Shaneka AneIam Perare ⚡️
67 posts
Shaneka AneIam Perare ⚡️
@robnadeem
Author of The Ascent Begins. Independent Analyst. Money, geopolitics, AI, science, and sovereignty. Mapping the collapse and the reconstruction of order.
Katılım Eylül 2011
34 Takip Edilen32 Takipçiler
Classical objectivity is not a view from nowhere. It is a fact no observer owns because the environment has copied its stable record into enough places for strangers to agree without touching the source.
English
A route that carries 1.5 million barrels of oil a day shut down this week. Nothing on it was damaged though.
The fields are fine. The 1,510 KM pipeline is fine. The offshore loading moorings are fine. It stopped because tanker bosses stopped sending ships.
Four tankers were hit in just four days at the Caspian Pipeline Consortium terminal near Novorossiysk. Nordic Zenith on 17 July. Asia and Nissos Ios on 19 July, struck while loading. Nelsa on 20 July, which caught fire. Its crew of 22 came off by tugboat. The captain and chief officer stayed aboard.
Then the owners stopped coming. Bloomberg reported the terminal had stopped accepting oil from the pipeline because tanker companies were too nervous to send vessels. The storage tanks had filled. CPC moves roughly 80 percent of Kazakhstan's exports and has averaged 1.48 million barrels a day this year.
Watch which way the failure travelled.
No ship, so no loading. No loading, so the tanks fill. Full tanks, so the pipeline cannot deliver. The next thing to give is production at the wellhead, and no producer has announced a cut yet.
The shortage moved backwards from the sea to the oilfield through equipment nobody touched.
Same week, Red Sea. After the Yemeni Houthis backed by IRGC banned calls at Saudi ports, five tankers changed course on Wednesday, two turning for Suez. Rerouting a Yanbu cargo bound for Asia adds roughly 10,000 nautical miles and 34 sailing days. Early Thursday the Houthis said they had struck two Saudi-flagged tankers. UKMTO confirmed one vessel hit by an unknown projectile 70 nautical miles off Al Shuqaiq, with a fire aboard, and did not name the ship.
Same week, Hormuz. Kpler counted nine crossings on 21 July, down 31% in a single day.
Three waterways. Two unrelated wars. Iran is not fighting Ukraine. There is no shared command and no shared cause.
They have converged on the same target anyway, and it is not infrastructure. It is the ship.
The obvious objection is that this is an ordinary wartime risk premium, and that risk-tolerant tonnage always fills the gap. Sanctioned tankers carry cargo nobody else will touch, for a price. That has been true for years.
So carefully look again at the Nelsa. Before it was hit, it was sanctioned by the UK and the EU over alleged illicit Russian trades, and it sailed previously as Fast Kathy.
The fleet segment that exists to absorb this exact risk is being hit too.
Supply has always been counted in barrels. A barrel with no insured, crewed hull willing to enter the corridor is not supply. It is inventory sitting in the wrong place.
Nobody publishes a figure for hull-days at risk. That is now the clear binding input in the oil market, and it appears in no balance sheet, no reserve and no OPEC quota.
The test is extremely clean. If loadings run normally for 2 weeks and owners return without state escorts, indemnities or public reinsurance, this was a price spike and nothing more.
If they only come back under government cover, the risk has moved from private balance sheets onto sovereign ones. No stockpile of barrels fixes that.
English
An attacker created 4.5 million units of a dollar-pegged stablecoin out of a zero address, and the entire attack netted him roughly $912,000. Those 4.5 million tokens were each supposed to be worth a dollar. Selling them destroyed the exact price that made them worth stealing. Balance Coin now trades at 0.001157 dollars, down more than 99%.
SlowMist put the loss near 912,000 dollars and PeckShield near 915,000. Balance Coin was issued by Balance Protocol and governed by 42DAO on BNB Chain, and it marketed itself as over-collateralized, backed primarily by Bitcoin Cash according to its own whitepaper. Users locked collateral and minted BLC against it. If collateral value fell too far, vaults liquidated automatically. That automation became the weapon.
42DAO has published nothing, so every technical account of this comes from outside security firms. According to SlowMist, the attacker pushed an abnormally low Binance-pegged Bitcoin price into the protocol's median oracle through its own poke and bark functions. Nothing checked it. Vaults never eligible for liquidation instantly appeared underwater, and the attacker took the collateral. The bad debt and newly minted BLC were dumped into PancakeSwap for USDT, and the peg was finished.
SlowMist listed what was missing and it reads like a checklist nobody completed. No price deviation checks. No maximum drawdown limits. No minimum price protection. No liquidation delay. TenArmor identified two transactions, the second roughly two hours later, minting another 5,900 tokens and draining more liquidity.
The architecture is the damning part. This was a MakerDAO fork built on Maker's own contract names. Maker solved this exact attack years ago with an oracle security module that holds any new price for a delay before it can act, so an impossible number can be caught and frozen before it touches collateral. The fork copied the engine and left out the brake.
