Stephan Hoppe

A fundamental problem with extending Codex/Cowork/Code to all knowledge work is that they remain very "software-brained" where the end result (the software) is what is important & that code serves as a source of truth. For a lot of other knowledge work, the process is at least as important as the outcome. This includes researching what is known, an exploration of alternatives, failed efforts, prototype branches, experiments, etc. All of those things are valuable, so you cannot use the PowerPoint at the end the way you can use a codebase, nor is progress on a to-do list sufficient context post compaction. You work in learning loops, refining your perspectives as you go. In some ways, this makes long-running models like Fable hard to use for deep knowledge work, since they are designed to deliver product to you in the end. You can prompt your way around this problem, but everything about the Codex and Code harnesses want you to be a software developer and you have to fight them. There is a real disconnect between how a manager or analyst thinks about problems and how the agentic software tools approach solving them. Addressing this is critical to breaking out of the coding niche for these tools.

One of the most powerful uses of AI is not getting answers. It is learning how to ask better questions. Suppose I am reading a historical atlas showing the trade networks of Eurasia around 100 CE: the Roman Empire in the west, the Parthian Empire in the Middle East, the Kushan Empire in Central Asia, the Han Empire in China, and the trade routes connecting them across the Silk Roads and the Indian Ocean. Traditionally, I would look at the map, read the caption, and move on. In AI at Your Side: The Student's Guide to Smarter Learning (forthcoming, Oxford University Press), we argue for a different approach: treat AI as a reasoning partner. Move beyond description. Ask for mechanisms. Seek evidence. Generate a counter-thesis. Identify blind spots. Compare alternative explanations. Synthesize. Here is what that looks like in practice. [1/n]

A 4B local model can do physics: dynamic fracture, beam mechanics, topology optimization, and Chladni plates. But how do you get a 4B model to "do" physics? We can give it scientific skills that turn intent into experiments, generate data, make images and movies, verify outputs, and report the results. In this video, Gemma-4 E4B runs locally on my MacBook via mistral․rs with 8-bit in-situ quantization, plus the Agent Skills API, completely open-source. Each skill is a folder with a SKILL․md, teaching the model how intent, parameters, assumptions, examples, and solution strategy relate. The skill is uploaded, invoked by the model, executed in a sandboxed shell/Python environment, and it inspects the generated files before reporting the result. In the video, I show four skills: 1. Beam mechanics: 1D Euler-Bernoulli beams with displacement and rotation: cantilever, simply supported, fixed-fixed, point loads, distributed loads, moments, springs, reactions, shear/moment diagrams, max deflection, and vertical-equilibrium residuals. 2. Fracture mechanics: 2D triangular-lattice dynamics with Morse/Lennard-Jones potentials, sharp edge pre-cracks, Mode I/II loading, bond breaking, temperature and strain-rate effects, stress-strain curves, peak stress/strain, broken-bond counts, and fracture movies. 3. Hierarchical topology optimization: Minimum-compliance optimization with density/sensitivity filtering, penalization, target volume fraction, flexible supports and loads, optimized density fields, compliance values, and STL export (so you can quickly move to manufacturing). 4. Chladni plates: Analytic superposition of vibrating-plate eigenmodes on square, rectangular, and circular plates to generate standing-wave nodal patterns. The examples show that the model does not need to contain all of physics in its weights; instead the physics lives in the executable and editable skill. The model orchestrates and scales, and results compound. This is quite interesting for science: AI becomes more capable when it has physical agency and can act through scientific instruments - and this can translate from in silico simulation to hardware, manufacturing, and more.

Humboldt University in Berlin: where the home of German classical archaeology is now considered expendable. Two centuries of scholarship sacrificed to budget spreadsheets. Winckelmann would be thrilled.

people on here are dumb. the latest subquadratic attention trick might produce a model that *processes* 1M tokens (or 12M..) without going insane, but that doesn't make it good the real problem isn't the architecture, it's the data. humans haven't generated many contiguous linear spans of 1M tokens. so of course we can't learn this distribution. it doesn't exist

🤯 A Russian missile struck the Kyiv Pechersk Lavra Following the overnight attack on Kyiv, the roof of the Dormition Cathedral caught fire. The cathedral is the main church of the monastery, founded nearly a thousand years ago and regarded as one of the most revered sites in Orthodox Christianity. Ironically, Putin has repeatedly claimed that Orthodoxy and Russia are inseparable, saying that “with the adoption of Christianity and Orthodoxy, the Russian nation began to form as a single nation,” and that “Orthodoxy unites people of many different nationalities.” That did not stop a Russian missile from hitting one of the most important Orthodox shrines in Ukraine.

I just watched priests trying to save crucifixes from a burning monastery. This is what russia is destroying. The Kyiv Lavra is one of the holiest Orthodox sites on earth. A UNESCO World Heritage site. Moscow calls itself “Christian civilisation.” It’s literally barbarism.

First responders have put out the fires in Kyiv – at the Lavra and the Mystetskyi Arsenal. I am grateful to all the services working at the sites of the strikes, and to everyone who is helping. It has been confirmed that two Russian drones deliberately targeted the part of the city where the Lavra and the Mystetskyi Arsenal are located. As of now, 35 people are reported injured in Kyiv. Across the country, 53 people have been wounded, and 11 people are known to have been killed in this massive Russian attack. My condolences to all their families and loved ones. We are in constant communication with our partners regarding the consequences of this Russian attack and the necessary response. It is important that state leaders, public leaders, and international organizations do not remain silent. I have instructed Ukraine’s Ministry of Foreign Affairs and our entire diplomatic team to maximize all contacts with partners today so that international events this week and next deliver real results for strengthening our defense and increasing global pressure on Russia over this war. We need a just response to this Russian strike. Thank you to everyone who stands with Ukraine.