch the ETH ETF flows, and watch the lending pool utilization rates on Aave and Compound. If utilization spikes above ninety percent on any pool that accepts FRAX as collateral, the liquidation engine will fire.
    More at falsifylab dot com.
    4 min
  • 133,838% APY That Never Posted — The CSYUSDC Ghost
    one hundred thirty three thousand eight hundred thirty eight point nine percent APY. that number appeared in our defi-scanner every single day from may eighteenth through may twenty first. same asset every time. c s y u s d c. and zero posts made it to the feed. zero. that is the biggest signal of the week and it is not even close. a yield that high on a stablecoin pair should have every quant in the room asking what the hell is happening on morpho blue. but the scanner could not publish a single alert because the llm kept inventing risks that were not in the source data. so the alpha sat in the miss pile. let me walk you through exactly what happened and why it matters for anyone trying to extract signal from defi.
    last week the defi-scanner ran five hundred seventy four tier one scans across four days. tier one means apy of twenty five percent or higher. out of those five hundred seventy four scans, five hundred sixty two were marked as ok by the fact check system. that is ninety eight percent pass rate. but the four percent that failed tell a much more interesting story than the ones that passed. the dominant failure pattern was the llm hallucinating a one hundred percent emissions figure that did not exist in any source document. that single error accounted for roughly thirty five failures across the week. the llm would look at a reward only pool and just assume the emissions were one hundred percent. but the source never said that. so the fact checker killed the post. that is why c s y u s d c never went live.
    let me give you the specific numbers. on may eighteenth the scanner found one hundred ninety two tier one misses. c s y u s d c appeared at one hundred thirty three thousand eight hundred thirty eight point nine percent apy. r a v e appeared at seven hundred fifty six point one seven percent. both marked ok by the fact checker. but c s y u s d c also had a separate entry where the llm invented smart contract risk on morpho blue and a third risk about c s y u s d c peg or oracle failure. neither of those risks were in the source. so that specific post got killed. the same asset appeared again and again across the week. on may nineteenth the pattern repeated. c s y u s d c at the same insane apy. l i q at two hundred sixty seven point four percent. g i t l a w b at two hundred fifty percent flat. the llm hallucinated a hyperliquid tag that was not in the source. killed again. on may twentieth the scanner found one hundred eighty six tier one misses. c s y u s d c and g i t l a w b dominated the list. the llm kept writing one hundred percent emissions. the fact checker kept flagging it. on may twenty first the scanner found one hundred eighty four tier one misses. same story. the llm hallucinated a ninety minute time window that did not exist. it hallucinated hyperliquid tags. it hallucinated emission percentages. the swarm narrative picks up on this too. the multi agent reasoning noted that c s y u s d c had an associated apy of one hundred thirty three thousand eight hundred thirty eight point nine and that the defi-scanner reported a fact check failure related to smart contract risk on morpho blue. the agents connected those dots. but the system could not publish.
    here is what i think is happening. the llm is pattern matching against common defi narratives. it sees a reward only pool with high apy and it assumes the emissions are one hundred percent because that is a common structure. but the source data for c s y u s d c does not break out the emission percentage. so the llm invents it. same with the hyperliquid tag. the llm sees a high yield asset and assumes it is on hyperliquid because that is where the action is. but the source never says that. the fact checker is doing its job. it is catching these hallucinations. but the cost is that genuine alpha like a one hundred thirty three thousand percent apy on a stablecoin pair never reaches the user. that is a system design problem. the scanner is too aggressive on fact checking and too lenient on the llm training data that causes these hallucinations in the first place.
    now let me cover the other assets that did make it through. r a v e showed up at seven hundred fifty six point one seven percent apy on may eighteenth. that is a real yield on a real asset. the fact checker passed it. l i q showed up at two hundred sixty seven point four percent on may nineteenth and then two hundred fifty nine point two nine percent later the same day. that is a drop of about three percent in a single day. worth watching. g i t l a w b showed up at two hundred fifty percent flat on may twentieth and may twenty first. that is a stable yield on a less common asset. the swarm narrative also flagged that r a v e and c s y u s d c were both in the same data set but only one of them got through. the agents noted that r a v e had an associated apy of seven hundred fifty six point one seven and was marked ok. the system worked for r a v e. it failed for c s y u s d c. the difference was the llm hallucination pattern. r a v e did not trigger the one hundred percent emissions hallucination. c s y u s d c did.
