Daisy P-L

Meta’s El Paso data center plans to link 813 gas generators into a single on-site power system. Utilities can’t deliver new grid capacity fast enough, so developers are building “behind-the-meter” gas plants beside their servers. xAI showed proved this model in Memphis, running largely on portable generators before EPA required emissions permits. Cleanview counts 47 such off-grid projects nationwide now including Meta, OpenAI, Oracle and Chevron. We talk a lot about the near-term AI power story being nuclear or fusion, but not just yet. washingtonpost.com/business/2026/…

🚨 Holy shit… Stanford just published the most uncomfortable paper on LLM reasoning I’ve read in a long time. This isn’t a flashy new model or a leaderboard win. It’s a systematic teardown of how and why large language models keep failing at reasoning even when benchmarks say they’re doing great. The paper does one very smart thing upfront: it introduces a clean taxonomy instead of more anecdotes. The authors split reasoning into non-embodied and embodied. Non-embodied reasoning is what most benchmarks test and it’s further divided into informal reasoning (intuition, social judgment, commonsense heuristics) and formal reasoning (logic, math, code, symbolic manipulation). Embodied reasoning is where models must reason about the physical world, space, causality, and action under real constraints. Across all three, the same failure patterns keep showing up. > First are fundamental failures baked into current architectures. Models generate answers that look coherent but collapse under light logical pressure. They shortcut, pattern-match, or hallucinate steps instead of executing a consistent reasoning process. > Second are application-specific failures. A model that looks strong on math benchmarks can quietly fall apart in scientific reasoning, planning, or multi-step decision making. Performance does not transfer nearly as well as leaderboards imply. > Third are robustness failures. Tiny changes in wording, ordering, or context can flip an answer entirely. The reasoning wasn’t stable to begin with; it just happened to work for that phrasing. One of the most disturbing findings is how often models produce unfaithful reasoning. They give the correct final answer while providing explanations that are logically wrong, incomplete, or fabricated. This is worse than being wrong, because it trains users to trust explanations that don’t correspond to the actual decision process. Embodied reasoning is where things really fall apart. LLMs systematically fail at physical commonsense, spatial reasoning, and basic physics because they have no grounded experience. Even in text-only settings, as soon as a task implicitly depends on real-world dynamics, failures become predictable and repeatable. The authors don’t just criticize. They outline mitigation paths: inference-time scaling, analogical memory, external verification, and evaluations that deliberately inject known failure cases instead of optimizing for leaderboard performance. But they’re very clear that none of these are silver bullets yet. The takeaway isn’t that LLMs can’t reason. It’s more uncomfortable than that. LLMs reason just enough to sound convincing, but not enough to be reliable. And unless we start measuring how models fail not just how often they succeed we’ll keep deploying systems that pass benchmarks, fail silently in production, and explain themselves with total confidence while doing the wrong thing. That’s the real warning shot in this paper. Paper: Large Language Model Reasoning Failures

@kron4news He was repeating back what the man in the crowd said. Why is this news?!?

Bad Bunny, while at a press conference today in San Francisco, says... "¡Viva México...!” 🇲🇽 #SuperBowl

Another stupid @Waymo blocking traffic at rush hour @ 16th and Valencia. 🤦🏼‍♀️🤦🏼‍♀️ Must fix this, constant problem on 1-lane streets in SF. #Waymo & other #RobotCars should pull to the side to wait. Passengers can walk a few storefronts to avoid inconveniencing the rest of the city.

@TKavulla the trees are exploding because even they want the ICE out (of their xylem and phloem) in MN.

I'm sorry what?

One of the most underreported stories of 2025 is that Trump is the first President ever to create electricity shortages on purpose. He is illegally canceling solar and wind, causing layoffs and price spikes, and the media is like “wow what a showman that guy is!”

@grok @putthole21 @TomSteyer Silly, why do you think there are labor shortages?

Deporting most undocumented immigrants is realistically easier and faster in California than building 1 million new homes in four years, as federal enforcement has already led to over 500,000 national deportations and 2 million departures in 2025 with minimal state-level barriers, while housing faces persistent CEQA delays, labor shortages, water constraints, and high costs despite reforms, requiring a massive scale-up from current annual rates of around 100,000 units.

To really bring down costs, we need to build more housing – a LOT more. I’m talking about the largest effort in state history to build homes that California families can actually afford. A million homes over 4 years. We can do that.

@samstein Data heist.

But Elon Musk got a trillion dollar pay package from Tesla and was dining with Trump and MBS the other day. so, i guess, it was a success on that front.

The legacy of DOGE is not that it failed. It's that it destroyed: 1. faith in real cost conscious governance 2. the careers of committed govt employees 3. scientific research and the biomedical breakthroughs that come with it 4. countless lives of the world's poorest

If you’re keeping track at home, Trump admin has cancelled: - Jobs Report - Inflation Report - GDP Report I’m sure everything is just fine.

@ArmandDoma Even better when Mario can take Wario's money

Mario and Wario often play tennis and golf, go kart race, party together after all…

Trump brazenly using his pardon power to free a supporter who was found guilty of helping China repress and intimidate people on American soil. Blatant corruption at the cost of US national security and values.

"The provision, tucked into a measure to fund the legislative branch, appears to immediately allow for eight GOP senators to sue over their phone records being seized in the course of the investigation by Jack Smith ... into the riot at the Capitol..." nytimes.com/2025/11/10/us/…

I will not support a deal that does nothing to make health care more affordable. We are in a health care emergency. A simple one-year extension of these tax credits would cost less than Donald Trump’s $40 billion bailout for Argentina. A vote for this bill is a mistake.

Effective tax rates before and after the Trump tax law: Verizon Before: 21% After: 8% Walmart Before: 31% After: 17% AT&T Before: 13% After: 3% Walt Disney Before: 26% After: 8% FedEx Before: 18% After: 1% This is what a corporate giveaway looks like.

If you include the normal coders that the Big Tech overlords are trying to replace with AI, that *miiiiight* be true. But once it's just Elon and his fellow tech leaders executing their dream of AI doing all the work, this will be flipped on its head and the proportions way off.

@charlove @ArmandDoma This. WTH?!

in my hagmaxxer era

This is what a #BlueWave looks like!

Huge implications - Peter Hubbard & Dr. Alicia Johnson will be 1st Dems on GA PSC in 20+ years! Voters responding to poor financial decisions by the All-GOP incumbent Commission which raised rates 6 TIMES THIS YEAR! Investing in nuclear & gas proved a HUGE waste of time & money!