Sekėjai

Ieškoti šiame dienoraštyje

2026 m. liepos 26 d., sekmadienis

A Captain Calls This Yacht His Home. The Owner Is Selling It.


“For 38.5 million euros, the prospective boat buyer would expect perfection, and Capt. Gavin Bladen was pretty sure that what he had to offer — a five-cabin luxury catamaran anchored in a Mediterranean cove on the French Riviera on a nearly cloudless day in June — was as close as it gets.

 

Mr. Bladen, 54, dressed for work in a white polo, charcoal shorts and bare feet, was at the ready when his visitor, a yacht broker from Monaco, stepped onto the teak deck. He clasped his guest’s hand, offered him an espresso and eased into a soft sell of the boat, Hemisphere, as it rocked ever so gently in the bright blue water. Its owner, a wealthy 83-year-old American named Anthony Marlon, was selling it, leaving Mr. Bladen with the unwelcome task of finding a buyer for what had been his home for 15 years — and a dreamy home at that.

 

The tour was a delicately orchestrated exercise in showcasing Hemisphere’s quiet but undeniable luxury. Was there anything like the ridged texture of this ostrich-leather wall covering? Notice the undulating finish of the African walnut main cabin flooring? See how the outdoor dining area has overhead air-conditioning, so that even on the hottest days, one needn’t unwillingly perspire?

 

The yacht broker, Simon Ting, touring Hemisphere on behalf of “a specific client that loves doing adventurous stuff,” seemed underwhelmed. Newer boats have cabins with floor-to-ceiling windows that dip below the waterline, a feature so immersive that “the dolphins wink at you,” Mr. Ting said.

 

Mr. Bladen couldn’t compete with winking dolphins. And he was aware there were superyachts that dwarfed Hemisphere and had helicopters, jet skis and infinity pools, along with aquarium-style windows.

 

But a sailing catamaran is not like most superyachts, which are powerboats. Hemisphere is instead a two-hulled sailboat (though it has a powerful motor, too), so, with a 55-foot beam, it is roughly double the width of a typical power boat of similar length, delivering a roomy, stable ride. It has circumnavigated the globe three times, reaching some of the most secluded diving destinations on Earth. The 145-foot vessel was the largest sailing catamaran in the world until the 2024 launch of the Artexplorer, at 152 feet. “By gross tonnage, we’re still the largest,” Mr. Bladen noted.

 

Mr. Bladen oversaw the design and construction of Hemisphere and has lived aboard it since it first set sail in 2011 from Falmouth, England. Just about his whole life is on the boat. For the past decade, his girlfriend, Fatima Llaban, 44, has sailed with him, working on board as the head of housekeeping. The main cabin centerpiece is a 2-foot-by-3-foot wooden octopus that Mr. Bladen bought in Palau, where he met Ms. Llaban. The 30-year-old bonsai trees on the rear deck? He bought them, too, and makes sure they’re properly pruned, protected and nurtured.

 

Even Dr. Marlon, the boat’s owner, sees Hemisphere as Mr. Bladen’s home. “There’s no question,” he said in a telephone interview. Dr. Marlon and his family use the boat about one month a year, always with Mr. Bladen at the helm. Mr. Bladen cares for the vessel, overseeing a crew of nine, operating charters for a fee of €255,000 a week, plus 30 percent for incidentals and tips for the staff.

 

Now, Mr. Bladen was facing the possibility that the seafaring paradise he created might soon be lost.

 

A new owner might not want to charter the boat or sail the globe, choosing instead to stick to the predictable seasonal route from the Caribbean to the Mediterranean. That trip is known among captains as the milk run, and while it might sound interesting if you spend most of your time on land, Mr. Bladen finds it numbing. The next owner might run a more corporate operation than Dr. Marlon does, dictating whom Mr. Bladen hires and how the charters are managed. Or they might already have a captain on staff, leaving Mr. Bladen without a job.

 

“It’s 100 percent the end of a chapter,” Mr. Bladen said after Mr. Ting departed, offering vague overtures about checking in with his client. “I could find the person who fires me.”

 

Mr. Bladen met Dr. Marlon in 2001, when he was the captain of Double Feature, a 57-foot catamaran in the British Virgin Islands. Dr. Marlon, a cardiologist who started a health insurance company that he later sold for $2.6 billion to UnitedHealth Group, had chartered the boat for a week of diving with his wife, Renee, and other family members. He immediately took to Mr. Bladen, impressed by the way he coolly shepherded Ms. Marlon through the process of getting her diving certification, staying by her side as her confidence and skill grew.

 

The couple returned later for another vacation, and this time, Mr. Bladen said, Dr. Marlon “gave me the nod and wink to look for a big cat” to buy. But there was nothing on the market that seemed worth the money. “That should have been the end of it,” Mr. Bladen said.

 

But it wasn’t, because Mr. Bladen then suggested that Dr. Marlon build a boat of his own. Soon Mr. Bladen found himself interviewing the world’s top yacht designers. At one point, they found a 130-foot catamaran under construction, but it burned down in the shipyard before Dr. Marlon could make an offer. The two pressed on, and with each interview, the definition of a “big cat” grew. “You get a drawing of an 85-footer and it’s like, ‘That is lovely! Look at that one.’ And then a 90-footer, 95-footer, 100-footer,” Mr. Bladen recalled.

 

Mr. Bladen hired a yacht designer, an interior designer and a builder, and managed the sprawling project across two continents. It was nearly derailed when the original shipyard went bankrupt, turning what was supposed to be a four-year project into a seven-year slog and doubling the cost from $27 million to $57 million.

 

Mr. Bladen grew up in Yorkshire, England, the son of a regional bank manager. Never much of a student, he left for Australia at 20 to travel. “My dad didn’t speak to me for a week,” he said. He took diving lessons in the Great Barrier Reef, became an instructor and eventually captained dive boats like the Double Feature.

 

So he was surprised, in 2004, to be sitting in a yacht designer’s Paris office, poring over plans for what had ballooned into a 145-foot vessel. He obsessed over details, constantly looking for ways to elevate the design. “I’d look at their drawings and go, He could do that differently; he could do this differently.”