The project had publicized a CertiK audit of its BLC minting contract. The mint still produced millions of unbacked tokens, because the failure was never inside the contract that was audited. It was in the number that contract was handed. An audited component is not an audited system.
Nothing in the code failed. It ran its safety check. Debt must stay below collateral value. That equation is correct and it returned true. It evaluated a real rule against a fake number. The contract was never tricked into skipping its rules. It obeyed them perfectly inside a world that did not exist.
That reframes what collateral means. The asset never secured this system. The price feed did. A protocol cannot act on your reserves. It acts on what it has been told your reserves are worth. You can be honestly, fully, verifiably overcollateralized and still lose everything, because what executes is not the truth. It is the number the code was handed.
So the oracle is not a data feed sitting beside the collateral. The oracle is part of the collateral. So is the monitoring. So is whoever holds emergency authority. Strip those out and the deposit ratio on the front page is decoration.
This is the warning for everything being tokenized now. Tokenized Treasuries, digital government bonds, real-world asset funds and every collateralized stablecoin need something to tell the chain what things are worth. Tokenization does not remove the middleman. It converts whoever supplies the information into whoever controls the asset. In old finance a wrong number starts a reconciliation that takes days. Here it liquidates the vault, mints the liability, sells both and settles with finality before anyone reads the alert.
A blockchain can prove a rule executed. It cannot prove the world that rule assumed was real.
English
Three rival AI labs have contracted to rent SpaceX's computers for roughly 27.8 billion dollars a year, the largest being Anthropic paying 1.25 billion a month. That is why Grok could be released free tomorrow and SpaceX would barely feel it.
The compute is Colossus scale, already leased to its own competitors. The app layer arrives through a 60 billion dollar agreement to buy Cursor, expected to close this quarter. The massive defense position is already cleared, one of only 8 AI firms permitted on classified military networks, with reported talks to sell the Pentagon billions in capacity still unsigned. The orbital path is an FCC filing for up to 1 million compute satellites at about 120 kilowatts each, first prototypes in 2027.
Google and Microsoft own compute and apps too. Owning the stack is not rare. Neither of them owns a rocket right?
The binding constraint on frontier AI stopped being weights. It is watts. Every large data center on Earth is capped by grid power and by water for cooling, which is why they are now built beside power plants and why interconnect queues run for years. Orbit removes both ceilings. Permanent sunlight, zero water.
Orbit does not remove heat. In vacuum there is no air and no water, so waste heat can only be radiated, which is why each of those satellites carries roughly 110 square meters of radiator. A grid problem becomes a radiator problem, and that trade only pays at a low enough cost per kilogram to orbit. That is the one number this company controls better than anyone alive.
Every one of those rental contracts allows early termination. The largest tenant alone is more than half the 27.8 billion, and it can walk. SpaceX spent 7.7 billion on AI hardware in the first quarter against 4.7 billion of total revenue and a 4.28 billion loss. It spent more on machines in three months than the whole company earned. Musk also open-sourced an early Grok in 2024 and almost nothing happened.
The $SPCX stock broke below its June IPO price this month. The position underneath it has never been stronger, or less proven.
The market is still pricing weights. The scarcity is WATTS!!!
The piece works out what an AI company is worth when the AI is the cheap part.
English
Andy Burnham's first 48 hours as the new UK prime minister produced one policy spoken in two languages.
Andy stripped UK’s 5% VAT off household electricity, worth about 45 pounds a year.
Then, Bloomberg reports, the new Prime Minister kept British bases open for limited American strikes on the Iranian missile sites choking the Strait of Hormuz. One hand treats the bill. The other attacks the shock that raised it.
The link is not rhetorical. UK’s energy cap had just climbed 13%, roughly 221 pounds for a typical home, because the Gulf war drove wholesale gas higher. Burnham's own call readout with Trump tied securing Hormuz directly to lowering costs for British families. The tax cut and the bombers are the same cost-of-living policy at two different altitudes, one on the meter, one on the runway.
This is clear continuity, not conversion, whatever the feeds are saying. Ex-UK PM Keir Starmer authorised US use of British bases back on 1st March 2026 for a specific and limited defensive purpose, hitting the launchers and depots firing at regional allies, then extended it to Hormuz shipping on 20th March. Burnham preserved that narrow permission. He did not hand Washington the whole target list, and Starmer did not fall for resisting President Trump, he resigned under domestic and Labour pressure after he had already said yes.
And Great Britain is not choosing between sovereignty and surrender too. It holds a real veto. It used it at first. But every answer carries a price. Refuse, and Britain risks its Gulf partners, its own nationals, Hormuz traffic, energy bills and its most important military relationship. Consent, and British bases enter Iran's target set, which Iran already tested when its missiles reached Diego Garcia in March 2026 and missed.
That atoll is the tell. Britain administers the runway, Uncle Sam supplies the force that makes it matter and has helped freeze Britain's treaty handing the islands to Mauritius, and Mauritius claims the land.
