AI for Science

TL;DR for operators A long-running research agent can produce an impressive manuscript while still leaving a manager unable to reconstruct what evidence was gathered, what failed, which claims were checked, or where a human should intervene. pAI/MSc1 is most useful as a response to that problem: although its fixed workflow uses 23 specialist agents across 30 graph nodes, its more consequential design choice is to preserve discovery, planning, theory, experimentation, synthesis, review, checkpoints, and budget accounting as named artifacts that can be inspected, resumed, audited, and structurally validated. ...

TL;DR for operators European XFEL did not find that the main obstacle to scientific AI assistance was simply insufficient documentation. Scientists and support staff had to combine project objectives, facility knowledge, instrument procedures, specialized software, computing constraints, and expertise distributed across people and documents. The resulting study1 points toward a different implementation model. An effective scientific agent needs grounded retrieval, but it also needs access to the actual execution environment, mechanisms for testing generated code, visible sources and plans, approval gates for sensitive actions, and components that can be replaced as models and tools change. ...

TL;DR for operators A technically sophisticated AI answer is still only a candidate. Coherent explanations, detailed equations, and confident conclusions cannot substitute for an independent check of whether the reasoning is valid. Kol and colleagues report such a check for a layered quantum optimization method called QAOA. Lean 4’s kernel verifies that, for an even ring with $2p+2 \le n$, the optimal approximation ratio at depth $p$ is exactly ...

TL;DR for operators Nguyen’s paper is not another “AI writes code” victory lap. It is more useful than that. It documents a 12-work-day, 57-session case in which a physicist supervised Claude Code, using Sonnet and Opus models, to build clax-pt, a JAX implementation of a differentiable one-loop perturbation theory module validated against the established C reference code class-pt.1 ...

Brainstorming Is Cheap; Research Judgment Is Not Brainstorming with an LLM is easy. Ask for ten research ideas, wait a few seconds, and receive a confident menu of things that sound just plausible enough to be dangerous. Turn up the temperature and the machine becomes “creative.” Wonderful. We have successfully automated the whiteboard intern. ...

Lab work has an old rule: never trust the first beautiful result. It may be correct. It may also be a measurement artifact wearing a lab coat. That rule becomes more important when the “research assistant” is an LLM that can write code, invent tests, explain errors, and occasionally hallucinate with the confidence of a junior consultant who has just discovered PowerPoint. The paper “AI-for-Science Low-code Platform with Bayesian Adversarial Multi-Agent Framework” takes this problem seriously.1 Its central claim is not that scientific automation needs a larger model, a longer prompt, or another cheerful agent named “Planner.” The claim is sharper: in AI-assisted scientific coding, both the generated code and the generated tests are uncertain. If the validator is also an LLM, then the system has not solved hallucination. It has merely hired hallucination as compliance staff. ...

A lab workflow is a terrible place to discover that your AI agent has been “remembering” chemistry as a conversation. That sounds unkind. It is also the point. In a casual chatbot, losing track of context means an awkward answer. In computational chemistry, losing track of context can mean a wrong molecular geometry, a missing imaginary-frequency check, an invalid charge or multiplicity, or a pKa estimate that looks numerically confident while being scientifically useless. The model did not necessarily become stupid. The workflow around it treated state as text. ...

A lab does not fail because the scientist forgot how to think. It fails more often for duller reasons: the data table is in the wrong format, the simulation script only works on one cluster, the instrument queue is opaque, the boundary condition was changed but not logged, the literature trail cannot be reconstructed, and the “promising result” lives in someone’s notebook like a small hostage. ...

A simulation job does not care how elegant the prompt was. It cares whether the input files are valid, whether the parameters are compatible, whether the previous step produced the right intermediate state, whether the solver converged, and whether the final number actually means what the workflow says it means. This is where the romance of “AI scientists” usually meets the concrete wall of scientific computing. The model can sound like a postdoc. The machine still wants the correct INCAR tag. ...

Packing a Punch: How Model-Based AI Outperformed Decades of Sphere-Packing Theory Expensive experiments have a nasty habit: they punish enthusiasm. In many AI success stories, the hidden luxury is cheap feedback. Generate a million candidates, test them quickly, keep the survivors, call it discovery. This is the comfortable world of many coding benchmarks, puzzle solvers, and evolutionary search systems. It is not the world of high-precision semidefinite programming, where evaluating one candidate can take days and where a bad guess does not merely waste a GPU minute; it quietly burns a serious slice of the research budget. ...