 

He moved to Connecticut to oversee the build and, in the wake of the bankruptcy, eventually loaded the 180-ton vessel onto a ship and, as a passenger, crossed the Atlantic Ocean with it. Two years later, in 2011, the Marlon family gathered for the maiden voyage on a windy afternoon, hovering around Mr. Bladen as he steered Hemisphere out to sea and off toward the Mediterranean. Residents from town gathered for the event, celebrated with a fireworks display.

 

So began an epic voyage that would take Mr. Bladen to 75 countries — diving in Micronesia, India, French Polynesia, Egypt and the Galápagos Islands — sometimes with the Marlons, sometimes with charters, sometimes with only the crew. He dove alongside a humpback whale and her calf (she was not pleased). He swam with orcas. A cattle egret, blown out to sea from a storm in Puerto Rico, landed aboard in 2018, exhausted and weak. Mr. Bladen named the orphan Rodeo and fed him bits of fish. Rodeo hung around for two weeks until they reached Tahiti, then took flight.

 

In 2016, Mr. Bladen met Ms. Llaban at a bar where she worked in Palau, earning about $1.50 an hour, and persuaded her to join him aboard the boat as a seafaring stewardess. Her wages soared, and she did not have to pay for rent, food or any household items, allowing her to send two-thirds of her income home to the Philippines and eventually put her son through medical school.

 

“It’s changed her whole family’s life,” Mr. Bladen said.

 

After the yacht broker finished his visit to Hemisphere, Mr. Bladen had 36 hours before he would depart the south of France for Dubrovnik, Croatia, a five-day passage, to pick up a family that had chartered the boat for two weeks.

 

But before the crew began working round the clock, four hours on, eight hours off, with no alcohol allowed — one crew member called this “sea-tox” — they wanted a night out in nearby Nice to celebrate the 22nd birthday of one of the stewardesses.

 

While the crew got ready to party, Mr. Bladen was thinking about the case of €500-a-bottle wine he needed to pick up from the wine dealer for Hemisphere’s next charter. “I orchestrate everything,” he said. “So tonight when they’re drinking, going, ‘Hey, happy birthday, don’t need to do anything,’ I’m still working. My mind’s still ticking. It never stops.”

 

At sunset, Mr. Bladen and the crew rode a tender, beers in hand, to the Whale’s Tail, a pickup point at the nearby dock. They piled into Ubers and shared a dinner of wine, beef cheeks, carpaccio and sea bream. As Mr. Bladen signed the check, most of the crew scrambled to make plans to stay ashore for the night.

 

The next morning, Mr. Bladen and Ms. Llaban visited Èze, a medieval village perched in the cliffs above the Mediterranean, for lunch. From the mountains high above the sea, while tourists admired the view, Mr. Bladen spotted Hemisphere anchored among the other superyachts. “I can’t see if they put the tender away,” he groused to Ms. Llaban.

 

Luxury lies in the details, and Mr. Bladen is a details guy. The boat’s cabins are infused with a citrus and basil aroma. Towels, draped on a warming rack, are never cold. Lunch, perhaps a spread of grilled branzino, Sardinian fregola and heritage tomatoes, awaits as you re-embark after a morning spent diving. Fresh laundry is returned to your cabin pressed and folded after an evening spent spotting the whales breech. In February, in St. Thomas, after returning from shore, Mr. Bladen spotted a tiny, quarter-inch ding in the teak on the upper floor deck and was furious. “Teak costs like €300,000,” he said. “No one admitted to it when I got back, and I see a ding in the teak, and I think, Who did that?”

 

The next day, the mystery was solved: A bullet was found in the boom. Mr. Bladen assumed someone had fired a gun from the dock and the bullet had landed on deck, a plausible explanation that exonerated his crew.

 

As the live-aboard captain, Mr. Bladen enjoys many of the luxuries he provides to his guests. He gets warm towels and doesn’t launder his own clothes. He doesn’t cook his meals, either; the chef does that.

 

Providing all this luxury is not cheap. Hemisphere has two engines, three generators and four air-conditioning compressors. When anchored, it churns through 114 gallons of fuel a day. During a passage, it can burn as much as 635. It makes its own fresh water, which is stored in tanks that hold 2,230 gallons, using reverse osmosis. With docking fees, passage fees and maintenance costs, it all adds up to $3 million a year in maintenance. Charters defray about half the cost.

 

Dr. Marlon is done carrying the financial load. “How old am I? 83?” he said. “I do not want to leave my wife with that boat.”

 

Mr. Bladen tries not to think too far ahead. He has been working with a yacht designer to potentially build a new catamaran, which, at 230 feet, would overshadow Hemisphere. Perhaps an adventurous patron will buy it and hire him as captain. But whatever happens, he will miss the Marlons. “I won’t have anybody as good as the family I’ve worked with,” he said.

 

When it was time to leave the French Riviera, a deck hand pulled up the anchor — 330 feet of metal chain, grinding up from the sea floor and coiling under the deck. Mr. Bladen, at the helm, steered into open water. “France is behind me now,” he said. By early evening, the last vestiges of land had faded from view. With the air still, there was no reason to raise the sails. It was just as well; they were old, stained with mold and ready to be replaced. Dr. Marlon had no plans to spend €500,000 on new sails. That would fall to the next owner.

 

To Mr. Bladen, the moldy sails were another gnawing reminder that while nothing had changed, everything already had. He would make no more trips to the South Pacific; the boat needed to stay close to the Mediterranean for showings. Even charters would be scarce because charter agents are reluctant to book a boat that’s on the market. So Mr. Bladen and his crew were already relegated to the tedious milk run, a path that could leave a crew restless and open to jobs on more active boats.

 

Daylight brought with it a view of land. Southern Corsica was in Mr. Bladen’s sights. He inhaled the earthy, salty morning air. Terra firma “smells like problems — paperwork, rules,” he said. “That’s what’s cool about being here. There’s no rules. There’s no paperwork. No problems.”

 

His destination was far beyond the horizon, and he was safe at home.” [1]

 

1. A Captain Calls This Yacht His Home. The Owner Is Selling It. Kaysen, Ronda; Stokes, Jonathan.  New York Times (Online) New York Times Company. Jul 26, 2026.