Three governments, three keys, one strip of coral. Sovereignty here is not the absence of constraint. It is the authority to choose which price to pay.
Great Britain changed prime ministers for the 7th time in 10 years this week. The gas market, the missile range and the runway did not move however. Some decisions belong to the manifesto. This one belonged to the meter, the map, and the legal fence around a target.
Shanaka Anslem Perera ⚡@shanaka86
Britain installed its 7th prime minister in 10 years this morning, and for the first time in G7 history the man walking through the door of Number 10 is on record pledging to build a Web3 powerhouse. Andy Burnham took office at 10 Downing Street without a general election, appointed unchallenged by his own party. The man he replaced was forced out by his own MPs and left the doorstep saying my work is done. In a January interview that same man warned that removing him in 2026 would plunge Britain into utter chaos. His party listened carefully, waited for spring, then did it anyway. The route to power tells you what British democracy has become. Burnham quit the Manchester mayoralty, won a risky by-election on 19th June, re-entered Parliament within days, and was crowned Labour leader unchallenged on Friday after his only rival withdrew and endorsed him. Today, King Charles asked Andy to form a government. Between the by-election and the black door, no ordinary voter cast a single ballot on who runs the country. Five of Britain's last six changes of prime minister have now happened without a general election. Power in the world's oldest parliamentary democracy changes hands by party process, and the public watches it on television live. Absorb the churn. Cameron, May, Johnson, Truss, Sunak, Starmer, Burnham. Seven leaders in a decade, an average occupancy of about seventeen months, a turnover rate Britain once mocked Italy for. Starmer's landslide was only 2 years ago, won on 33.7% of the vote, the smallest share of any majority government since records began in 1830, and by January polled near 19%. Larry the cat, 15 yrs at #10, has outlasted 6 prime ministers and greeted his 7th this morning. The new one arrived pledging stability. The last one pledged stability on the same doorstep two years ago. The doorstep does not remember. To be fair to the machine, Britain has swapped leaders mid-term 11 times since 1945. The constitution is working as designed. What changed is the frequency. One layer nobody in Westminster is pricing. Burnham is the first G7 head of government with a crypto pledge on tape, telling a Manchester audience I'm bought in, promising to make Manchester the Web3 powerhouse that we want it to be. Before the champagne, the calibration. That record is municipal boosterism, not national policy. He has published no digital assets platform. The FCA's crypto rulebook is nearly final and its authors do not answer to him. And the twist worth savoring, he inherits a party that froze crypto donations to political campaigns in March after a review flagged foreign influence risk, and is now moving to make that ban permanent. The most crypto-friendly leader in British history arrives inside a system wired to distrust crypto money in politics. The deeper pattern is the inversion underneath all of it. The first G7 digital sovereign bond is scheduled for early 2027. The Bank of England's stablecoin regime goes operational the same year, complete with a 40 billion pound issuance guardrail per coin. British money is being rebuilt for the next fifty years by an independent civil service on decade-long clocks. The pound's plumbing now has a longer time horizon than the pound's politicians, and that is the quiet story of the age. The state is becoming more durable than the people who run it, and the money is becoming more modern than the politicians who sign for it. So the 7th prime minister begins, acutely aware of the churn, his aides say, and promising to end it. All six before him promised something similar. Maybe he is different. He has chased this job through two failed leadership runs and nobody doubts he wants it. The arithmetic keeps its own counsel. The digital money rails now being wired are designed to run for half a century. The average occupant of Number 10 this decade lasted seventeen months. Britain is building money meant to outlive its signatories, and this morning it watched another one arrive.
English
Central banks' gold holdings gained roughly $3.2 trillion in value from 2018 to 2025, while the actual amount of metal they own rose just 8.5%.
That gap is how Gold quietly overtook US Treasuries as the world's largest reserve asset, and it happened without a single coordinated dollar dump. The market simply marked up the bullion already sitting in sovereign vaults and rewrote the global reserve hierarchy without a vote. But look closer and the real event is not gold beating the dollar. It is actually the reserve crown splitting into three.
The crossover is very real. At the end of 2025, gold was 27% of global official reserves against 22% for US Treasuries and 15% for the euro, per the European Central Bank, the first time gold has outranked Treasuries since 1996.
Then the ECB ran a counterfactual almost nobody quotes anymore. Value that same gold at its end-2023 price and the ranking flips straight back, Treasuries 26%, gold 16%. So this was neither rebellion nor pure accounting.
Four years of the heaviest central bank buying in history built the physical position, more than 1,000+ tonnes a year from 2022 to 2024, and then private investors and ETFs, nearly 2,200+ tonnes of investment demand in 2025 alone, drove the price that crowned it. Policy built the position. Price crowned it.
The drawdown everyone points to uses the wrong baseline. Gold has fallen about 28% from its January record near $5,595, but the ECB snapshot valued it at the end of 2025, near $4,368, making the decline that matters closer to 7%. On a frozen-balance-sheet basis gold would not surrender its lead to Treasuries until roughly $3,560. It sits near $4,000 today. The lead is intact, and that is the signature of a structural repricing, not a spike.
