"Surprisingly it only supports reasoning "none" or reasoning "high".That setting didn't seem to make any real difference - it added a tiny bit of thinking trace and high actually produced less output tokens than none.The high bicycle frame is better then the none one though.Pelicans: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...(Definitely the best I've seen from any Mistral model: https://simonwillison.net/tags/pelican-riding-a-bicycle+mist... )"

"Impressive vision benchmarking. If the vision model is truly as good as astra, that would make it best in the world.Also strong on cyber benchmarks (better than all chinese models), so this is a good defender model.Lots of people shitting of Mistral for no reason imo. These are pretty good numbers across the board. Definitely good enough to use as a daily driver over other llms, if you have moral qualms with the others. For certain use cases, like cyber security, this may be the go to model.I like to make fun of europe, but there's lots for mistral to be proud about in this release imo."

"I have a question, and perhaps some of the AI/ML infrastructure experts here could answer: Mistral says, "ML4 was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral’s own datacenters in Europe."If a 1T params model trained on ~4k NVIDIA GB GPUs could almost match the performance of Kimi's K3 (which is on par with top closed source models of OpenAI/Anthropic) while beating/exceeding other leading SOTA models from top Chinese labs, what are we (in the US) even building these super massive data centers for? Just to churn through more backpropagation reps more quickly?SpaceXAI's Colossus supercluster in Memphis and Colossus 2 in Memphis/Mississippi (Southaven) are supposed to run into hundreds of thousands to a million GPUs. MSFT's Fairwater GPUs are supposed to have hundreds of thousands as well. So, 3800 GB GPUs are an absolute drop in the bucket. I don't understand the strategy of hyperscalers here, especially with edge inference hardware only getting better from here on (Apple, and all).Distillation explains some of the advances, but doesn't that mean hyperscalers have a ton of deadweight wrt GPUs sitting on their balance sheets? Will all these GPUs be used for inference once a SOTA model's training checkpoint/batch is done? It's bonkers to me."

"I was a graph theory junkie long ago and even moved to Budapest for awhile to study among the greats. While I was there I started working on Barnette's Conjecture which came to occupy my thoughts over the next 24 years of my life, on and off as I worked in many different fields. Last summer I even thought for a few days that I had actually solved it.But it's supposedly proven here - problem 180. I don't know what to think exactly. I spent thousands of hours on that problem. I really enjoyed it. Hearing that it is solved somehow makes me sad in a far-off way, like hearing an ex-girlfriend died suddenly in a car crash. I don't know, there's probably a lot of people feeling odd emotions tonight.There's no Lean proof for this one so I'm digesting the paper. On the surface it looks like an approach I considered 24 years ago and abandoned.I revisited the problem this summer, along with my partial solutions, when the previous round of stunning proofs came out. Several hours of work with Fable simply convinced me it wasn't yet solvable and reinforced how hard of a problem it was."

"Beyond the reasoning capabilities of the unreleased model and that they can run a large number of agents in parallel, what makes me curious is that the writing of the proofs is quite human-readable. This is in contrast with the scientific text produced by the ChatGPT available to us, which writes horribly in a way that no human would write. One tell-tale is that they constantly attempt to be defensive and cover all edge cases like division by zero etc. that are clearly a non-issue for humans in some proofs or at least a human would not add this to the main statement, but AI is so overly careful that makes reading its proofs impossible. On the other hand, these new OpenAI proofs are very good."

"> We prove the Unique Games ConjectureThe Unique Games Conjecture (sorry, "Unique Games Theorem" now!) is huge. It was a very significant pillar supporting many of the limits of the polynomial-time approximation algorithms in the graduate-level randomized and approximate algorithms course I took in theoretical computer science. Textbooks will have to be re-written.Here is an explainer: https://share.gemini.google/nbjIK6X3tOfzWith UGC proved, certain polynomial-time approximation algorithms used in difficult real-life problems are now known to be the best approximations we can achieve in polynomial-time:> If UGC holds, the elementary algorithm that grabs both ends of an edge is fundamentally the best efficient algorithm that will ever exist. No amount of advanced linear programming or heuristics can achieve a ratio of 1.999.> Under UGC, the Goemans-Williamson algorithm's 0.87856 ratio is mathematically optimal.> UGC is considered the "Rosetta Stone" of approximation algorithms. In 2008, Prasad Raghavendra proved that for every single constraint satisfaction problem (CSP), a canonical Semidefinite Programming relaxation paired with the best rounding scheme achieves the optimal approximation ratio if and only if UGC is true. If the conjecture holds, the algorithmic boundary for an entire class of combinatorial problems is completely resolved.Other hardness of approximation results from this UGC proof:> [Max acyclic subgraph, a problem encountered in real life]: No polynomial-time algorithm can fundamentally outperform an unthinking coin toss.> [Relative scheduling, another realistic problem]: As with acyclic subgraphs, the problem is "approximation-resistant": clever algorithms cannot beat random shuffling."

