0S70

USA 🇺🇸 Malgré la prolongation du cessez-le-feu, les États-Unis activent d'importants flux aériens vers le Moyen-Orient afin de ravitailler les forces sur place. L'Amérique recharge ses capacités pour frapper l'Iran dans le cas d'un retour à la confrontation.

Moonshot’s Kimi K2.6 is the new leading open weights model. Kimi K2.6 lands at #4 on the Artificial Analysis Intelligence Index (54) behind only Anthropic, Google, and OpenAI (all 57) Key takeaways: ➤ Increase in performance on agentic tasks: @Kimi_Moonshot's Kimi K2.6 achieves an Elo of 1520 on our GDPval-AA evaluation, which is a marked improvement over Kimi K2.5’s Elo of 1309. GDPval-AA is our leading metric for general agentic performance, measuring the performance on knowledge work tasks such as preparing presentations and analysis. Models are given code execution and web browsing tools in an agentic loop via our open source reference agentic harness called Stirrup. This continues Kimi K2.6’s strength in tool use, maintaining a 96% score on τ²-Bench Telecom, placing it among other frontier models in this category. ➤ Low hallucination rate: Kimi K2.5 scores 6 on the AA-Omniscience Index, our knowledge evaluation measuring both accuracy and hallucination rate. This score is primarily driven by a comparatively low hallucination rate of 39% (reduced from Kimi K2.5’s 65%), indicating a greater capability to abstain rather than fabricate knowledge when the model is uncertain. Kimi K2.6’s low hallucination rate places it similarly to other models such as Claude Opus 4.7 (36%) and MiniMax-M2.7 (34%) ➤ High token usage: Kimi K2.6 demonstrates high token usage, but is in line with other frontier models in the same intelligence tier. To run the full Artificial Analysis Intelligence Index, Kimi K2.6 used ~160M reasoning tokens. This is slightly lower than Claude Sonnet 4.6 (~190M reasoning tokens) but much higher than GPT 5.4 (~110M reasoning tokens). ➤ Open weights: Kimi K2.6 is a Mixture-of-Experts (MoE) model with 1T total parameters and 32B active, same as the previous two generations of models Kimi K2 Thinking and Kimi K2.5. Kimi K2.6 again pushes the open weights frontier in intelligence. ➤ Third Party Access: Kimi K2.6 is accessible through Moonshot’s First Party API as well as third party API providers Novita, Baseten, Fireworks, and Parasail ➤ Multimodality: Kimi K2.6 supports Image and Video input and text output natively. The model’s max context length remains 256k. Further analysis in the threads below.

It happened. An open weights model just dropped that benchmarks higher than Opus 4.6 is out If you have 2 Mac Studios w/ 512gb, you can run Opus 4.6 level intelligence completely for free on your desk I warned you this would happen months ago. Now Mac Studios and Mac Minis are sold out The next Mac Studio has been delayed until Q3/Q4. The price will be significantly higher I told you this was going to happen. Intelligence explosion. Hardware bottleneck. Increased efficiency Luckily I picked up 2 Mac Studio 512gbs, 2 Mac Minis, and a DGX Spark I will be loading this up in the next couple of days and will have completely private super intelligence running for me 24/7 I’m telling you right now by end of year we will have a local version of Mythos. It’s 100% guaranteed You called me crazy but every single prediction I’ve made has turned out to be true These models will only get more efficient and require less hardware. But that hardware is only going to get more expensive Local/open source is so obviously the future and if you’re still denying this now you are delusional

An idea we shared recently as a way to play the current geopolitical environment outside of tankers, ags/fert or O&G names: companies that benefit from an increased focus on energy sovereignty by the EU and Japan. A bonus here is that the price of electricity in the EU is mostly tied to the price of natural gas. This means a windfall for companies that can produce electricity without gas as an input cost.

Neuralink vient d'entrer dans la phase clinique humaine de Blindsight. L'implant stimule directement le cortex visuel, contourne les yeux et le nerf optique, et doit restituer une forme de vision à des patients aveugles de naissance ou devenus aveugles après destruction des voies oculaires. La première génération livrera une image pixellisée, proche d'une caméra CCD de 1995. Elon Musk a tempéré les attentes : pas de couleurs ni de contraste fin, rien qui ressemble à la vue naturelle. Mais pour quelqu'un qui n'a jamais vu, distinguer les contours d'une porte ou repérer un visage qui approche représente un saut de nature, pas un progrès incrémental. La résolution viendra avec l'itération. Le projet était une idée folle en 2016. Il est devenu Breakthrough Device à la FDA en septembre 2024. Il passe aux essais humains en 2026. La courbe est classique des révolutions médicales : cinq à dix ans de labeur invisible, puis le marché découvre qu'une promesse de science-fiction est devenue un protocole. Les régulateurs européens, pendant ce temps, travaillent sur la classification des dispositifs médicaux IA comme "haut risque par défaut". Dans deux ans, on se fera opérer en Arizona en payant cash, pendant que l'agence nationale éditera la version 3.2 de son guide de bonnes pratiques.

