DHH is a few weeks away from realizing Java is also great when you don’t look at the code

GPT-6 just cracked a 60-year-old fusion physics conjecture. For six decades, physicists have wrestled with the math governing magnetic confinement in fusion reactors. The equations describing plasma turbulence and heat loss are notoriously chaotic. A foundational conjecture proposed in the 1960s remained unproven. Human mathematicians hit a hard wall. The complexity was simply too massive to untangle. Until now. Researchers pointed a frontier AI model at the problem. Instead of blindly guessing, the model generated a rigorous, step-by-step mathematical proof that verified the 60-year-old hypothesis. It didn't just calculate numbers. It discovered abstract mathematical structure. Clean, limitless fusion energy is the holy grail of human civilization. The single biggest bottleneck has always been our inability to fully model and control high-temperature plasma. If AI can solve six-decade-old theoretical physics problems in its spare time, the timeline for commercial fusion just shifted forward. We spent the last century trying to force the universe to reveal its secrets through human intuition. Now, we just have to ask the model.

Math is easy now

Lattice Deduction Transformers has been accepted at @NeurIPSConf! 🎉 We're excited to present this work and grateful to everyone who has engaged with it so far. Here's a glimpse of LDT in action and how it works. We’re also preparing an improved revision of the paper, more soon!

Die Energiewende kostet eine Familie soviel wie eine Eisdiele.

Guys it’s over you don’t have to do this anymore

Singapore realising their education system raises NPCs with the survival skills of a baby panda and cannot survive outside a controlled environment. Solution: just plan their whole life as if it were secondary school. Next up, points for participating in adult CCAs and CIP.

Adam Neumann is a very problematic founder, but as a business thinker he is unironically excellent.

The reason the US succeeds is because we have people with biology Ph.D.s who can quit on a dime and go work in software, not because we have them locked into aerospace engineering at 14.

the ending is always the same

CS academia is dead. So where does that leave AI PhDs like me? Did my ICLR reviewer bidding today. Skimmed a few abstracts, and the methods are the exact same recipe I learned when I got into 3DV two years ago. Swap in a newer video gen base model and boom, new paper 🥲 Makes you wonder how many of the 60k submissions were actually thought up by AI. Meanwhile, World Labs' Atlas has basically solved 4D scenes. Academia is so far behind it's not even funny. I still remember the day GPT-6 Astra dropped. My feed was flooded with GPT + Blender doing inverse graphics and GPT driving robot arms through manipulation tasks. The results were so good I literally had to sit down. A year ago, I was dead sure LLMs could never have spatial intelligence. A year later, Astra slapped that belief right out of me 💀 There's no doubt Astra was post-trained on tons of 3D and manipulation data, and it's only going to get bigger and faster. To me, that means any domain that can be represented symbolically, with clean benchmarks for RL, is going to get swallowed by LLMs. Next to real LLM intelligence, most academic papers that add a bit of inductive bias and tune their way to SOTA are just roadkill waiting to happen. Sadly, these papers keep piling up and flooding every conference. It's inertia, plain and simple. We've been chasing SOTA for so long that it's hard to stop overnight. But make no mistake: the paradigms in a lot of areas converged long ago. What used to be research is now pure engineering, and the room for academics to add inductive biases is only going to shrink. So as a PhD student, I see two paths left. One: if you can't beat them, join them. Clean data, build infra, then go to industry and train foundation models. Two: go back to real science. Stop caring about squeezing out another 0.1% on a benchmark, and start asking why this works and that doesn't, with foundation models themselves as the object of study. As a friend put it: AI research might end up looking more and more like biology, except the organisms are silicon-based 🤣 Beyond these two, it's hard to see anything that won't get eaten by LLMs. Still, I'm pretty pessimistic. I can feel the value of human knowledge being eroded bit by bit. Everyone will get their own AlphaGo moment. What will academia even look like after this... 😭

boomer origin stories are always like "Finally at 35 years old I decided to get serious with my life -- fortunately there was a national shortage of hedge fund analysts"

PSA: Math and coding are the two most black-and-white fields there is, and AI's success in them does not easily generalize to others. Jumping from "AI solved a math problem" to "AI will kill us all" is not just absurd; it's deranged.