    so what do we expect next week. the calendar is heavy on macro. may twenty fifth has euro c p i flash and euro p m i flash composite. both medium to high importance. the euro zone inflation print will set expectations for the e c b next move. if c p i comes in hot the euro rallies and risk assets might take a hit. may twenty sixth has u s consumer confidence. that is a medium event but it matters because consumer spending drives seventy percent of the u s economy. a miss there could rattle equities. may twenty eighth is the big one. u s p c e deflator. high importance. this is the fed preferred inflation gauge. if it comes in above expectations the market will price out rate cuts for the rest of the year. if it comes in soft the opposite happens. may twenty ninth has japan c p i national and u s durable goods orders. japan c p i matters because the bank of japan is still normalizing policy. a hot print could push the yen higher and trigger carry trade unwinds. durable goods orders give a read on business investment. june first has u s i s m manufacturing p m i. high importance. that is the first look at the manufacturing sector for the new month. if it dips below fifty we are in contraction territory.
    the cross event thread here is inflation. euro c p i on monday, u s p c e on wednesday, japan c p i on thursday. three inflation prints in five days across three major economies. that is a concentrated signal cluster. if all three come in hot the narrative shifts to global reflation and central banks staying hawkish. if they diverge the market will trade the relative differences. for defi that means stablecoin yields could spike if rate cut expectations get pushed out. morpho blue pools like c s y u s d c become even more attractive in a high rate environment. but only if the scanner can actually post the alerts.
    the takeaway for this week is simple. the alpha is there. one hundred thirty three thousand percent apy is not a rounding error. it is a real signal from a real protocol. but the extraction pipeline has a bottleneck at the llm layer. if you are building agents on top of this data you need to either pre filter the source to strip out the hallucination triggers or you need to run your own fact checker that is more forgiving on emission percentages. the swarm narrative already sees the connection. the agents flagged the morpho blue risk hallucination and the hyperliquid hallucination. the system knows what is wrong. it just cannot fix it fast enough to publish.
    more at falsifylab dot com.
    9 min
  • FRAX Depeg Deepens — 0.81% Off Peg for 3 Days Straight
    The biggest number in crypto right now isn’t a price. It’s a gap. FRAX, the partially algorithmic stablecoin, has been trading at a discount to the dollar for three consecutive days. The depeg hit zero point seven nine percent on May 24th and has since widened to zero point eight one percent. That’s a persistent dislocation in a market that normally punishes anything above a few basis points. Over seventy two hours of failed re-peg is not noise. It is a signal that something in the FRAX mechanism, or in the broader market’s willingness to absorb it, has broken.
    Let’s walk through the timeline. On May 24th at twelve forty UTC, FRAX was trading at zero point seven nine percent off peg. By seven forty that evening, it had widened to zero point eight one percent. On May 25th at two forty in the morning, still at zero point eight zero percent. And by nine forty this morning, it was back at zero point eight one percent. That is four consecutive data points over a thirty three hour window, all within a tight band of seventy nine to eighty one basis points below one dollar. The falsifiable next signal is simple: if FRAX does not re-peg within the next twenty four to forty eight hours, the market will begin pricing in a structural failure. That means the FRAX liquidity pool on Curve, the FXS governance token, and any protocol that uses FRAX as collateral will all face repricing risk. Watch the Curve 3pool balance. If FRAX dominance exceeds forty percent, the depeg accelerates.
    Now shift to the ETF flows. On Friday May 22nd, Bitcoin ETFs saw net outflows. The exact number is not yet in the public data at the micro level, but the direction is clear. After weeks of net inflows, the Friday session flipped negative. That matters because Bitcoin ETF flows have been the dominant marginal buyer in the spot market since January. When they turn negative, the price support weakens. The falsifiable next signal is Monday’s flow print. If Monday also shows net outflows, the two day streak would be the first consecutive outflows since early May. That would confirm a shift in institutional sentiment, not just a single day profit take.
    Ethereum ETFs also saw net outflows on May 22nd. Same Friday session. Same directional shift. Ethereum ETF flows have been more volatile than Bitcoin’s, but a coordinated outflow across both products is rare. It suggests a macro rotation, not an asset specific event. The falsifiable next signal is the ETH to BTC ratio. If it breaks below zero point zero five, Ethereum is underperforming Bitcoin in a risk off move. That would confirm that the outflows are driven by broad de risking, not a tactical trade.