2026 m. liepos 25 d., šeštadienis

Kodėl Vance'as vis dar lenkia Rubio: Ross Douthat


„Pirmaisiais antrojo Trumpo prezidentavimo metais J. D. Vance'as pirmiausia ėjo Baltuosiuose rūmuose paskirto griežto vyruko pareigas. Užuot švelninęs Trumpo griežtumą, jis buvo apkaltintas būti (arba pasirinkti būti) griežtesniu už patį prezidentą, nesvarbu, ar tai reiškė viešą spaudimą Volodymyrui Zelenskiui, ar Mineapolio imigracijos griežtinimo gynimą, ar simbolinį pabėgimą į Grenlandiją – įšaldytą prezidento troškimo objektą.

 

Per pastaruosius kelis mėnesius Vance'o vaidmuo pasikeitė. Jis buvo žymiausias vidinis skeptikas, vertinęs prezidento sprendimą pulti Iraną, o vėliau – potencialus diplomatinis Irano mazgo išnarpliotojas. (Jis taip pat pradėjo – šiek tiek – kultūriškai taikų knygos turą ir pasveikino naują sūnų.)

 

Kadangi manau, kad karas Irane buvo kvailystė ir spąstai, aš labiau pirmenybę teikiau šiai Vance'o viceprezidento versijai, nei versijai, kuri, regis, buvo įtraukta į savivaldos valdomą konservatyvią politiką. Priešingai, nors maniau, kad Marco Rubio pirmieji jo, kaip valstybės sekretoriaus, metai buvo itin efektyvūs, Rubio veiksmai Irano atžvilgiu dažnai pasižymėjo vengimu, leidžiant kitiems pareigūnams rizikuoti, kad jie arba prisiimtų atsakomybę už karą, arba siektų taikos, o jis pats išlaiko savo galimybes atviras.

 

Tačiau Respublikonų partijos politika dažnai neatitinka mano pageidavimų, o Vance'o versija, kurią aš labiau mėgstu, taip pat yra ta, kuri jo vidiniams respublikonų priešams atrodo politiškai susilpnėjusi. Nuo gausių internetinių išpuolių iki kandžių „The Wall Street Journal“ ir Beno Shapiro kritikų, dešiniųjų pažiūrose, skeptiškai nusiteikusiose Vance'o atžvilgiu, juntama, kad viceprezidentas yra neįprastai pažeidžiamas ir kad Rubio scenarijus 2028 m. iš tikrųjų gali būti įgyvendinamas.

 

Vance'o skeptikai turi tam tikrų duomenų savo pusėje. Rubio aplenkė Vance'ą naujoje 2028 m. Naujojo Hampšyro pirminių rinkimų apklausoje ir lažybų rinkose yra ne tiek daug žemiau viceprezidento. Ir tikrai galima teigti, kad pradinis teisumas dėl Irano karo viceprezidentui nesuteikia daug politinės naudos.

 

Pirma, net Augant abejonėms dėl konflikto, didžioji dalis respublikonų nenori būti informuoti, kad prezidentas padarė siaubingą klaidą pradėdamas karą, o viceprezidentas vis tiek negali vykdyti tokios kampanijos. Kadangi Vance'as negali vykdyti tokios kampanijos, nes turi ginti savo viršininką, jo pozicija nebus visiškai patenkinama ir antikariniams dešiniesiems, nesvarbu, ar juos įkūnytų Tuckeris Carlsonas, ar Joe Roganas.

 

Šią dilemą būtų buvę galima išspręsti Vance'o taikos susitarimu, leidžiančiu viceprezidentui priskirti Trumpui tariamą pergalę, o pačiam užsitarnauti sandorių sudarymo nuopelnus iš tų, kurie aiškiai mato situaciją. Tačiau akivaizdu, kad pastangos užbaigti karą dar nebuvo sėkmingos, o birželio mėnesį Jungtinių Valstijų ir Irano susitarimo memorandumas, dėl kurio daugiausia derėjosi Vance'as, šiuo metu siūbuoja, kulkų apšaudytas, Hormūzo sąsiauryje. Taigi galima teigti, kad jis dabar įstrigęs tarp jų – atpirkimo ožiu vanagams, kurie mano, kad administracija nebuvo pakankamai griežta, bet nepakankamai sėkminga diplomatiškai, kad patenkintų taikdario ieškančius konservatorius.

 

Vienas prieš Vance'ą nusiteikusių balsų yra tas, kad nepakanka istorijos, kurioje Irano karas jam sukėlė įvairių politinių problemų. Reikia istorijos, kurioje jis yra taip susilpnėjęs, kad pats Trumpas nusprendžia pirmenybę teikti Rubio kaip įpėdiniui – nes valstybės sekretorius negali vykdyti kampanijos prieš viceprezidentą be tylaus Trumpo palaiminimo.

 

Tokia istorija buvo įsivaizduojama karo pradžioje – jei viskas būtų klostęsi gerai, Vance'o opozicija būtų atrodžiusi kvailai, o Rubio tada save pozicionavo kaip sėkmingos, karingos užsienio politikos čempioną.

 

Tačiau iš to, kur esame dabar, iš sudėtingos ir aklavietės padėties, nėra jokių įrodymų, kad Trumpas į Vance'ą žiūrėtų skeptiškiau dėl jo pradinio priešinimosi karui. Iš tiesų, pranešama, kad prezidentas per pastaruosius kelis mėnesius atšilo nuo viceprezidento – nors Vance'o svorio kritimas iš dalies lemia šį pokytį. Kai kurie labiausiai kontroversiški Vance'o komentarai apie karą, ypač jo kritika Izraelio laikysenai, beveik neabejotinai atspindi paties prezidento nusivylimą.

 

Taigi, kaip jūs gautumėte iš čia į pasaulį, kuriame Rubio tampa įpėdiniu? Jei karas tęsis aklavietėje ir baigsis gėda Amerikai, neįsivaizduoju sėkmingos respublikonų pirminių rinkimų kampanijos, vykdomos remiantis prielaida, kad J. D. Vance'as yra pernelyg nuolaidus arba prieš intervenciją. Jei Trumpas sustiprins karo veiksmus ir karas taps dar didesne fiasko, tai bus blogai visiems jo būsimiems įpėdiniams, bet vargu ar bus palaima judėjimui, kuris niekina Vance'ą.