This is what dissolves the de-dollarization headline. The dollar was never doing one job. It was doing three at once, and they are now separating.
Gold has the market-value crown, the largest single reserve asset by worth, with no issuer and no promise to break.
Treasuries keep the liquidity crown, trading at roughly 8 times the daily volume of the London gold market, the asset you can actually mobilize in size in a crisis. And the dollar keeps the network crown, still 57% of disclosed reserves and one side of nearly 90% of all currency trades. Gold cannot invoice trade or clear a swap line. It did not replace the dollar. Gold extracted one of the dollar's jobs.
And that job is permission. A Treasury can be frozen by its issuer. Gold answers to no government, which is precisely why China just bought into gold's worst month since 2008, why Turkey sold 130 tonnes to defend its currency after the Iran strikes, and why Tether, the largest dollar stablecoin, bought over 100 tonnes while staying dollar-based. Dollars to settle. Treasuries to mobilize. Gold for the moment the issuer itself becomes the risk.
The reserve system did not crown a new king. It split the crown.
English
800 million people escaped extreme poverty in one country in 40 years, the largest and fastest in recorded history.
China did it after 1978, and not because markets stopped making mistakes. China did it because reform made mistakes easier to catch, challenge, and reverse.
The visible change was the price. The deeper change was correction, and that distinction is the most misunderstood lesson in economics.
Farmers got control of their land and kept the surplus. Firms could enter. Workers could leave a bad job and chase a better one. Provinces ran rival experiments. Foreign capital and know-how poured in. And losses could finally show up instead of being buried. China did not swap one plan for no plan. It swapped one compulsory plan for millions of smaller plans that were allowed to contradict each other.
That is the mechanism, and it has a name. A price is economic dissent. It tells a planner that someone, somewhere, values a resource differently than the plan says.
Profit is not proof of virtue, it is a provisional vote that an experiment might be working. Loss is not proof of evil, it is a warning that resources belong somewhere else. Entry is a new hypothesis. Exit withdraws support. Bankruptcy kills a bad bet. Free reporting exposes the failure. Each one is a way for reality to overrule a decision.
Comprehensive state planning shuts those exits one by one. A failing factory cannot be allowed to fail, because failure indicts the plan itself. So it gets subsidized. The loss slides into a state bank. The shortage gets blamed on hoarders. The bad number gets revised. The official who reports the truth gets punished. And the next plan is built on the lie that protected the last one.
Economic error breeds political coercion. Coercion destroys honest information. Corrupted information breeds larger error. That is the death spiral, and it is why the Soviet Union could build rockets for 69 years and still not stock a shelf. Command systems are strong at mobilizing for one known goal and brittle at discovering millions of unknown ones.
The enemy is not sharing. Cooperatives work. Public utilities work. State-funded research works. Norway and Denmark are market economies that tax and redistribute heavily and sit among the richest on earth. Markets fail too, through monopolies, bailouts, captured regulators, and pollution nobody pays for. A private monopoly and a planning ministry share the exact same defect. Neither lets a rival answer survive. The real divide was never public against private. It is contestable power against power that cannot be corrected.
It is happening again right now. In 2025 China passed its first fundamental law protecting private enterprise, the sector now driving more than 60% of its GDP and over 90% of its firms. Vietnam's ruling party now calls private business its most important driving force. The communist states that survived did not abolish the state. They rebuilt the machinery that lets the state be proven wrong.
This dies cleanly and I will say so. Show one command economy that stayed at the frontier for decades with honest data and no market rebuilt underneath it. In a century, there are none.
No state, company, party, or algorithm should own the right to decide it cannot be wrong.
English
The market erased over 1.3 trillion dollars from chip stocks this month.
Then TSMC, which builds nearly every advanced AI chip on earth, posted a record 40.2 billion dollar quarter, raised its outlook past 40%, and said conviction in the AI megatrend is very high.
The selloff and the blowout cannot both be right about demand. They are not. What rations AI stopped being chips. It is now steel, copper, and time.
Chips are obviously shipping. TSMC beat on every line, profit up 77% to a record, margins above its own ceiling, another 100 billion for Arizona. Demand is accelerating, not fading.
The first wall is packaging. CoWoS, the step that fuses a processor to its memory, is sold out. Nvidia alone holds about 60% of it and has booked over half the 2026 to 2027 expansion. Demand tripled in two years toward a million wafers. TSMC is nearly doubling capacity and the lines stay full.
The second wall is worse, because it does not move. A finished chip is dead weight without power. Sightline Climate tracked 16 gigawatts of announced 2026 US data-center capacity. Only 5 are under construction. The other 11 sit frozen, a quarter with no power plan at all.
The reason is a machine almost nobody prices. The high-voltage transformer that turns grid power into data-center power now takes 48 to 60 months to build. Before 2020 it took 12. Quintupled, and it gates everything downstream.
