Extracted system prompts from Anthropic – Claude Fable 5.1, Opus 5, Claude Design, Claude Code. OpenAI – ChatGPT GPT-6-Astra, Codex. Google – Gemini 3.8 Flash, 3.1 Pro, Antigravity. xAI – Grok, Grok Bot, Cursor, Kimi and more! Updated regularly.
Archives: Bookmarks
How an AI math breakthrough ignited a controversy | Science | AAAS
Mathematician Tristan Buckmaster of New York University and Levent Alpöge, a researcher affiliated with Anthropic, one of OpenAI’s major competitors, had made significant progress solving the problem over the past month using AI tools adapted from OpenAI and Anthropic. But before they could publicize their findings, OpenAI mounted its own colossal effort—pouring millions of dollars of computational resources into the problem. OpenAI says it didn’t knowingly tap into Buckmaster and Alpöge’s latest work, but Buckmaster has released public statements questioning that position.
Source: How an AI math breakthrough ignited a controversy | Science | AAAS
🔬 The Lab of the Future Should Feel Like a Data Center — Andy Beam & Rafa Gómez-Bombarelli, Lila Sciences
Their goal is a scientific superintelligence wired directly into the wet lab. They are all in on the bitter lesson, and the thesis follows from it: a lab is an infinite token generator. Produce data at scale, and the synergies give you a general reasoner that can tackle any scientific problem. They are committing hard. Biology, chemistry, drug discovery, and materials science, all at the same time. Time will tell if it works, but it is an exciting hypothesis.
Gallery of Concept Visualization
Source: Gallery of Concept Visualization
A/I Shuts Down Stay Human · Keep the Internet free
We cannot engage in a fight or expose ourselves to manipulations that threaten the lives of these people and their loved ones, leaving them at the mercy of autocrats, fascists and agencies that take extra-legal courses of action. The possibility that our work may cause legal and financial consequences to those who are close to us – or even only have something to do with us – leaves us no choice. A/I is shutting down. The Autistici/Inventati collective is shutting down and will soon discontinue all of the services we provide.
How Claude is accelerating protein design and analytical chemistry Anthropic
We ran Opus 4.8 and Mythos Preview in multi-target mode with 48 hours of wall time and up to 12,500 NVIDIA H100 hours of compute for running specialized protein design and folding models. We also ran Mythos Preview in single-target mode with 24 hours of wall time and up to 2,500 NVIDIA H100 hours of compute for each target.5
Source: How Claude is accelerating protein design and analytical chemistry Anthropic
Guide to Preferred Sources in Google Search for Web Publishers | Google Search Central | Documentation | Google for Developers
If you’re a website owner, you can help your audience find your publication as a preferred source in Google Search. When a user selects your site as a preferred source, your content is more likely to appear in “Top Stories”, highlighted with a “preferred” badge. In AI Mode and AI Overviews, your content can be highlighted with a “preferred” badge for users who have selected your site as a preferred source.
eviews-mcp — Drive EViews from Python and AI assistants
Drive EViews 13 from Python — or let an AI assistant drive it for you. A Model Context Protocol server and a Python library, sharing one engine. Every number comes from EViews itself.
Source: eviews-mcp — Drive EViews from Python and AI assistants
[2602.22631] TorchLean: Formalizing Neural Networks in Lean
Neural networks are increasingly deployed in scientific, safety critical, and mission critical pipelines, yet verification and analysis are often performed outside the programming environment that defines and runs the model. This creates a semantic gap between the executed network and the analyzed artifact: guarantees can depend on implicit conventions about operator semantics, tensor layouts, preprocessing, floating-point behavior, graph transformations, accelerated kernels, and external certificates. We present TorchLean, a unified framework for formalizing, executing, and verifying neural networks in Lean 4.
Source: [2602.22631] TorchLean: Formalizing Neural Networks in Lean
TorchPhysics – Deep Learning for PDEs
Welcome to TorchPhysics, an open-source PyTorch library that implements deep learning methods for partial differential equations, enabling engineers and applied mathematicians to use modern AI-based methods in their usual workflow.