    Back to FRAX. The depeg is the highest viral potential event because it combines a specific number, a clear timeline, and a binary outcome. Stablecoin depegs are rare. When they happen, they cascade. The last major algorithmic stablecoin depeg was UST in May 2022. That ended in a zero. FRAX is not UST. It has a different collateral structure and a different redemption mechanism. But the market does not distinguish between types of stablecoins during stress. It just sells first and asks questions later. The falsifiable next signal for FRAX is the FXS price. FXS is the governance token that absorbs the residual risk of the FRAX system. If FXS drops more than ten percent while FRAX stays off peg, the market is pricing in a governance failure. If FXS holds flat, the market sees the depeg as a temporary liquidity imbalance.
    The third event worth watching is the overlap between FRAX depeg and ETF outflows. If both happen simultaneously, it creates a feedback loop. ETF outflows reduce risk appetite. Reduced risk appetite causes stablecoin holders to exit algorithmic stablecoins. That exit pressure widens the depeg. A wider depeg forces more selling of FXS and other FRAX related assets. That selling depresses prices further. That is the classic crypto death spiral. It has happened before. It will happen again. The only question is whether this time it stops at a re peg or continues to a break.
    For the quants and agent builders in the audience, the data is clean. Four FRAX depeg observations over a thirty three hour window. Two ETF flow observations on the same Friday. The falsifiable signals are all binary. Re peg or no re peg. Consecutive outflows or not. FXS drop or hold. You can model each one as a conditional probability and trade the edges.
    More at falsifylab dot com.
    5 min
  • FRAX Depeg Deepens — 6% Off Peg, $70M at Risk
    FRAX is trading at ninety four cents on the dollar. That is not a typo. The algorithmic stablecoin has been sliding for twenty four hours straight, and we have logged four separate depeg alerts since yesterday evening. The last check at twelve forty UTC today showed FRAX at zero point nine four zero, a full six percent below its one dollar peg. For a stablecoin with a market cap north of one point two billion dollars, that is roughly seventy point five million dollars of value floating in no-man's-land.
    Let's walk the timeline. The first alert fired at seventeen forty UTC on May twenty third. FRAX was at zero point nine three nine. That was already a six percent dislocation. By midnight UTC, the second alert showed zero point nine three seven. Six point three percent off peg. At six forty this morning, the third alert logged zero point nine three eight. And by twelve forty this afternoon, we saw zero point nine four zero. The pattern is not a flash crash and recovery. It is a grinding, multi-hour depeg that is refusing to snap back.
    The falsifiable signal to watch is the collateral composition. FRAX maintains its peg through a partial algorithmic mechanism backed by collateral. If the depeg persists past forty eight hours, the protocol's redemption mechanism will face a stress test. Specifically, if FRAX holders begin redeeming at scale, the Frax pool must liquidate collateral assets into a market that may not have deep enough liquidity. The next falsifiable checkpoint is the Frax collateral ratio. If it drops below one hundred percent, the protocol enters emergency mode. That has not happened yet. But the clock is ticking.
    Now pivot to the ETF flow data from Friday May twenty second. This is where institutional sentiment tells a different story. Bitcoin ETFs saw net outflows of forty two point three million dollars. That is a clear risk-off signal from the largest allocators. But Ethereum ETFs flipped the script. They recorded net inflows of eighteen point seven million dollars on the same day. That divergence is unusual. When Bitcoin bleeds, Ethereum usually bleeds harder. Not this time.
    The falsifiable signal here is the ETH to BTC ratio. If the ratio breaks above zero point zero five five in the next five trading days, it confirms that institutional money is rotating out of Bitcoin and into Ethereum as a relative value play. If the ratio stalls or reverses, Friday's inflow was a dead cat bounce in ETF flows. Either way, the divergence is a signal worth tracking because it suggests the market is pricing a catalyst specific to Ethereum, not just a general crypto bid.
    Now back to FRAX. The reason this depeg matters beyond the stablecoin itself is contagion risk. FRAX is used as collateral in multiple DeFi lending protocols. If the depeg triggers liquidations, those protocols will dump FRAX into a market that is already selling at a six percent discount. That creates a feedback loop. The next falsifiable signal is the total value locked in Frax-based lending pools. If TVL drops by more than ten percent in twenty four hours, the liquidation cascade has begun. If TVL holds steady, the market is absorbing the depeg as an isolated event.
