Slovenija potrebuje davčni sistem, ki ne bo kaznoval znanja in pripravljenosti delati več. Če želi povečati produktivnost in dolgoročno gospodarsko rast, mora razbremenitev dela postati del širše razvojne strategije. Reforma obdavčitve plač je zato lahko eden izmed gradnikov za prehod v gospodarstvo z višjo dodano vrednostjo.

Slovenija se danes nahaja v specifičnem davčnem položaju znotraj Evropske unije: obdavčitev dela, zlasti za visoko kvalificirane kadre, je med višjimi, medtem ko je obdavčitev kapitala in premoženja relativno nizka v primerjavi z večino držav članic. Takšna struktura davčnega sistema ima pomembne razvojne posledice. Visoka obremenitev dela zmanjšuje donosnost znanja, inovacij in podjetnosti, hkrati pa relativno nizka obdavčitev kapitala omejuje možnosti za bolj uravnoteženo davčno strukturo. Posledično to vpliva na nižjo produktivnost, slabšo konkurenčnost in dolgoročno omejuje potencial gospodarske rasti.

Nadaljujte z branjem→

Apple published a paper in June 2025 that called out the entire AI industry.

And the industry has not recovered from it since.

The paper is called “The Illusion of Thinking.” Six Apple researchers. Months of controlled experiments. One conclusion that landed like a grenade.

Frontier reasoning models face a complete accuracy collapse beyond certain complexities.

Complete. Not partial. Not gradual. Complete.

Here is what that actually means.

For two years, every major AI lab has been racing to build reasoning models. OpenAI’s o1, o3. Anthropic’s Claude 3.7 Sonnet Thinking. DeepSeek R1. Google’s Gemini Thinking. These models do not just answer questions, they visibly think first. They show their work. They reason step by step through a problem before arriving at an answer. The entire industry marketed this as the next evolution of intelligence.

Apple tested whether it was real.

They did not use math benchmarks or coding tests, the standard evaluations every AI company optimizes against during training. They built clean, controllable puzzle environments. Tower of Hanoi. River Crossing. Checker Jumping. Blocks World. Problems with precise, verifiable correct answers and zero possibility of data contamination.

Then they systematically turned up the complexity. And watched what happened.

For simpler, low-complexity problems, standard LLMs demonstrated greater efficiency and accuracy, the reasoning models were beaten by regular models that do not think at all. As complexity moderately increased, reasoning models gained an advantage. But when problems reached high complexity, both model types experienced complete performance collapse.

The thinking models, the ones that cost more, take longer, and are marketed as more intelligent, lost to basic models on easy tasks. Then both collapsed completely on hard ones.

But the finding that truly alarmed researchers was not the collapse itself.

Nadaljujte z branjem→

I bought an apple in New York. I brought it home to Europe and placed it next to an Italian apple on my kitchen counter.

The Italian apple was gone in two weeks.

The American apple sat there for almost two years.

I am not a scientist. But I started asking questions. What followed was three years of research into one of the most important and least discussed differences between life in America and life in Europe.

The short version: the food industry has been running two different experiments on two different populations for fifty years.

One population was protected by strict government regulation. The other was left to the industry to regulate itself.

The results are now visible in the hospitals.

In this piece: the eight substances banned in Europe still found in everyday American food, the two gut bacteria your diet is systematically destroying, what a Norwegian hospital discovered when it transferred bacteria from healthy people into IBS patients, why Americans are dying of heart disease at ages Europeans are not, the trick the food industry uses on every health-conscious shopper, and the embarrassingly simple conclusion hundreds of research studies have reached.

You are not doomed. But you do need to know what you are dealing with.

Povezava na članek

NBC je poročal, da je ameriški predsednik Trump nenadoma in na hitro “pavziral” napovedano akcijo “osvoboditve” ladij iz Zaliva z vojaško silo (akcija “Freedom”) , ker sta Savdska Arabija in Kuvajt zaprli zračni prostor za ameriška letala.

Spodnja razlaga, zakaj sta to naredili, zveni smiselno: Ne zato, ker se z akcijo nista strinjali, pač pa zato, ker jima ZDA niso zagotovile zaščite v primeru iranskih povračilnih ukrepov. Ker Trump uporablja zalivske države kot pogajalski drobiž v želji doseči “dogovor” z Iranom.