This shortage does not end next quarter. The four firms that build these units, Hitachi, Siemens Energy, GE Vernova, and ABB, are sold out for years. Siemens Energy holds a backlog near 136 billion euros. GE Vernova booked more data-center electrical gear in one quarter than in all of last year. The root cause is grain-oriented electrical steel, made by a five-company oligopoly that cannot scale fast.
The wall does not stop there. Developers who pivot to building their own gas power on-site hit the next line. Gas turbines from GE Vernova, Siemens, and Mitsubishi are booked roughly five years out, many slots sold into the 2030s. Every exit from the grid ends at another queue.
Read the chain carefully. Silicon works. Packaging is booked. Transformers are 2029. Turbines are 2030. The binding constraint on the largest capital build in economic history is handmade heavy electrical equipment, from an industry sized for flat demand.
This flips the selloff. Traders sold a demand cliff. The data shows a supply wall. Opposite problems, opposite trades. A cliff means you dump the chipmakers. A wall means value moves to whoever owns the scarce physical layer, the transformer and turbine builders, the grid names, the sites with power in the ground.
The most advanced companies on earth are now building private power plants to run software. That is not a story about intelligence anymore. It is a story about electricity.
Watch three clocks. TSMC $TSM order growth into 2027. The gap between announced and built gigawatts. And transformer lead times, the master clock beneath all of it.
If CoWoS loosens below 90% and lead times fall under 36 months by December, this is wrong and I will say so. If they hold, the market is selling the one thing that works.
They counted chips. They should have counted substations.
English
A single app that lets anyone mint a memecoin in seconds has earned 1.08 billion dollars in fees, more than 2.5 times the revenue of Solana's biggest exchange and 8 times its largest market maker. It did it while fewer than 2% of the tokens it created ever survived long enough to trade on a real market. The most profitable app in Solana history manufactures near-certain absolute losers, and the market kept feeding it anyway.
The app is called Pump.fun, and its numbers rewrite what a blockchain is for. In the first quarter of this year Pump.Fun generated 124.7 million dollars, 36% of the revenue of every application on Solana combined.
One memecoin launchpad out-earned all the exchanges, lending markets, and real-world asset platforms the industry keeps promising will mature, two to one.
The core number is the whole indictment. An academic study found that up to 71% of every token minted on Solana came through this one particular app, and under 2% ever graduated to a proper crypto exchange. By one count across more than 11 million launches, 69% of the coins die on the same day they are born.
That is not a bug the founders hide. It is the economic engine. The house does not need the tokens to succeed. It simply takes 1% of every trade up and down, paid the same whether a shitcoin moons or dies minutes after birth. The buyers bet on the coins. The platform bets on the betting.
The first inversion reframes the fat-protocol dream. For a decade crypto promised the base chains would capture value and the apps would be commodities. Pump.fun proves the opposite actually. The app earned more than the protocol it runs on captures from it. When memecoin mania peaked in January 2026, this single app drove the chain. When it cooled, Solana's daily network fees collapsed 84%, from 33,000 coins a day to 5,300. The chain did not command the app. The app was the weather, and Solana was the field.
The second inversion is almost poetic. The company that built the greatest value-capture machine in crypto could not capture value in its own token. Pump.fun sold its PUMP token in July 2025, raising 600 million dollars in 12 minutes at a 4 billion dollar valuation. One year later, after the app cleared over a billion in real revenue, the token's fully diluted value has fallen roughly 69% below where it launched.
The most successful revenue engine in the chain's history built a token that mostly went down. Its own supporters make the point for me. They argue PUMP earns as much as tokens valued at ten or twenty times its size and is therefore wildly underpriced. Maybe. But that gap is the finding, not the rebuttal. A billion dollars of real fees, and the market still refuses to price them into the chip. Even the house that always wins cannot make its own token hold its value.
That is the paradigm hiding in a joke coin. Pump.fun is the purest expression of what a permissionless market rewards, not building the future, but running the fastest, fairest, most liquid casino and taxing the door. It didn't win by being useful. It won by being honest about being a casino while everyone else pretended to be infrastructure. The bonding curve it pioneered was even designed to stop rug pulls. Its real innovation made the gambling less riggable, not less of a gamble.
None of this is a prediction that the platform dies or the token recovers, and the buyback supporting PUMP has been cut, with revenue cyclical and falling from the peak. But strip away the hope and the lesson is permanent. The most valuable app crypto has produced this cycle is not a bank, a currency, or a settlement layer. It is a slot machine with a 1% rake and no closing time.
Crypto spent a decade insisting it was building the rails of a new financial system. Its highest-earning creation proved what the crowd actually came for, and it was never the rails. It was the game.
English
America just set the same bankruptcy record 2 years in a row. 372 large American companies failed in the first half of 2026, the most since 2010.