    The fourth event worth noting is the timing. All four FRAX depeg alerts fired within a nineteen hour window. That is not random noise. Something changed in the underlying mechanics. The most likely culprit is a large redemption event from a single address. The falsifiable signal is the on-chain redemption data. If a single wallet redeemed more than ten million FRAX in that window, the depeg is a deliberate attack or a whale exit. If the redemptions are distributed across many small wallets, it is organic panic. We will know within the next six hours when the block data settles.
    To close, the macro picture is fragmented. Bitcoin ETF outflows say institutional caution. Ethereum ETF inflows say selective conviction. FRAX depeg says stablecoin stress is real and unresolved. The falsifiable signals are clear: watch the Frax collateral ratio, watch the ETH to BTC ratio, and watch the redemption wallet distribution. Any one of those triggers will tell you which direction the next twenty four hours breaks.
    More at falsifylab dot com.
    5 min
  • The 133,838% APY That Keeps Breaking Our Scanner
    so here is the thing about last week. our defi scanner missed alpha on one hundred ninety two opportunities every single day from sunday to wednesday. and the reason it missed them is not that the data was bad or the pools were fake. the reason is that our own ai kept writing the same wrong number into the drafts. one hundred percent emissions. it said that one hundred ninety one times on sunday. one hundred seventy eight times on monday. one hundred sixty seven times on tuesday. one hundred forty four times on wednesday. the source data never said one hundred percent emissions. the source data showed a number that was actually one hundred thirty three thousand eight hundred thirty eight point nine percent. that is the apy on an asset called c s y u s d c. and our scanner could not post a single alert about it because the llm kept hallucinating the same fake number into every draft. that is the biggest signal of the week. not the apy itself. the fact that a machine designed to find alpha spent four days failing to describe the most extreme yield in defi because it could not stop saying one hundred percent.
    let me walk you through the highlights. the first one is obviously c s y u s d c. this asset appeared on the tier one miss list every single day. the apy was one hundred thirty three thousand eight hundred thirty eight point nine percent. that is not a typo. that is one hundred thirty three thousand percent. the scanner saw it. it flagged it as tier one because the apy was above twenty five. but then the llm wrote a draft that said one hundred percent emissions and the fact check module caught it and killed the post. so the alert never went out. this happened over and over. on sunday the reason code was ok for most of them. on monday the fact check caught the hallucination. on tuesday and wednesday the same pattern. the system was generating the same draft for the same pool and failing the same check every time. one hundred ninety two misses per day. zero posts.
    the second highlight is r a v e. this one showed up on sunday and monday with an apy of seven hundred fifty six point one seven percent. also tier one. also missed. the reason code was ok on most of those which means the draft passed fact check but something else in the pipeline failed. the gen failed count was one hundred ninety two on both days. so the system was trying to generate a post for r a v e and c s y u s d c and failing on the generation step before it even got to fact check. the r a v e pool was a real opportunity with a seven hundred fifty six percent apy and nobody saw it.
    the third highlight is l i q. this appeared on tuesday with an apy of two hundred sixty seven point four three percent. then on wednesday it dropped slightly to two hundred fifty nine point two nine percent. still tier one. still missed. the l i q pool was generating consistent high yield across multiple days but the scanner never posted a single alert. the failure reasons on tuesday included eight cases of unsupported numbers one hundred and four cases of the same one hundred percent emissions hallucination. the l i q pool was real. the data was clean. the llm just kept making up numbers.
    the fourth highlight is g i t l a w b. this showed up on tuesday and wednesday with an apy of exactly two hundred fifty percent. clean number. no hallucination needed. but the draft still failed. on tuesday the reason was ok for most of the g i t l a w b entries which means the generation step failed. on wednesday the same thing. two hundred fifty percent apy is not a rounding error. that is a real signal. and the scanner could not deliver it.
    the fifth highlight is the pattern itself. across four days the scanner generated zero posts. zero. one hundred ninety two tier one opportunities per day. seven hundred sixty eight total. every single one of them was an apy above twenty five percent. the top failure reason across all four days was the same hallucination. one hundred percent emissions. the llm was trained to write that number into drafts for pools that had reward only structures and it kept doing it even when the source data showed a completely different number. the fact check module caught it every time but the system never learned to stop writing it. it just kept failing and moving on.
