Alex

🚨 New Paper Alert 🚨 We’ve just released "Hyperbolic Brain Representations" on arXiv. This research explores how hyperbolic geometry can transform artificial neural networks (ANNs) by better aligning them with the human brain’s structure. 🧠 arxiv.org/abs/2409.12990 🧵 (1/6)

Tencent has killed fine-tuning and RL with a $18 budget. Right now, if you want an AI agent to become an expert at a specific, complex real-world task, you have to use Reinforcement Learning. You let it try, fail, and update its internal parameters over and over again. This is the exact optimization technique (GRPO) that DeepSeek used to build their massive reasoning models. But there is a massive problem. Updating model weights is insanely expensive. It requires massive GPU clusters. And worst of all, when you train a model to be highly specialized at one thing, it often "overfits" and forgets how to be good at everything else. Tencent killed this bottleneck forever.. by building Training-Free GRPO. Instead of spending thousands of dollars to permanently alter the AI's brain, they asked a simple question: What if we just distill the experience of learning, and inject it as a memory? Here is how it works. They run the AI through the exact same trial-and-error process. But instead of updating the weights, they extract the "semantic advantage"—the actual logic of why one answer was better than another. They compress this winning logic into a "token prior”, a tiny package of high-quality experiential knowledge. Then, they just attach that knowledge directly into the API call. The results are staggering. Tested on DeepSeek-V3, this method required only a few dozen training samples to turn the AI into a specialized expert in complex math and web searching. It didn't just compete with models that were actually fine-tuned. It outperformed them. Zero parameter updates. Zero expensive training runs. Zero base-model amnesia.

beware of putting too much stock in your heroes. greatness is a transitory phenomenon. it is never consistent the gods briefly act through Men and then leave them to their ordinary fate

Imagine every pixel on your screen, streamed live directly from a model. No HTML, no layout engine, no code. Just exactly what you want to see. @eddiejiao_obj, @drewocarr and I built a prototype to see how this could actually work, and set out to make it real. We're calling it Flipbook. (1/5)

Scientists often make breakthroughs by synthesizing ideas across papers. In our new paper, we ask whether a language model can anticipate this process: given two parent papers, can it generate the core insight of a future paper built on them? 🧵⬇️

I am glad to see this. Recursive Transformer generalizes pretty well to other domains. MoR provides further free benefits. link: arxiv.org/pdf/2507.10524

Hierarchical planning unlocks long-horizon, non-greedy behavior in JEPA world models. Paper: arxiv.org/pdf/2604.03208 Website: kevinghst.github.io/HWM/ Code: github.com/kevinghst/HWM_…

Can we turn part of an LLM's weights into long-term memory that continuously absorbs new knowledge? We took a small step toward this with In-Place Test-Time Training (In-Place TTT) — accepted as an Oral at ICLR 2026 🎉 The key idea: no new modules, optional pretraining. We repurpose the final projection matrix in every MLP block as fast weights. With an NTP-aligned objective and efficient chunk-wise updates, the model adapts on the fly — complementing attention rather than replacing it. 📄 Paper: arxiv.org/abs/2604.06169 with amazing @Guhao_Feng @Roger98079446 Kai @GeZhang86038849 Di @HuangRubio

Researchers invented a fake eye disease called "bixonimania" with papers thanking "Professor Sideshow Bob" and crediting "Starfleet Academy." Every major LLM told users it was real. It got cited in an actual peer-reviewed journal. Source: nature.com/articles/d4158…

Reuters: "DeepSeek's efficiency gains are impressive" — yeah and a cheeky bit of training piracy. Anthropic caught them running 16M queries through Claude via 24,000 fake accounts to distill its capabilities. Industrial-scale copying. anthropic.com/news/detecting…

We found a task where LLMs struggle massively! Give them a coding problem in Python and they'd work great. Give the same problem in brainfuck and zero-shot their performance is ~0% +[--------->+<]>+.++[--->++<]>+.