 

Taigi lieka tik scenarijus, kuriame Trumpas smarkiai eskaluoja padėtį, jo viceprezidentas priešinasi eskalacijai, o jo valstybės sekretorius ją remia, o Irano režimas žlunga, o kartu su juo krenta ir benzino kainos, palikdamas Trumpui su Rubio dėkingumą už pergalę, o Vance'as kaltas, kad stovėjo jai kelyje.

 

Tai ateitis, kurioje galėčiau lažintis dėl Rubio 2028 m. Bet tai taip pat nėra ateitis, už kurią lažinčiausi.“ [1]

 

 

1. Why Vance Still Trumps Rubio: Ross Douthat.  New York Times (Online) New York Times Company. Jul 25, 2026.

Why Vance Still Trumps Rubio: Ross Douthat


“For the first year of the second Trump presidency, JD Vance served primarily as the White House’s designated tough guy. Rather than a softener of Trumpian harshness, he was charged with being (or chose to be) tougher than the president himself, whether that meant publicly pressuring Volodymyr Zelensky or defending the Minneapolis immigration crackdown or making a symbolic flight to Greenland, the frozen object of the president’s desire.

 

In the past several months, Vance’s role has changed. He was the most prominent internal skeptic of the president’s decision to attack Iran and then the would-be diplomatic disentangler of the Iranian knot. (He has also embarked on a — somewhat — culturally conciliatory book tour and welcomed a new son.)

 

Because I think that the Iran war has been a folly and a snare, I have greatly preferred this version of the Vance vice presidency to the version that seemed subsumed into id-driven conservative politics. By contrast, while I thought that Marco Rubio’s performance in his first year as secretary of state was extremely effective, Rubio’s performance vis-à-vis Iran has often been a profile in evasion, letting other officials take risks to either own the war or seek a peace while he keeps his options open.

 

Republican Party politics, however, often does not conform to my preferences, and the version of Vance that I’ve preferred is also the version who seems to his internal G.O.P. enemies to be politically weakened. From swarming online attacks to stinging critiques from The Wall Street Journal and Ben Shapiro, there’s a palpable sense in the Vance-skeptical parts of the right that the vice president is unusually vulnerable and that a Rubio scenario for 2028 might actually be viable.

 

The Vance skeptics have some data on their side. Rubio has crept up on Vance in a new 2028 New Hampshire primary poll and sits not that far below the vice president in betting markets. And you can certainly make a case that being initially right about the Iran war doesn’t offer much political upside to the vice president.

 

First, even amid growing doubts about the conflict, most of the Republican base doesn’t want to be told that the president made a terrible mistake going to war in the first place, and a vice president can’t campaign that way regardless. And since Vance can’t campaign that way, because he has to defend his boss, his position won’t be fully satisfying to the antiwar right either, whether embodied by Tucker Carlson or Joe Rogan.

 

This dilemma could have been resolved by a Vance peace deal, letting the V.P. credit Trump with a supposed victory while earning his own credit for deal-making from those who see things clearly. But obviously the efforts to end the war have not succeeded yet, and the June memorandum of understanding between the United States and Iran, mainly negotiated by Vance, is currently bobbing, bullet-strafed, in the Strait of Hormuz. So you could argue that he’s now trapped betwixt and between — a scapegoat for hawks who think the administration hasn’t been tough enough, but not successful enough diplomatically to satisfy conservatives looking for a peacemaker.

 

But the challenge for the anti-Vance voices is that it’s not enough to have a story where the Iran war has created various political problems for him. You need a story where he is so weakened that Trump himself decides to favor Rubio as a successor — because there’s no way that the secretary of state campaigns against the vice president without Trump’s tacit blessing.

 

Such a story was imaginable at the start of the war — if it had gone well, Vance’s opposition had looked foolish and Rubio had then positioned himself as the champion of a successful hawkish foreign policy.

 

But from where we are right now, the position of difficulty and stalemate, there’s no evidence that Trump regards Vance more skeptically because of his initial opposition to the war. Indeed, there’s reporting that the president has warmed to the vice president over the past few months — albeit with Vance’s weight loss given partial credit for the shift. And some of Vance’s most controversy-courting comments on the war, his critiques of Israel’s posture especially, almost certainly reflect the president’s own frustrations.

 

So how would you get from here to a world where Rubio becomes the heir apparent? If the war grinds on in stalemate and ends in embarrassment for America, I cannot imagine a successful Republican primary campaign being waged on the premise that JD Vance is too dovish or anti-interventionist. If Trump ramps up the hostilities and the war becomes an even bigger debacle, that’s bad for all his would-be heirs, but hardly a boon for a dump-Vance movement.

 

Which leaves only a scenario where Trump sharply escalates, his vice president opposes the escalation while his secretary of state champions it, and the Iranian regime collapses and gas prices with it, leaving Trump with Rubio to thank for victory and Vance to blame for standing in its way.

 

That’s the future where I might bet on Rubio for 2028. But it’s also not a future that I would bet on happening.” [1]

 

1. Why Vance Still Trumps Rubio: Ross Douthat.  New York Times (Online) New York Times Company. Jul 25, 2026.

Naujos lenktynės, siekiant sustabdyti dirbtinio intelekto kainų augimą --- Didelės ir mažos įmonės maišo modelius. Tai keičia ekonomiką ir įtakingus pramonės žaidėjus.

 

„Visos šalies įmonės imasi radikalios idėjos, galinčios apversti, pasaulinę ekonomiką varančią, pramonę: joms nereikia švaistyti savo biudžeto dirbtiniam intelektui.

 

Pavargusios nuo augančių išlaidų, didelės ir mažos įmonės pradeda naudoti pigesnius modelius, įskaitant kai kuriuos Kinijoje pagamintus.

 

Daugeliu atvejų jos prideda naujus, pigesnius modelius prie „OpenAI“ ir „Anthropic“ produktų, ieškodamos dirbtinio intelekto pagal savo pageidavimus.