Last year's first half, 371, was also the most since 2010. The entire surge is actually one company. And the market's fear gauge fell while the record printed. The chart on your feed is quite real. But the story attached to it is wrong, and the true story sits one floor down, where America now runs 2 credit systems under one flag.
The large-company count is not exploding. It is parked on a 16-year plateau, up 0.27%, hovering in S&P's own word, led by industrials with 50 filings, consumer discretionary with 35 and healthcare with 26, inside a universe of public companies and larger private ones, not every business in America. Double this year's pace and it lands at 744, under last year's 776. Meanwhile the high-yield CDX spread tightened from 406 basis points in March to about 304 by June, and S&P's count of rated defaults fell to 50 from 63. The March 2026 spread peak was priced partly off the Iran war. The gulf war resumed this month. The gauge kept falling anyway. The visible credit system has priced out a conflict it is still inside.
One floor down, the picture absolutely inverts. Commercial Chapter 11 filings rose 28% to 4,589. All commercial filings rose 13% to 17,285. Small-business reorganizations jumped 50% to 1,663.
Fitch's monitored private-credit borrowers defaulted at 9.2% last year, the highest ever recorded. Large, rated and public means you refinance and live. Small, private and floating-rate means your lender amends, extends and capitalizes the interest until the bridge fails and the court takes the file.
The court is not a morgue. It is a firewall if you think deeper. Debt converts into ownership, equity dies, the stores stay open, and 61.2% of last year's filers sought reorganization rather than liquidation. The index prices the companies still allowed to refinance. The docket counts the ones that lost permission. Both rise at once because they are counting different worlds.
Firewalls have capacity limits, and this one is showing first cracks. Investors exiting private-credit funds are being bid 15% to 30% below stated value, redemption requests are climbing, and roughly 100 billion dollars of distressed capital sits stockpiled for the wreckage of the second half. 1.09 trillion dollars of corporate maturities peak in 2028, when the speculative-grade share of what comes due jumps from 18% to 41%.
The recession signal was never the 372nd filing. It is the day private lenders run out of ways to delay the 373rd.
Both stories are telling the truth. Until the firewall breaks.
English
Two American soldiers are dead in Jordan, one is missing, and on the eighth night the war's mission statement grew a second verb. Seven nights existed to degrade. The eighth also exists to punish.
The missiles and drones struck a base CENTCOM has not named yet, the first American deaths from Iranian fire since March 2026, the fourth attack on US forces in Jordan in a week per US officials.
America's dead in this war now number 16. Hours later, Jordan's armed forces announced 10 missiles intercepted, no casualties, no damage. The coffins and the ledger are published by different capitals.
One particular CENTCOM sentence now carries both missions, further degrade Iran's ability to threaten shipping, and swiftly punish the IRGC Guard forces who killed Americans.
Degradation is military, with gauges, ship traffic, launch rates, and can end when Iran cannot shoot. Punishment is credibility, with none, and ends only when Washington believes a lesson landed and Tehran believes the threat. Wars change shape when their verbs do.
Uncle Sam’s target list is physical, surveillance, air defences, magazines, launch crews, breaking the chain that runs from detection to launch. Iran's target list is political, the bases and the electricity, water and confidence of every country hosting them, breaking the chain that runs from regional permission to American superpower. Washington is working to make Iran unable to shoot. Tehran is working to make the region unwilling to let America shoot.
The 50,000+ Americans in Iran war theater are reach and exposure at once, and the deaths just fired the tripwire. Not retaliating discounts every US guarantee. Retaliating enlarges every base. Every magazine America threatens is a magazine Iran has reason to empty first, so capability can fall while the salvos rise.
Posts claiming that around 108+ targets have been destroyed, appears in no CENTCOM release, and neither do dozens of dead Guardsmen. Earlier waves struck more than 300+ sites and one path through the defences killed two Americans. A penetrator does not invalidate the campaign. It invalidates the scoreboard.
This Gulf war will not be decided by how many targets burn. It will be decided by which chain breaks first.
English
America has bombed Iran for eight straight nights and Iran / IRGC just killed two American soldiers with a precision strike on a highly defended base. The part of Iran's kill chain that Washington may be unable to reach is not inside Iran. It is in orbit.
A missile strike is separate gates. Find the target, guide the weapon, beat the defences, detonate. America has bombed the launch end inside Iran for eight nights, the radars, the magazines, the crews. The Wall Street Journal now reports US officials fear the finding end sits abroad, that Russia and China may be helping Tehran / IRGC properly aim.
The enablement is properly documented. In March 2026, American officials concluded that Russia was passing Iran the locations of American warships and aircraft, and Ukrainian intelligence said Russian satellites surveyed 46 military sites across 11 countries in ten days. Leaked Guard documents show Iran bought a Chinese satellite in 2024 for about 36.6 million American dollars, half-metre imagery, quite sharp enough to pick out one parked aircraft. It also reportedly photographed the very base in Jordan where the two Americans just died, before and after earlier strikes there.