Computer science will stop being taught as a useful craft; it has become an explanation of how the world works, like physics, biology and chemistry. This change is long overdue.

Very soon, all software will be free since it'll be easier/better to build in house. I pulled two all nighters this past weekend recreating *all* of our retool dashboards on Herald, our homegrown + more flexible alternative. Few realize how much leverage AI gives them.

We underestimate how much "abstract" thought is just repurposed sensorimotor control circuitry. A lot of reasoning is essentially about moving through idea-space the way we move through physical space.

🚨 𝗕𝗥𝗘𝗔𝗞𝗜𝗡𝗚: Arsene's Wenger new offside rule has been APPROVED and will be used in the Canadian Premier League! A player will ONLY be offside if he has FULLY passed the last defender, so it won't be decided based on a body part anymore. If succesful, the offside rule could change across the entire world from the 2027/28 season on.

Software engineering changed more in the last 3 months than the preceeding 30 years. Everything about running a software company needs to be rethought from first principles.

He's not just defending AI energy use. He is smuggling in a whole anthropology where humans are basically inefficient meat computers that you have to pour food and years into before they become useful. And once you accept that, the next move is obvious. If people are just costly biological training runs, then burning mountains of electricity to build synthetic intelligence starts to feel not only equal, but superior, even if it negatively impacts actual humans. That is the dystopian. It makes human development sound like a bug in the system, and it makes sacrificing human and creational flourishing for more computational power sound logical. To him, the grid gets strained, prices go up, ecosystems get hit, but hey, humans eat too, so what's the difference? The difference is that humans aren't an inefficient line item. They're the point. If your worldview can look at a child growing into an adult and describe it as energy spent to train intelligence, you haven't said something profound. You've revealed a horrifically rotten worldview.

Claude can now run your app, look at it, review the code, fix the bugs, and merge the PR a full engineering team that never sleeps, never argues about tabs vs spaces, and costs $20/month we just witnessed a mass layoff announcement disguised as a product update

The math on this project should mass-humble every AI lab on the planet. 1 cubic millimeter. One-millionth of a human brain. Harvard and Google spent 10 years mapping it. The imaging alone took 326 days. They sliced the tissue into 5,000 wafers each 30 nanometers thick, ran them through a $6 million electron microscope, then needed Google’s ML models to stitch the 3D reconstruction because no human team could process the output. The result: 57,000 cells, 150 million synapses, 230 millimeters of blood vessels, compressed into 1.4 petabytes of raw data. For context, 1.4 petabytes is roughly 1.4 million gigabytes. From a speck smaller than a grain of rice. Now scale that. The full human brain is one million times larger. Mapping the whole thing at this resolution would produce approximately 1.4 zettabytes of data. That’s roughly equal to all the data generated on Earth in a single year. The storage alone would cost an estimated $50 billion and require a 140-acre data center, which would make it the largest on the planet. And they found things textbooks don’t contain. One neuron had over 5,000 connection points. Some axons had coiled themselves into tight whorls for completely unknown reasons. Pairs of cell clusters grew in mirror images of each other. Jeff Lichtman, the Harvard lead, said there’s “a chasm between what we already know and what we need to know.” This is why the next step isn’t a human brain. It’s a mouse hippocampus, 10 cubic millimeters, over the next five years. Because even a mouse brain is 1,000x larger than what they just mapped, and the full mouse connectome is the proof of concept before anyone attempts the human one. We’re building AI systems that loosely mimic neural networks while still unable to fully read the wiring diagram of a single cubic millimeter of the thing we’re trying to imitate. The original is 1.4 petabytes per millionth of its volume. Every AI model on Earth fits in a fraction of that. The brain runs on 20 watts and fits in your skull. The data center required to merely describe one-millionth of it would span 140 acres.

.@usegrade is building payroll for performance, helping companies pay their workers based on results. They started with paying creators based on views and in the last 30 days have paid out $380k, up 120% month over month. Congrats @lottsnomad & @jvheaney on the launch! ycombinator.com/launches/PSG-g…