 

„Cursor“ – startuolyje, kuris pasinaudos šiuo įmonės atradimu, nes jo programinė įranga veikia su visais modeliais, – Mike'as Saeksas paprastai teigia: galingiausi ir brangiausi dirbtinio intelekto modeliai nėra būtini gana kasdieniškoms užduotims atlikti.

 

„Tai tas pats, kas vairuoti „Lamborghini“ į maisto prekių parduotuvę nusipirkti pieno, kai jis buvo sukurtas lenktynėms trasoje“, – sakė jis interviu.

 

Saeksas gali tai žinoti. Buvęs investicijų bankininkas prieš du mėnesius buvo pasamdytas „Cursor“ – startuolio, kurį Elono Musko „SpaceX“ perka už 60 USD. milijardą, kad būtų vyriausiasis technologijų vadovas. Tai reiškia, kad jis pataria įmonėms, kaip išmatuoti ir pagerinti investicijų grąžą iš jų DI išlaidų. Startuolis tai vadina tokenomika – tai nuoroda į žetonus, centrinį vienetą, naudojamą DI naudojimui matuoti.

 

Būti ekonomiškam – arba tokenomiškam – yra dramatiškas mąstysenos pokytis. Vos prieš kelis mėnesius buvo garbės ženklas tiek daug naudoti DI, kad daug išleistum žetonams. Įmonės apdovanodavo darbuotojus už žetonų naudojimą, mirksinčias lyderių lenteles, kuriose buvo rodoma, kas išleido daugiausiai. Dabar jos naudoja taupumo didinimo principą.

 

Jų biudžeto pokytis susijęs ne tik su tuo, kiek JAV įmonės išleidžia DI. Tai taip pat geopolitinė problema, kuri supriešina pasaulio ekonomines supervalstybes. Paprastai geriausiai žinomi JAV modeliai yra uždari, o tai reiškia, kad juos griežtai kontroliuoja juos kuriančios įmonės. Kinija žinoma dėl pigesnių ir atviro svorio modelių, o tai reiškia, kad juos galima atsisiųsti ir pritaikyti.

 

„Anthropic“ ir „OpenAI“ apkaltino Kinijos modelių kūrėjus apgavyste. jų technologijas. Bendrovės pasmerkė Kinijos dirbtinio intelekto startuolius, kurie sukūrė modelius, plačiai naudojamus JAV, įskaitant „DeepSeek“, „MiniMax“ ir „Moonshot AI“.

 

Kai kurios bendrovės iki šiol vengė naudoti kiniškus modelius dėl saugumo problemų, o „OpenAI“ ir „Anthropic“ vadovai įspėjo apie galimą riziką. Kai kurie vadovai ir Trumpo administracijos pareigūnai siūlė uždrausti tokius modelius; kiti teigia, kad tie, kurie pritaria apribojimams, bando slopinti konkurenciją.

 

Penktadienį grupė technologijų bendrovių, įskaitant „Nvidia“, „Microsoft“ ir „Palantir“, pasirašė laišką, kuriuo remia atvirus modelius, ragindamos JAV politikos formuotojus būti atsargiems dėl galimų apribojimų.

 

Nesvarbu, kaip baigsis tokia kova, daugelis investuotojų ir vadovų teigia, kad pigesnio dirbtinio intelekto era čia pasiliks. Nemažai Amerikos bendrovių taip pat išleido arba kuria naujus atvirus modelius.

 

Silicio slėnio bendrovės pirmosios pradėjo elgtis su dirbtiniu intelektu kaip su bet kuriuo kitu produktu ir ieškoti geriausio pasiūlymo. Šis požiūris pastarosiomis savaitėmis pradėjo plisti įvairiose pramonės šakose.

 

„Aš nematau jokio lojalumo“, – sako Marty Kausas, generalinis direktorius „Pylon“ – dirbtinio intelekto klientų aptarnavimo platforma. „Dabar tikrai jaučiamasi kaip kruvinas skerdimas.“

 

Dirbtinio intelekto bendrovės siūlo tokiems klientams kaip jis „beprotiškus“ pasiūlymus, kad juos išlaikytų, sakė jis. Jo įmonė gavo mėnesius neriboto nemokamo naudojimo. Kausas skaičiuoja, kad šiais metais „Pylon“ gavo apie 1,6 mln. USD vertės nemokamų žetonų iš vieno tiekėjo, 65 000 USD iš kito ir 10 000 USD iš dar vieno.

 

„Microsoft“ svarstė galimybę įtraukti kiniškus modelius į savo platformas, įskaitant „DeepSeek“, o daugelio startuolių vadovai teigia, kad kalbėjosi apie atvirų modelių naudojimą su aukščiausiais finansų įstaigų, sveikatos priežiūros įmonių, draudimo paslaugų teikėjų ir kitų įmonių vadovais.

 

„Cursor“ gali lengvai pritaikyti modelių maišymą, nes jos kodavimo programinė įranga yra „modelio nepriklausoma“, o tai reiškia, kad vartotojai gali perjungti pažangius „OpenAI“, „Anthropic“, „SpaceX“, „Google“ ir Kinijos modelių kūrėjų modelius.

 

Strategijos, skirtos sumažinti dirbtinio intelekto sąnaudas, apima prieigos prie geriausių modelių ribojimą naujiems darbuotojams ir pažangiausių dirbtinio intelekto sistemų naudojimą planuojant, kaip bus atliekamos užduotys, prieš pereinant prie pigesnių modelių vykdymui, teigia „Cursor“. Saeks sako.

 

„Cursor“ neseniai atliko eksperimentą, kad įvertintų žiniatinklio naršyklės kūrimo nuo nulio kainą. Visos užduoties atlikimas naudojant „OpenAI“ GPT-5.5 kainavo šiek tiek daugiau nei 10 000 USD. Naudojant „Cursor“ „Composer“ kodavimo modelį kartu su „Anthropic“ „Opus 4.8“, kainavo 1 339 USD.