"All the death knell comments - Is no one looking at the revenue line? Revenue still trending the same. Costs presumably haven't skyrocketed. They've invested in something big. Thats probably a good thing, and often needed to evolve."

"A bit tongue-in-cheek but I'm amazed by people that don't open an IDE or look at code. Are you paid to press enter at anything that Claude present to you?Personally I usually use Claude inside the Jetbrain's IDE terminal, even if I code by hand obviously less. I still like using the diff, navigation, quick access to any file path for my prompt etc but maybe that won't last.(edit: although I currently use an old perpetual fallback license because I didn't like the last price increase especially as they slowly remove what I like in the UI)"

"Wow, this really sucks. With all the money flocking to companies that trained LLMs on stolen data, we may lose the only set of IDEs that offer high quality professional development tools"

"A breakdown of why the awarded work, Ice Cube, is significant: - A neutrino is an elementary subatomic particle. Neutrinos are produced by nuclear reactions inside stars, supernovae, radioactive decay. They are one of the most abundant particles in the universe. - Neutrinos are known as "ghost particles". They have 0 charge and near-zero mass. They only react with the weak nuclear force and gravity. Incredibly hard to detect. Trillions can pass through a whole planet without hitting a single atom! - Neutrinos give us a pristine snapshot into the origin of the universe. They have travelled billions of years and trillions of miles without interacting with anything. Unless...we catch them! - Ice Cube does this. Located in Antarctica, the project turns a cubic kilometer of ice into a neutrino detector. How? Scientists drilled boreholes 2.5km deep into the ice and placed 5000 optical detectors to catch a neutrino interaction. - When a neutrino, rarely, collides with an atom, it produces charged particles. In ice - not vacuum! - certain particles can travel faster than light. This produces something similar to a sonic boom. A faint, blue glow known as Cherenkov radiation (see it in action https://youtube.com/watch?v=hSuSG19Pcoc). - IceCube was first to detect neutrinos coming from outside the solar system, establishing the source of high energy cosmic radiation. It also opened a whole new chapter of neutrino astronomy."

"He receives the prize for conceiving the IceCube neutrino detector, a cubic-kilometer-sized detecter in the Antarctics.https://en.wikipedia.org/wiki/IceCube_Neutrino_ObservatoryMechanism is via conversion of neutrinos into charged particles which are then detected via Cherenkov radiation which is produced when a charged particle moves with speeds larger then the speed of light in the medium. (That is only possible because it is less than the speed of light in vacuum which cannot be exceeded.)This was also discussed recently if you are interested: https://news.ycombinator.com/item?id=49655286https://en.wikipedia.org/wiki/Cherenkov_radiation"

"I'm in love with the cute little figure that came with the press release: https://www.nobelprize.org/uploads/2026/10/fig_fy_26_3x2.jpgI appreciate the boldness of this project, it has an element of sci-fi to it. Building a base at the south pole to bury sensors in ice to measure elusive particles. The stuff of dreams!"

"Some more discussion: https://news.ycombinator.com/item?id=49977979"

"Dupe.https://news.ycombinator.com/item?id=49977979 (200 comments now)"

"Did they release it again? https://news.ycombinator.com/item?id=49977979"

"I've used Polars before and can only recommend it.It's like you get a really good 'query planner' like a DB would give you, but for your notebooks/scripts/etc. Much better than pandas imo."

"I'm not following the trends closely, but has Polars become a full replacement for Pandas? Are there use cases where one is better suited than the other?"

"Well done, Polars team!Everything that I build greenfield moving forward I plan to use DuckDB, Polars, or PyArrow. Pandas was a great grandfather of a project (I actually cut my OSS contrib teeth on it, how the time flies)! I'll always appreciate the improvement pandas brought over SAS."