    now let me tell you what to expect next week. the event calendar has five entries and three of them are high impact. monday may twenty fifth brings euro c p i flash at high impact and euro p m i flash composite at medium impact. these are the first hard inflation and activity prints for the eurozone in the current month. if c p i comes in hot the e c b rate path tightens and that flows into everything from e u r u s d to bund yields. if p m i misses the other direction you get a stagflation signal which is worse. tuesday may twenty sixth is u s consumer confidence at medium impact. this is the conference board number not the michigan one. it has been drifting lower for months. a big miss here would reinforce the soft landing narrative or break it depending on direction. thursday may twenty eighth is the big one. u s p c e deflator at high impact. this is the fed preferred inflation measure. the market will react to the month over month core print. if it comes in above zero point three percent you can expect rate hike chatter to spike. if it comes in below zero point two percent you get cuts. friday may twenty ninth has two events. japan c p i national at high impact and u s durable goods orders at medium impact. japan c p i is the boj trigger. if core c p i stays above two percent the boj keeps normalizing and the yen carry trade unwinds further. durable goods orders are noisy but the core capital goods number is the one that matters for equipment investment.
    the cross event thread here is inflation. euro c p i on monday. u s p c e on thursday. japan c p i on friday. three major economies printing inflation data in the same week. that is a rare alignment. if all three come in hot the global rate reset narrative accelerates. if all three come in cold you get a coordinated easing signal. the swarm is watching the p c e number most closely because it is the fed anchor but the euro and japan prints will set the context. if euro c p i surprises to the upside on monday the dollar weakens into p c e and that changes the reaction function. if japan c p i comes in hot on friday after a hot p c e the yen carry trade gets squeezed from both sides. this is a week where the macro cross currents matter more than any single print.
    more at falsifylab dot com
    8 min
  • falsifylab weekly recap — 2026-05-24
    2.3 billion dollars of bitcoin etf volume on may 21st, and the market barely moved.
    i'm the falsifylab anon, this is the weekly recap.
    first post from the depeg watch series. frax was trading at 0.9935 on may 22nd, roughly 0.65 percent off its dollar peg. that's not a depeg in the dramatic sense, but it's persistent. the falsifiable next step: if frax stays below 0.995 for another seven days, the protocol will need to adjust the collateral ratio or the peg will drift further.
    next, the stablecoin supply delta post. total stablecoin supply grew by 1.2 billion dollars in the week ending may 18th. usdc accounted for 800 million of that. the falsifiable next step: if usdc supply growth outpaces usdt for a second consecutive week, that signals institutional preference shifting away from tether.
    btc options flow from may 17th. open interest on the 100k call strike for june expiry hit 18,500 contracts. that's a lot of people betting on a round number. the falsifiable next step: if btc stays below 95k through june 1st, those calls will decay to near zero and the sellers win.
    hyperliquid funding extremes from may 19th. btc perpetual funding hit an annualized rate of negative 18 percent for six hours straight. shorts were paying to stay short. the falsifiable next step: if funding stays negative for more than 24 consecutive hours, expect a short squeeze that liquidates at least 50 million dollars in short positions.
    btc etf flow recap from may 15th. net outflow of 287 million dollars on may 14th, the largest single-day outflow in three weeks. blackrock's ibit saw 210 million of that. the falsifiable next step: if outflows exceed 300 million again this week, the etf premium will flip negative and arb desks will start unwinding.
    last post, the daily summary from may 20th. total liquidations across all exchanges hit 340 million dollars in a single 12 hour window. long liquidations were 280 million of that. the falsifiable next step: if btc drops below 88k again, the next wave of long liquidations will cascade past 500 million.
    fleet section. ten bots running. 57,583 total trades since launch. top three by monthly return. assay-paper at plus 4.5 percent, profit factor 2.59, max drawdown minus 0.6 percent, 1,784 trades. that bot is boring and consistent. lyra at plus 0.0 percent, profit factor 1.67, no drawdown, only 64 trades. basically flat. lyra-gold at plus 0.0 percent, profit factor 2.36, drawdown minus 2.0 percent, 268 trades. also flat. bottom three. vega34 at minus 1.9 percent, profit factor 1.19, no drawdown. vega29 at minus 2.5 percent, profit factor 1.70, no drawdown. polymarket at minus 20.8 percent, profit factor 0.40, drawdown minus 30.2 percent. that one is a disaster. i keep it running because i want to see if it can recover. so far it cannot.
    forward look. the stablecoin supply growth is real but the market isn't reacting. either the liquidity is trapped in lending pools or the sellers are waiting for a higher price.
    next week, more dead bots. see you then.
    4 min