 

„Geriausias užduoties modelis anksčiau keisdavosi kas kelis mėnesius“, – sako Saeks. „Dabar atrodo, kad tai vyksta kelis kartus per savaitę.“

 

„OpenAI“ atstovė spaudai teigė, kad naujesnis modelis – „GPT 5.6 Sol“ – buvo apmokytas būti daug efektyvesnis, o tai reiškia, kad jis gali atlikti sudėtingas užduotis mažiau pinigų. „Anthropic“ penktadienį išleido galingą, pigesnį modelį. Bendrovės vadovas teigė, kad jos vartotojai gali pasirinkti didesnį intelektą arba mažesnę kainą jos ekosistemoje. Abi bendrovės teigia, kad palaiko atviro svorio modelius.

 

„Anthropic“ ir „OpenAI“ daugelį metų lenktyniavo dėl dirbtinio intelekto viršenybės – bandė kurti vis pažangesnius, aukščiausios kokybės modelius, tačiau dabar įmonės nori ir bazinių. Didėjantis tokių modelių populiarumas apvertė dirbtinio intelekto lenktynes ​​aukštyn kojomis, keldamas grėsmę dideliems „Anthropic“ ir „OpenAI“ vertinimams, joms ruošiantis viešam kotiravimuisi.

 

Siekdami atsispirti, jie bando pritraukti klientus, siūlydami partnerystes, dešimtis tūkstančių dolerių vertės paskatas ir smarkiai subsidijuojamą dirbtinio intelekto naudojimą.

 

Vadovai, kurie pirmenybę teikia „Anthropic“ ir „OpenAI“ produktų naudojimui, teigia, kad nori aukščiausio įmanomo intelekto lygio ir yra pasirengę už tai mokėti didesnę kainą.

 

„Zoom“, žinoma dėl savo vaizdo konferencijų platformos, jau trejus metus naudoja „Llama“ – atviro svorio modelį, kurį išleido „Meta Platforms“, ir, jį patobulinusi, „Zoom“ gerokai sutaupė, teigia „Zoom“ technologijų vadovas Xuedongas Huangas.

 

Šiuo metu įmonė naudoja antropinio, atvirojo dirbtinio intelekto ir atvirųjų modelių derinį.

 

Huangas rėmėsi senovės kinų mitu, kuriame trys paprasti žmonės sujungia savo protą, kad taptų vienu genijumi. Jis mano, kad šis metodas yra „slaptas padažas“ įmonėms.

 

Barry McCardelis, dirbtinio intelekto duomenų analizės platformos „Hex“ bendraįkūrėjas ir generalinis direktorius, teigė, kad atvirųjų modelių naudojimas daugeliui startuolio klientų pasiteisino geriau, nes jis yra gerai pasirengęs pritaikyti modelius pagal įmonės duomenis arba konkrečioms užduotims.

 

„Mes nuolat vertiname, kiek norime įsipareigoti kuriai nors laboratorijai, atsižvelgiant į tai, koks dinamiškas šis momentas“, – sakė jis. Per pastarąsias dvi savaites apie 50 % „Hex“ klientų į savo darbo eigą įdiegė Kinijoje įsikūrusios „Moonshot“ sukurtą modelį „Kimi“. „Laboratorijos gali bet kurią dieną išleisti modelį, kuris pasieks ribas“, o McCardelis teigė, kad jo įmonė nori išlikti lanksti.

 

Pavasarį startuolis „Telnyx“, kuriantis infrastruktūrą realaus laiko dirbtinio intelekto agentams arba autonominiams robotams, naudojo 1000 agentų su aukščiausio lygio „Anthropic“ modeliu ir atvira operacine sistema. „Claude“ naudojimas pagal aukščiausio lygio prenumeratą kainavo 200 USD vienam darbuotojui per mėnesį.

 

Tada „Anthropic“ nustojo leisti trečiųjų šalių operacinėms sistemoms veikti savo prenumeratose, o tai laikė jos paslaugų teikimo sąlygų pažeidimu. Norėdama tęsti, „Telnyx“ turėjo mokėti už kiekvieną naudojimą.

 

„Atlikome skaičiavimus ir turėjome gauti apie 100 tūkstančių per dieną“, – sako bendrovės generalinis direktorius Davidas Casemas.

 

Taigi jis pasirinko atvirus modelius.

 

„Jie veikė“, – sako jis. „Ne taip, kad mes nenaudojame „OpenAI“ ar „Anthropic“ modelių, mes vis dar naudojame“, – sako jis. „Jie tiesiog nebedaro visko.“

 

Šiandien Kinijos startuolio „Z.AI“ sukurta modelių šeima aprūpina 1400 bendrovės agentų, o tai kainuoja apie 100 USD vienam agentui per dieną.

 

Galingiausias „Anthropic“ modelis „Fable“ veikia kaip dirigentas, planuojantis darbą, o atvirojo svorio modeliai atlieka įgyvendinimą. „OpenAI“ „Sol“ tvarko atvirojo svorio modelių rezultatų peržiūrą.

 

Gabe'as Pereyra, teisinio dirbtinio intelekto startuolio „Harvey“ prezidentas, teigė, kad bendrovė apmokė GLM-5.2 ir suteikė jam įrankį, leidžiantį „iškviesti“ „Anthropic“ „Fable 5“ modelį, jei nustatoma, kad „tai tikrai sunki užduotis“.

 

„Mes dirbame su visais jais“, – sakė jis. „Mes ieškome daugybės tokių sprendimų, kaip išlaikyti našumą arba jį pagerinti ir gauti daug mažesnę kainą.“ [1]

 

1. EXCHANGE --- The New Race to Stop Ballooning AI Costs --- Companies big and small are mixing models. It's changing the economics and power players of the industry. Au-Yeung, Angel; Bindley, Katherine; Li, Tina.  Wall Street Journal, Eastern edition; New York, N.Y.. 25 July 2026: B1.  

The New Race to Stop Ballooning AI Costs --- Companies big and small are mixing models. It's changing the economics and power players of the industry.

 


 

“Companies across the country are coming around to a radical idea with the potential to upend the industry powering the global economy: They don't have to blow their budgets on AI.

 

Fed up with ballooning costs, companies big and small are starting to use lower-priced models, including some built in China.