Doesn't work Never will Natural language by design isn't sufficiently precise

GPT3 writing code. A compiler from natural language to code. People don't understand — this will change absolutely everything. We're decoupling human horsepower from code production. The intellectual equivalent of the discovery of the engine. player.vimeo.com/video/42681…

Holy mother of EV transition

Black Forest Labs founder @robrombach: the startup is registered in the US. Why? A pure German GmbH wasn’t an option: Registering the GmbH took longer than training their first model, Flux 1. That’s just too slow. They need top investors – and for that, a US company is the gold standard. source: handelsblatt

Yesterday I asked Astra to categorize several business meal expenses and to add more description based on short snippets I provided. It did not add more description and instead spammed some slop into my spreadsheet, and I pointed this out, and it apologized. Do people with the view below actually do any real work whatsoever?

Since there's a lot of RLM fandom going around, I hope no one gets mad at me for asking the obvious. I read the paper, I read the blog post, and I worked through some examples. I am still scratching my head about: 1) what's new here than what other agentic harnesses (CC mainly) are doing, 2) what's one straightforward example where RLM wins that all models and agentic harnesses fail? For example, the "quickstart" example feels like an RLM counterexample.

whenever something is crazy hype and is followed up by an armada of PR, you gotta bring out the bullshit detectors so i cloned the repo and did a little digging and found a lot of sound and fury. please enlighten me if any of these facts are untrue, would love to be educated RLM claim either overblown or false this doesn't even appear to be RLM (arxiv.org/abs/2512.24601)? RLM is "cool" as a concept because there's a premise of going infinitely deep to do arbitrary decomposition of a problem. in practice though it's not really hard, you just make subagents callable as functions and wrap that in a code execution tool. the reason RLMs have not been productionized and is just seen as a research ditty is that 1) going arbitrarily deep down the stack is a shitty thing to do in production without guardrails, 2) most problems do not need more than 2 layers of decomposition, and 3) agents are bad at banana phoning each other and will fail to preserve the subtleties of human requests, leading to a lot of inefficiencies and bad behavior. to solve these problems, there needs to be a fair bit of innovation, either on model training end or on harness end, to mitigate these issues. so what were the innovations in prime agent? they set RLM_MAX_DEPTH = 1? ... ok what? so theres no innovations, it's just calling subagents like every other harness in existence. if RLM MAX DEPTH is 1, it's not even RLM? it's just... a harness? so now i try to look up to what depth the ARC AGI 3 hillclimbing was done at. not surprisingly this doesnt seem to be disclosed on the blogs / posts. if it's just 1 layer deep, that's just shady marketing. a real innovation or contribution to harness engineering would be to detail the things that were improved to get a max depth of >3 to work without blowing up your computer, wasting a bajillion tokens, taking forever, or having agents go off the rails. i didnt find any 97% is just overfitting after hillclimbing on a public eval ok folks. it's nice that opus5 does better than sol than terra, but at the end of the day, we're looking at a public eval in which problems and solutions are open access, so any agent harness can arbitrarily overfit to the task set however much it wants to. if you do a codex or claude loop with some semi shady prompting to just get 100%, im pretty sure it will too. now obviously this is almost certainly not what the prime intellect team instructed their agents to do (otherwise you'd see that terra get 100% after a couple iterations), let's just all remember that there is no train / test split, and that ARC explicitly says - Public-set scores are vulnerable to task-specific overfitting. - They are “emphatically not” valid evidence of progress toward AGI. - Real generalization should be tested on the 55 semi-private or 55 fully private games. It's difficult to assess the degree of overfitting in Prime Agent's self-improving harness (which, btw, is not terribly different from nous agent or really any self-improving system that just looks at old JSONL agent trajectories and suggests skills / memories / system prompt appendments / tools / extensions). the more overfitting, the less impressive the 97% performance is. a counter-example that i found to be legitimately impressive was github.com/alexisfox7/PRO-LO… -- minimal / almost no overfitting, stupidly simple and general solution, and 95% on ARC AGI 3. i think the fairest thing to say about this system, given it's independent daemon system for managing subagent lifecycles, is that Prime Agent contributes a fairly robust persistent asynchronous agent-process tree. It does not demonstrate a solution to scalable deep recursive agency