Having spoken to a senior Saudi official about the NBC article regarding Project Freedom, I honestly think the article completely misunderstood what actually happened because it was written almost entirely from a US perspective rather than from a GCC perspective.

First of all, contrary to the impression being created, the GCC were NOT blindsided by Project Freedom.

They knew about it beforehand. Roughly half a day before. The airspace was opened. The facilities were available. Nobody objected. There was broad support for the idea because, at least publicly, Project Freedom was supposed to be a limited humanitarian-security operation aimed at relieving the 22,000 sailors trapped around Hormuz and allowing shipping lanes to breathe again.

Nobody in the GCC had a problem with that.

But here is the issue .. and this is the part the NBC article completely misses.

Nadaljujte z branjem→

Zahtevana pristojbina za prečkane Hormuške ožine je relativno nizka – med 0.5 in 1 dolarjem za sodček, kar predstavlja manj kot 1 % cene surove nafte pri “normalnih cenah” nafte (okrog 60 $/bl) oziroma pol odstotka cene pri povišanih cenah nafte (okrog 100 $/bl). Ta pristojbina ne bo imela omembe vrednega vpliva na cene nafte. Iran vztraja pri tej pristojbini iz treh glavnih razlogov:

1. s tem pridobi de iure priznano pristojnost nad Hormuško ožino in s tem de facto kontrolira kontrolira pomemben segment globalnega trga z nafto (20 %), s čimer ima vzvod, s katerim lahko prisiljuje ZDA in zahodne države, da proti njemu ne uvedejo sankcij (kontrola na d Hormuško ožino je za Iran podobna posedovanju jedrskega orožja);
2. na ta način lahko zagotovi plačilo vojaških reparacij za škodo, ki sta jo povzročila ZDA in Izrael z nesprovociranim in nezakonitim napadom na Iran; iluzorno je pričakovati, da bi Trump in ZDA sicer psristala na kakršnokoli plačilo vojne škode, pri tem pa pristojbino dejansko plačujejo porabniki nafte izven ZDA;
3. s tem Iran lahko doseže odpravo finančnih sakncij proti svojim bankam, saj morajo plačniki pristojbino sprocesirati prek iranskih bank. 

Seveda Iran s pobiranjem pristojbine sebi zagotovi stalne fiskalne prihodke na letni ravni okrog 100 milijard dolarjev, vendar je to zgolj koristna eksternalija ob zagotovitvi prvih treh točk.

_______

Iran is well aware of the difficulty of getting Trump to agree to pay war reparations, and it is also aware of his manipulation in lifting the freeze on frozen funds.

Trump canceled the previous agreement in 2018 because he doesn’t want to pay; this is a stingy, domineering man who wants to take only and will never pay anything to anyone else under any circumstances.

Collecting the fees ensures the following:

• Obtaining massive reparations
• Collecting an amount equivalent to the frozen funds in Western banks
• Forcing America and its allies to lift sanctions on Iranian banks and the Iranian financial system, which is exclusively authorized to collect those fees.

Iran has stated that only its banks are authorized to collect transit revenues from the Strait, so every country that wants passage must deal with the Iranian financial system.

And this means there are no real financial sanctions, and the world will be forced to deal with Iranian banks despite the United States.

The Strait of Hormuz and the issue of fees are a matter of life or death for the Iranians; it is the gateway through which they will open up to the world, and it is the financial resource through which they rebuild the infrastructure.

Therefore, even if America pays the frozen funds or even lifts the sanctions, the Iranians will not abandon collecting the fees; this is the only guarantee that preserves for them a resource alongside oil with which they compensate for the losses of 47 years of siege.

Barak Ravid je izraelski novinar, rojen v Izraelu, ki trenutno deluje kot Global Affairs Correspondent za Axios, politični analitik za CNN in dopisnik izraelskega Channel 12 News. Je avtor knjige Trump’s Peace in eden najbolj vplivnih diplomatskih poročevalcev o bližnjevzhodni politiki ZDA in Izraela. Njegovo delo temelji na globokih virih v Washingtonu in Jeruzalemu, zaradi česar pogosto objavlja ekskluzivne novice, ki vplivajo na javno razpravo in trge.