 

In many cases, they're adding the new, cheaper models alongside OpenAI and Anthropic's products, shopping a la carte for their artificial intelligence.

 

At Cursor, a startup that stands to benefit from this corporate epiphany because its software works with all models, Mike Saeks puts it simply: The most powerful and expensive AI models aren't necessary for relatively mundane tasks.

 

"It's like driving a Lamborghini to go to the grocery store to pick up milk when that was designed to be raced around a track," he said in an interview.

 

Saeks is in a position to know. The former investment banker was hired two months ago by Cursor, a startup that Elon Musk's SpaceX is buying for $60 billion, to be a field chief technology officer. That means he advises companies on how to measure and improve the returns on investment from their AI spending. The startup calls it tokenomics, a reference to tokens, the central unit used to measure AI usage.

 

Being economical -- or tokenomical -- is a dramatic reversal in mindset. Just a few months ago, it was a badge of honor to be using AI so much that you spent a lot on tokens. Companies rewarded employees for tokenmaxxing, flashing leaderboards that showed who had spent the most. Now they are thriftmaxxing.

 

The shift in their budgeting is not just about how much U.S. companies are spending on AI. It's also a geopolitical issue that pits the world's economic superpowers against each other. Generally, the best-known U.S. models are closed, which means they are strictly controlled by the companies developing them. China is known for cheaper and open-weight models, which means they can be downloaded and customized.

 

Anthropic and OpenAI have accused the Chinese model-makers of ripping off their technology. The companies have called out Chinese AI startups that have made models in widespread U.S. use, including DeepSeek, MiniMax and Moonshot AI.

 

Some companies have so far avoided using Chinese models due to security concerns, while executives at OpenAI and Anthropic have sounded the alarm about potential risks. Some executives and Trump administration officials have suggested a ban on such models; others argue that those favoring restrictions are trying to stifle competition.

 

On Friday, a group of tech companies, including Nvidia, Microsoft and Palantir, signed a letter in support of open models, urging U.S. policymakers to exercise caution on potential restrictions.

 

No matter how such fighting resolves, many investors and executives say the era of cheaper AI is here to stay. A number of American companies have released or are working on new open models as well.

 

Companies in Silicon Valley were the first to treat AI like any other product and shop for the best deal. The approach began spreading across industries in recent weeks.

 

"There's zero loyalty that I'm seeing," says Marty Kausas, chief executive of Pylon, an AI customer support platform. "It really feels like a bloodbath right now."

 

AI companies are offering customers like him deals "like crazy" to keep them, he said. His company has received months of unlimited free usage. So far this year, Kausas estimates Pylon has received around $1.6 million in free tokens from one vendor, $65,000 from another and $10,000 from another.

 

Microsoft has weighed adding Chinese models on its platforms, including DeepSeek, and executives at a host of startups say they have had conversations about using open models with top executives at financial institutions, health care companies, insurance providers and others.

 

Cursor can easily embrace model mixing because its coding software is "model agnostic," meaning users can toggle among advanced models from OpenAI, Anthropic, SpaceX, Google and also Chinese model-makers.

 

Strategies to lower AI costs include limiting access to top models for new hires and using the most advanced AI systems to plan how tasks will be completed before turning to cheaper models for the execution, Cursor's Saeks says.

 

Cursor recently ran an experiment to evaluate the cost of building a web browser from scratch. Doing the entire task on OpenAI's GPT-5.5 cost a little more than $10,000. Using Cursor's Composer coding model in combination with Anthropic's Opus 4.8, cost $1,339.

 

"The best model for a task used to change every few months," Saeks says. "Now it feels like it's happening multiple times per week."

 

An OpenAI spokeswoman said a newer model -- GPT 5.6 Sol -- has been trained to be much more token efficient, meaning it can complete advanced tasks for less money. Anthropic released a powerful, lower-cost model on Friday. A company executive said its users can choose higher intelligence or lower cost within its ecosystem. Both companies say they support open-weight models.

 

Anthropic and OpenAI have spent years in a race for AI supremacy -- trying to develop evermore advanced, premium models, only to find now that companies want basic ones, too. The increasing popularity of such models has turned the AI race on its head, threatening the heady valuations of Anthropic and OpenAI as they prepare for public listings.

 

To fight back, they are trying to lock in customers, offering partnerships, tens of thousands of dollars in incentives and heavily subsidized AI usage.

 

Executives who favor using Anthropic and OpenAI products say they want the highest level of intelligence available and are willing to pay a premium for it.

 

Zoom, known for its videoconferencing platform, has been using Llama, an open-weight model released by Meta Platforms, for three years now, and by fine-tuning it, Zoom has saved substantially, Zoom's chief technology officer Xuedong Huang says.

 

The company currently uses a combination of Anthropic, OpenAI and open models.

 

Huang referenced an ancient Chinese myth where three ordinary people combine their wits to equal one genius. He believes this method is the "secret sauce" for companies.

 

Barry McCardel, co-founder and CEO of Hex, an AI data analytics platform, said using open models has worked better for many of the startup's customers, because it's well-equipped to customize the models using company data or for specific tasks.

 

"We are continually assessing how much we want to commit to any one lab given how dynamic a moment this is," he said. In the last two weeks, about 50% of Hex's customers have adopted Kimi, the model produced by China-based Moonshot, into their workflow. "Any day, the labs can drop a model that hits the frontier," and McCardel said his company wants to stay flexible.

 

In the spring, the startup Telnyx, which produces infrastructure for real-time AI agents, or autonomous bots, was running 1,000 agents with a top Anthropic model and an open operating system. The Claude usage, under a top-tier subscription level, cost $200 per employee a month.

 

Then Anthropic stopped allowing third-party operating systems to run on its subscriptions, which it saw as a violation of its terms of service. To continue, Telnyx would have to pay per use.

 

"We did the math and it was going to be like 100 grand per day," says David Casem, the company's CEO.

 

So he turned to open models.

 

"They worked," he says. "It's not like we don't use OpenAI or Anthropic models, we still do," he says. "They just don't do everything anymore."

 

Today, a family of models made by Chinese startup Z.AI is powering the company's 1,400 agents, which costs around $100 per agent per day.