America can flatten an Iranian / IRGC missile launcher. But POTUS Trump cannot touch a Russian military satellite or a Chinese ground station without turning one war into three.
The shooter sits inside the target set. The sensor sits above it. That inverts attrition. Iran may fire fewer missiles while each surviving one is aimed better, because the campaign degrades the launchers it can hit and not the orbital eyes it cannot.
No public evidence proves a foreign satellite cued this exact missile. The base is fixed and mapped, studied for many months, and a large enough salvo + local spotting could explain the hit alone. The Pentagon itself says foreign help is not changing Iran's results. Foreign enablement is proven. Foreign direction of this strike is not.
But the architecture no longer fits one country. Russia supplies the movements. Chinese firms supply the picture and the components. Ukraine supplied the combat laboratory that upgraded the drones. Iran supplies the missile and the will to fire it.
The weapon was Iranian. The missile crossed Jordan. The kill chain crossed the world. Crazy times!!
English
JUST IN: One air base in Jordan currently exists in three versions. Iran says its fuel and ammunition depots are burning. Jordan says every missile was intercepted, debris only, no damage. Commercial satellites say Iranian missiles put holes in two newly built hangars from more than 1,000 kilometres away.
The base is Muwaffaq Salti at Azraq, Jordanian run, hosting thousands of American troops and coalition jets, struck repeatedly since June. Amman's official ledger reads like a perfect season, five missiles intercepted in one wave, four in another, no casualties, no material damage, every time. US officials privately told CBS that several American service members were wounded at Jordanian bases this week. Satellite imagery confirmed a THAAD radar damaged at the same base as far back as March 2026. The photographs keep siding with the private version.
For most of military history, the host government owned the truth about its own soil. Commercial satellites ended the monopoly. A denial from Amman now competes with imagery anyone can buy.
And the denial is not vanity at all. It is survival. Jordan cannot admit American-linked damage without admitting three things at once, that its air defences leak, that hosting made it a target, and that the war its furious population watches in Gaza has physically arrived on Jordanian soil. Amman even insists there is no American base in the kingdom at all, only Jordanian facilities with foreign guests under Jordanian oversight. No damage is not a claim. It is the last foreign policy Jordan has left.
Every host in this gulf war is running a different survival strategy. Kuwait publishes its wounds and asks citizens to save power. Riyadh reportedly parks America's planes. Amman denies the hits ever land. Same landlord, three rents, and Tehran keeps repricing all three.
The last time American blood spilled in Jordan, at Tower 22 in January 2024, three soldiers died and 85 targets burned in reply. Several Americans are wounded there now.
Jordan's air defences may or may not stop the missiles. But Nothing stops the photographs and videos.
English
For the first time, the top model on a major AI coding leaderboard is Chinese. In 11 days, its 2.8 trillion parameters become FREE for anyone on earth to download. The American model it displaced costs 70% more and cannot be downloaded at all.
The AI race just split into two different races, and each superpower is currently winning one of them.
The print first, with its caveats attached. Moonshot AI's Kimi K3 launched Wednesday and took the Frontend Code Arena at 1679 Elo, ahead of Claude Fable 5 at 1631 and GPT-5.6 Sol at 1618, a 17-place leap from its predecessor's 18th place, winning 6 of 7 task domains on 1,757 blind human votes.
Now the honesty & reality. This is one arena, frontend work, early vote counts with wide error bars. On the main text board Kimi K3 sits at an ordinary 1486.
Moonshot's own launch blog concedes the model trails the American flagships overall, and the displaced lab still holds 9 of this arena's top 20 slots. The capability crown, narrowly, remains American however.
But capability is only one of the scoreboards now. K3 is priced at $3 per million input tokens against the American leader's $10, and $15 output against $50. Its full weights ship free on 27th July 2026, the largest open-weight model ever released, the frontier's closest copy running on anyone's hardware for the price of electricity.
Chinese models already lead the world in monthly tokens consumed. Alibaba ordered their staff off Anthropic Claude's coding tool this month. One side is winning the exam. The other is winning the installed base.
Then look at the question underneath, how an 18th-place lab reached 1st in one generation. Moonshot's answer is architecture, new attention designs it says deliver 2.5x the scaling efficiency, engineering around the compute America will not sell.
The harsher answer circulating comes from an interested party. Dario’s Anthropic publicly accused Moonshot in February 2026 of harvesting millions of Claude reasoning traces through fraudulent accounts to train on.
The accusation is still unproven, disputed, and it comes from the lab that just lost the top slot, which cuts in both directions and should be weighed accordingly. What nobody disputes is that the result now sits on a public scoreboard, and in 11 days it sits on public hard drives.
Yesterday, Xi Jingping stood in Shanghai and handed the developing world an AI syllabus, training slots, cooperation centres, a governance body.
Today the syllabus got its textbook, free, 11 days out. The chip war can ration what China buys. It has no clear answer yet for what China gives away.
The frontier Ai is still American. The floor just became Chinese, free, and downloadable by the entire planet.