Pred Axiosom je Ravid delal za izraelske medije: začel je na NRG (2005), nato deset let kot glavni diplomatski dopisnik pri Haaretzu (2007–2017), kjer je bil znan po kritičnem pogledu na Netanjahuja, potem pri Channel 10/13 News (2017–2020). Leta 2020 je prestopil k Axiosu kot Middle East correspondent (sprva iz Tel Aviva, kasneje iz Washingtona), kjer piše tudi za Walla News. Njegov slog je “smart brevity” – kratke, z viri podprte novice, ki pogosto postanejo referenčne v diplomatskih krogih.

Pri Axiosu in CNN-u je Ravid ključni glas o ameriško-izraelskih odnosih, iranskem jedrskem programu, Abrahamovih sporazumih in vojni v Gazi. Objavil je številne “scoops”, ki so vplivali na politiko v Washingtonu, Jeruzalemu in Evropi – na primer podrobnosti o pogajanjih Trumpove administracije z Iranom, Hamasom in Ukrajino. Leta 2024 je prejel nagrado White House Correspondents’ Association za odličnost v pokrivanju Bele hiše. Njegovi viri so pogosto anonimni “uradniki ZDA”, “izraelski viri” ali “viri z znanjem”, kar mu omogoča hitre ekskluzivne informacije.

Ravid pa v zadnjih osmih tednih ni nastopal zgolj kot diplomatski poročevalec, temveč kot eden ključnih prenosnikov tržno občutljivih signalov iz Bele hiše o vojni z Iranom, Hormuški ožini in možnem ameriško-iranskem dogovoru. Njegova vloga je kontroverzna, ker so se pred nekaterimi objavami pojavile nenavadno velike short pozicije na terminskih trgih z nafto. Axios je sam poročal o “epidemiji sumljivega trgovanja” pred Trumpovimi najpomembnejšimi odločitvami, pri čemer naj bi se trgovanje pojavljalo minute ali ure pred tržno občutljivimi političnimi premiki. (Axios)

Nadaljujte z branjem→

Raziskovalci Dongyang Fan, Sebastien Delsad, Nicolas Flammarion, Maksym Andriushchenko so v pravkar objavljeni raziskavi »HalluHard: A Hard Multi-Turn Hallucination Benchmark« pokazali, da LLM modeli umetne inteligence v veliki večini primerov (med 60 % in 75 %) na vseh zajetih strokovnih področjih »halucinirajo«. Torej v 60 % do 75 % primerov (v 3 izmed 4 primerov) si izmislijo odgovor, ki ni v skladu z dejstvi. Po domače: zlažejo se. In dlje kot jih sprašujete, bolj se lažejo in svoje prejšnje – izmišljene – odgovore uporabijo kot dejstva. Ko so raziskovalci uporabili modele UI, ki imajo neposreden dostop do interneta in od modelov zahtevali, da verificirajo svoje odgovore s citati na dejansko objavljena dejstva, so še vedno halucinirali v več kot 30 % primerov (v 1 od 3 primerov). Predstavljajte si, da zdravstveni in pravni nasvet ali pomoč pri kodiranju programov iščete pri UI.

Spodaj je najprej povzetek članka:

Large language models (LLMs) still produce plausible-sounding but ungrounded factual claims, a problem that worsens in multi-turn dialogue as context grows and early errors cascade. We introduce HalluHard, a challenging multi-turn hallucination benchmark with 950 seed questions spanning four high-stakes domains: legal cases, research questions, medical guidelines, and coding. We operationalize groundedness by requiring inline citations for factual assertions. To support reliable evaluation in open-ended settings, we propose a judging pipeline that iteratively retrieves evidence via web search. It can fetch, filter, and parse full-text sources (including PDFs) to assess whether cited material actually supports the generated content. Across a diverse set of frontier proprietary and open-weight models, hallucinations remain substantial even with web search (≈30% for the strongest configuration, Opus-4.5 with web search), with content-grounding errors persisting at high rates. Finally, we show that hallucination behavior is shaped by model capacity, turn position, effective reasoning, and the type of knowledge required.

Če iste nič razumeli, ali ne dovolj, je spodaj en tak poljudni povzetek članka.

Nadaljujte z branjem→