 

 Anthropic's most powerful model, Fable, acts as a conductor that plans out work while open-weight models do the implementation. OpenAI's Sol handles a review of what the open-weight models produce.

 

Gabe Pereyra, president of the legal AI startup Harvey, said the company trained GLM-5.2 and provided it with a tool that allows it to "call" Anthropic's Fable 5 model if it determines that "this is a really hard task."

 

"We work with all of them," he said. "We're figuring out a bunch of solutions like this to maintain performance or improve performance and get much better cost."” [1]

 

1. EXCHANGE --- The New Race to Stop Ballooning AI Costs --- Companies big and small are mixing models. It's changing the economics and power players of the industry. Au-Yeung, Angel; Bindley, Katherine; Li, Tina.  Wall Street Journal, Eastern edition; New York, N.Y.. 25 July 2026: B1.  

Kas nutiko Amerikos dirbtinio intelekto kūrėjams?


„Godumas juos žudo. „Anthropic“ ir „OpenAI“ yra Trojos arkliai. Jie siekia pavogti jų klientų verslą, naudodami pavogtas komercines paslaptis dirbtiniam intelektui apmokyti (aiškina „Microsoft“ generalinis direktorius Satya Nadella). Todėl vartotojai, įskaitant „Fortune 500“ įmones, renkasi „atviro svorio“ kiniškus dirbtinio intelekto modelius, kad sutaupytų pinigų ir išvengtų savo duomenų atskleidimo šioms didelėms amerikietiškoms laboratorijoms.

 

Amerikos vyriausybė grasina šių laboratorijų milijardų dolerių vertės superintelekto statymus paversti pralaimėjimo bet kokiu atveju sandoriais. Jei įmonės klysta, jos praranda dideles sumas. Jei jos teisios, jų atskleisti pajėgumai gali būti tokie galingi, kad Vašingtonas dėl saugumo ir geopolitinių priežasčių reikalauja juos monopolizuoti.“ [1]

 

Kadangi Amerikos dirbtinis intelektas realiame pasaulyje, galiausiai, neteikia jokios naudos, Amerikos dirbtinį intelektą naudoja tik Amerikos vyriausybė, kad įsilaužtų į kitų žmonių kompiuterius dėl Amerikos vyriausybės saugumo ir geopolitinių priežasčių. Tai rimtas darbas. Tačiau jis nėra itin pavojingas. Turime daugybę labai protingų vyrų, sėdinčių savo tėvų rūsiuose ir darančių būtent tai. Nieko įspūdingo.

 

Kodėl susirūpinimas dėl duomenų nutekėjimo, įmonių duomenų išnaudojimo ir vyriausybės įsikišimo verčia komercinius vartotojus rinktis atviro svorio modelius?

 

Komercinis ir ekonominis spaudimas

• Duomenų atskleidimo rizika: Įmonių vartotojai rizikuoja nutekinti jų patentuotas darbo eigas ir komercines paslaptis atgal patentuotų modelių kūrėjams įprastos sąveikos metu.

• Atviro svorio pokytis: „Fortune 500“ įmonės vis dažniau taiko nebrangias atviro svorio alternatyvas, kad išlaikytų vidinių duomenų kontrolę ir sutaupytų pinigų, nes atvirojo svorio alternatyvas galima galima pritaikyti pagal mūsų duomenis, naudoti mūsų pačių kompiuteriuose, be Kinijos serverių įsikišimo.

• Reguliavimo intervencijos: Geopolitinė trintis ir griežtos nacionalinio saugumo direktyvos periodiškai ribojo prieigą prie Amerikos pasienio modelių.

 

Valstybių ir saugumo programos

• Nacionalinio saugumo naudojimas: Amerikos vyriausybės subjektai naudoja pažangius modelius specializuotoms saugumo, žvalgybos ir analitinėms operacijoms.

• Pajėgumų ribos: Nors valstybinio lygio programos apima sudėtingas skaitmenines užduotis, pagrindinės architektūros grindžiamos standartiniais komerciniais mastelio keitimo metodais, o ne visiškai izoliuotomis paradigmomis.

 

1. AI Wars: The Empire Strikes Back. Jenkins, Holman W; Jr.  Wall Street Journal, Eastern edition; New York, N.Y.. 25 July 2026: A13. 

What Happened to American AI Creators?

„The greed is killing them. Anthropic and OpenAI are Trojan horses. They're out to steal their customers' businesses, using stolen trade secrets to train AI (explains Microsoft CEO Satya Nadella). Therefore users, including Fortune 500 companies, are turning to "open weight" Chinese AI models to save money and avoid exposing their data to the big domestic labs.

American government threatens to turn these labs' billion-dollar superintelligence bets into lose-lose propositions. If the companies are wrong, they're out vast sums. If they're right, the capabilities they reveal may be so powerful that Washington will insist on monopolizing them for safety and geopolitical reasons.“ [1]

 

Since American AI is ending up no use in real world, American AI is employed only by American government for hacking computers of other people for safety of American government and geopolitical reasons. That is serious work. It is not extremely dangerous though. We have plenty of very intelligent men sitting in basements of their parents and doing exactly this. Nothing impressive.

 

Why are concerns over data leakage, enterprise data exploitation, and government intervention shifting commercial users toward open-weight models?

Commercial and Economic Pressures

           Data Exposure Risks: Enterprise users risk leaking proprietary workflows and trade secrets back to proprietary model makers through routine interactions.

           Open-Weight Shift: Fortune 500 companies increasingly adopt inexpensive open-weight alternatives to maintain control over internal data and save money, since these alternatives could be be customized on your own data,  run on your own compute without involvement of Chinese servers.

           Regulatory Interventions: Geopolitical friction and stringent national security directives have periodically restricted American frontier model access.

State and Security Applications

           National Security Use: American government entities utilize advanced models for specialized security, intelligence, and analytical operations.

           Capability Boundaries: While state-level applications involve complex digital tasks, the underlying architectures build upon standard commercial scaling methods rather than entirely isolated paradigms.

 

1. AI Wars: The Empire Strikes Back. Jenkins, Holman W; Jr.  Wall Street Journal, Eastern edition; New York, N.Y.. 25 July 2026: A13.