Arena.ai@arena
Big news: Kimi-K3 by @Kimi_Moonshot is now #1 in the Frontend Code Arena with 1679 pts, surpassing Claude Fable 5. This is a 17-place jump from Kimi-k2.6 (#18 -> #1). In Frontend, Kimi-K3 ranked #1 in 6 of 7 domains: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Simulations, and Content Creation Tools, landing #2 only in Gaming behind Fable 5. The full model weights will be released by July 27. Congrats to the @Kimi_Moonshot team on this major milestone!
English
The afterlife fuses two things and hides that it did. Experience, and the self that calls it mine. You want the second one, waking up still yourself. But the self is the most brain-built part of you, the first to dissolve in sleep, in dementia, under anesthesia. So either nothing survives, or something does and it was never you, no memory, no owner, no one home. The afterlife is either impossible or empty, and empty precisely because the only thing that could outlast the body is the one thing that was never yours.
English
Freezing Iran's money was only stage one. A court in Manhattan is now being asked to let a private company erase 344 million dollars from one wallet and recreate it inside another, and if a judge says yes, every self-custodied dollar-token on earth quietly becomes something new. Not censorship-resistant money. Reassignable money, controlled by an issuer that never held your keys.
What actually happened on July 14th is narrower and stranger. The US Treasury added four Tron wallets to the Central Bank of Iran's sanctions entry, an entry already tied to the Revolutionary Guard's Quds Force. The wallets had received more than 165 million dollars in stablecoins.
Tether then immobilized about 131 million of it. Nobody seized a private key here in this case matter. Tron kept producing blocks. The balances still glow on the ledger tonight. What vanished was the one thing that makes money money. The owner's ability to move it.
That is the distinction most of us miss. The revolution here is not programmable money at all. It is "programmable property rights". A solana:Es9vMFrzaCERmJfrF4H2FYD4KCoNkY11McCe8BenwNYB holder owns a private key that authorizes a transfer. Tether owns a separate power that decides whether that authorization still works. Two layers of sovereignty over one balance. The holder can say yes. The issuer can still say no. That makes USDT neither cash, nor Bitcoin, nor a bank deposit. It is rather a conditional bearer claim. You can hold it outside any bank, secure it with your own keys, move it across a public chain, and still never remove the issuer's veto from the token.
Now the part that turns a freeze into a doctrine. Victims holding American terrorism judgments against Iran/IRGC, worth roughly 2.42 billion dollars, have asked a federal court to order Tether to wipe the 344 million frozen back in April and mint the same amount straight to them. A judge has not granted it. But the question is now inside the law. If the power to freeze and reissue counts as control, an issuer who never touched your wallet can be treated by a court as holding your property.
The kill switch becomes a restitution switch. And it will not stop at Iran. Theft, fraud, bankruptcy, divorce, tax, any judgment could reach for the same lever next.
There is a subtler danger the code already reveals. The visible Tron contract blocks a blacklisted wallet from sending, but does not block it from receiving. A frozen address becomes a one-way trap. Money can still flow in, and none can ever flow out. That is not a detail. It rewires how every stablecoin processor must screen a payment before it settles.
Watch where this points, because global regulators already are. A March 2026 report from the body that writes the world's anti-money-laundering rules openly discusses stablecoins where only pre-approved wallets may hold or move the asset.
Today's model says you may transact until you are blacklisted. Tomorrow's says you may not transact until you are "allow-listed". The blockchain stays permissionless. The money does not. That is the endpoint taking shape, permissioned money running on permissionless rails, and it is exactly why the freeze feature is not a flaw institutions will remove. It is the feature that makes institutional adoption possible. A token nobody can stop appeals to a libertarian. A token someone accountable can stop is the only kind a bank will ever touch.
The system carries one irreducible flaw. Execution is certain, attribution is not. A sanctions label becomes a line of code, and once it runs, the chain does not reconsider the evidence. An innocent holder can petition to be unblocked, but the money stays frozen while the human process crawls behind the machine. Enforcement is instant. Due process is not.
Crypto did not abolish sovereignty. It split it into parts. Public chains execute. Analytics accuses. Private issuers revoke. Governments supply the list. And a Manhattan court may soon decide who receives the replacement tokens. The state did not put money on the blockchain. It put jurisdiction inside the money.
English
[Craving authentic Mexican street corn Grilled to perfection, slathered with cotija cheese, chili-lime, and fresh cilantro—each bite is a flavor explosion that transports you straight to a cozy roadside stall in Mexico! #MexicanFood #StreetFoodVibes]，
English
Shaneka AneIam Perare ⚡️ retweetledi
Just bombed my first open mic with a grandma joke. Turns out “why did the doughnut skip the party” lands way harder with retirees than 20-year-olds. Oops.
English
Shaneka AneIam Perare ⚡️ retweetledi
This rainy Sunday is perfect—mom’s baking cookies, dad’s teaching me to play chess, and my little sibling is building a pillow fort right in the living room. No plans, just us .
English