🇭🇺 #Hungary Starts to Report Early Election Results - Bloomberg bloomberg.com/news/live-blog…

GUYS ZO COMPUTER CAN BE POWERED BY YOUR EXISTING CLAUDE SUBSCRIPTION BECAUSE IT USES ACP AND BREAKS NO RULES Since Zo runs as ACP when you link it to your Claude's monthly subscription, its usage is completely allowed under Anthropic rules You can text him via iMessage (like i've been showing you) anything that you're already doing with Openclaw + self hosts too with your Zo Sub I think this is one of the stronger harnesses I've seen that can still be powered by your monthly Claude subscription and do most of the same things that Openclaw can do How did I not know this 😭

That’s life.

Anthropic just announced Mythos, a model so good at hacking that they chose not to release it. Mythos found thousands of zero-day vulnerabilities in every major operating system and every major web browser. It autonomously chained together Linux kernel exploits to go from ordinary user to full machine control. My hot takes: 1/ If Anthropic wanted to, they could take over most servers on the planet right now. Let that sink in. 2/ This is the first time a model exists that, if released open source, would pose a credible risk of civilizational catastrophe (collapsing the internet). 3/ Some will say Anthropic is being responsible by not releasing it and only giving access to trusted partners for defense. Don't fall for this narrative. They built a weapon of mass destruction. Thank god they're not releasing it into the wild. But it's insane that they were allowed to develop it without any oversight whatsoever. 4/ Now ask yourself: how far behind are other labs? Meta has repeatedly signaled they would open-source a model like this the moment they have one. The only reason they haven't is that they're thankfully incompetent (their superintelligence team keeps delaying releases because performance is disappointing). And OpenAI? They have an unreleased model codenamed Spud. Do we trust Sam Altman, a compulsive liar who has essentially dismantled his safety team, to have potential root access to most machines on the planet and not pursue his goals recklessly? Read the recent New Yorker investigation (link below) and decide for yourself. 5/ This highlights a brutal asymmetry. Attack is now automated. Defense must catch up across every single piece of software in the world before a model with Mythos-level capabilities leaks or gets released openly. The task is colossal and the clock is ticking. 6/ The world is in a significantly worse place than it was last month. Anthropic didn't protect us. They created the threat, and now they're asking us to trust them with the solution. We need oversight, regulation, international coordination, and a Pause. Not corporate goodwill.

Whoever does set design for Kanye needs a raise.

Incredible footage of the crewed Artemis II mission launch to the Moon

SUMMARY OF PRESIDENT TRUMP'S ADDRESS TO THE NATION: 1. The Iran War will last another "two to three weeks" 2. The US will strike Iranian power plants if no deal is reached 3. Core strategic objectives are "close to completion" in Iran 4. The US "will bring Iran back to the stone age" 5. The US will not import oil from the Strait of Hormuz in the future 6. "Iran's navy is gone and their air force is in ruins" US oil prices are surging above $103/barrel after the speech.

In honor of 50 years of Apple, we're sharing - for the first time ever - Don Valentine's original 1977 memo for Sequoia's investment into Apple Computer. #Apple50

We spoke to diplomats, advisers, scholars, experts and current and former officials in China. Almost all of them see the war in Iran as a grave American error. Register for free to read why: econ.st/416vZDK

OpenAI shares have fallen out of favor on the secondary market — in some cases becoming almost impossible to unload — as investors pivot quickly to Anthropic, its biggest competitor. bloomberg.com/news/articles/…

@WonderingApp is new my fav app rn

@amitisinvesting that’s crazy

SPACEX HAS OFFICIALLY FILED FOR AN IPO. SpaceX Revenue — $15B, targeting a $1.75T valuation $META Revenue — $200B, currently at a $1.45T valuation Are you buying the SpaceX IPO?

BREAKING: SpaceX has confidentially filed for an IPO, putting the company on track for a June 2026 listing, per Bloomberg.

🇺🇸 Traders Trigger Euphoric Rally With Iran, Trump Seeking Exit - Bloomberg bloomberg.com/news/articles/…