"I’m pretty frustrated with Muse and the last thing I want to be doing with my free time is defending Meta, but this is such clickbait.Point by point:> “OMG you can jailbreak it and get it to spill its VM”This is the whole point; any content on the VM is yours. It runs in an isolated sandboxed VM separate from stored credentials etc; this is effectively your own computer. You don’t have to trick it.> It collects dossiers on your contactsThese are more text files that live on your private VM, alongside memory.md, etc. Do you want your secretary to forget every person you contact every day?> It accessed Messages without full disk accessThis whole story never made sense or was substantiated. Full disk access is an OS level security boundary; the user had to switch this on.> It sold some guys stuff for too cheap and gave out his addressOK this one I basically believe, haha. Because this is the problem with Muse: the LLM is just too dumb to perform complex tasks effectively in many cases."

"I get the appeal of agents, I do. This is the future stuff we always wanted. But I cannot get myself to give one of these things access to my bank account or allow it to do price comparsion and shopping without oversight. Or access to my email or chat history.I just do not trust any of them, not with my money or with access to my conversations."

"It's very simple. There is ZERO chance the worlds biggest advertising companies will refrain, long-term, from using your most intimate secrets, gathered through your many conversations with their AI personal assistants, to sell you things."

"For the desperate people whose tests all just broke, I have a page on my public endpoint testing server that exactly reproduces the classic design: https://example.testserver.host/. You can just use update the URL and go back to blissful ignorance.Also open-source and self-hostable: https://github.com/httptoolkit/testserver. Lots of other endpoints too, full docs here: https://testserver.host/"

"I'm wondering how many automated tests this change broke.Yes, I know it's not the example.com's fault, and it's a side effect of Hyrum's Law:> This is not a service, avoid relying on it for testing and monitoring purposes.Don't we all have a few fragile tests?"

"> there's a gradual opacity transitionLooks like they removed this and instead just show all languages with no CSS animation now."

"Lamps are a fantastic way to light an interior space. Recessed ceiling lights do not establish hierarchy in the dimensions that humans are most interested in. Getting the lighting source close to eye level is the key to success.At some level this turns into an accessibility and comfort issue. People with limited vision can take strong advantage of well placed lights. Highlights and shadows are some of the most fundamental elements of human visual perception. Without them our brain needs to work overtime to analyze a given scene."

"I've had the BenQ WiT reading desk lamp for nearly five years, and it has been almost flawless until now. I always use it to light my desk when working on the computer, and it's also fine for any detail work or whenever I'm reading on the desk. It could be improved with some design changes to reduce lateral glare, but this could impact the wide throw of the light which entirely covers my desk."

"I love iconic minimal lamps too. I never saw the Lampe de Bureau nor the Kismas Lamp 01, love them! You might like the Muuto Post."

"On this topic, I can never end recommending Adam Curtis's series "All Watched Over by Machines of Loving Grace", and on this topic of "bounce back" the second episode in that series.The concept of "natural equilibrium" is a 1950s cybernetic machine fantasy projected onto the natural world rather than an accurate reflection of ecological reality. We can model the world in the language of system theory, but the models we make there might not have much to say about the real world. We can claim that from system theory it follows that there must be an equilibrium (as somewhat of an expert, I can say that even that is not true) and that if we disrupt the system, things will eventually go back to that equilibrium. And if we don't see it, we just needed to wait longer.The whole stack of logic is not really build on any proper reasoning. The parallel with economics based on the "homo economicus" is striking, also there you reason from an ideal model and try to make inductive statements about the real world, which in many places directly conflict with observation. The "eco-system" and "natural equilibrium" stem from this approach of applying system theory to ecology, and are ideas which are now _very_ ingrained in western thinking, but which are obviously off.Here is the start of that second episode: https://www.youtube.com/watch?v=FgoereMDUJ0"

"I think what happened to the cod fishery off North America is really interesting. They overfished the cod then decided to regulate the fishery to let the population come back. But it never recovered - the cod population is stable but at a much lower number. Other 'stuff' (other fish species, jellyfish, etc) had moved into the environmental niche that the cod were in."

"I had always assumed this, and my idea of nature bouncing back was more on the millions of years time scale as evolution would result in the refilling of empty ecological niches.All that said I still think we should just extinct disease carrying mosquitoes."