AI for Economists — curated by Jesse LastunenNew resources in the curated AI & LLM library for economists.https://policy.fi/ai-econ/2026-07-23T00:00:00ZJesse Lastunenhttps://policy.fiMeasuring AI Innovation with Trademark Data (arXiv)https://arxiv.org/abs/2607.187952026-07-01T00:00:00ZC. Castaldi, F. Castellacci, A. Fronzetti Colladon, L. Segneri, F. VenturiniResearch note proposing AI trademark data as a timely, cross-sector complement to patents for measuring the development and diffusion of AI.AI Premium (NBER)https://www.nber.org/papers/w354512026-07-01T00:00:00ZNicola Borri, Aleh Tsyvinski, Yukun LiuNBER paper mining 380 trillion AI tokens across 400+ LLMs to build an AI Factor and document a large, tradable AI Premium in stock returns.Grand-to-Ground (G2G): An Adversarial Research Loop for AI-Assisted Social Sciencehttps://github.com/HongtaoMa1/grand-to-ground2026-07-01T00:00:00ZHongtao MaOpen-source Claude plugin running an 8-phase, Lakatos-inspired adversarial research loop with critic agents for economics and finance research.empirical_agent: Sandboxing and Scoring Agentic LLMs on Econometric Replicationhttps://github.com/SimonFirestone1/empirical_agent2026-07-01T00:00:00ZSimon FirestoneOpen framework to sandbox and score agentic LLMs (Claude, Codex) on econometric replication tasks, with Docker isolation and benchmark grading.AI in Economics Research: From Frontier Ideas to a Working Agentic Pipeline (wiiw seminar)https://lusiki.github.io/AI_workshop_wiiw/2026-07-01T00:00:00ZLuka Šikićwiiw seminar with slides, handout, and a cloneable repo reproducing an agentic event-study pipeline (R/Quarto/targets) with plan and provenance gates."We Must Act Now": Sixteen Nobel Laureates Join Leading Economists and AI Researchers in Call to Prepare for AI's Economic Transformationhttps://digitaleconomy.stanford.edu/news/wemustactnow/2026-07-01T00:00:00ZJesse Lastunen (curator)Statement by 16 Nobel laureates and 200+ economists and AI researchers urging urgent action so AI's economic transformation benefits society broadly.Where the Returns to AI Gohttps://www.economicforces.xyz/p/where-the-returns-to-ai-go2026-07-01T00:00:00ZBrian AlbrechtPrice-theory take on who actually captures AI's economic returns — model labs, workers, or owners of scarce complementary inputs.AI and development economics: Early evidence and how to keep uphttps://voxdev.org/topic/ai-and-development-economics-early-evidence-and-how-keep2026-07-01T00:00:00ZOliver HanneyVoxDev editor Oliver Hanney's living bank of the best sources on AI and development economics — evidence, tools, and how to keep up.The Microeconomics of Artificial Intelligencehttps://direct.mit.edu/books/oa-monograph/6067/The-Microeconomics-of-Artificial-Intelligence2025-01-01T00:00:00ZJoshua GansJoshua Gans's open-access MIT Press book analysing AI as prediction through microeconomics — demand, supply, pricing, market structure, and policy.AI Research Feedback: Claude Code Skills for Peer Reviewhttps://github.com/claesbackman/AI-research-feedback2026-07-01T00:00:00ZClaes BäckmanOpen-source Claude Code skills for academic peer review: 8-agent paper review, reproducibility audits, pre-analysis plan and grant proposal reviews.Integration and Collaboration in AI Research Workhttps://paulgp.substack.com/p/integration-and-collaboration-in2026-07-01T00:00:00ZPaul Goldsmith-PinkhamCapstone of Goldsmith-Pinkham's Claude Code series: a git/GitHub workflow for verifying AI-generated research and managing "verification debt."The Economic Benefits and Costs of AI and Policies to Mitigate AI's Impact on Inequality (arXiv)https://arxiv.org/abs/2607.011012026-07-01T00:00:00ZMatthew O. Jackson, Zafer KanikHow productive AI reshapes wages and inequality, and which tax/regulatory policies yield Pareto improvements under competitive vs. monopolistic AI.A Practitioner's Guide to Using Large Language Models and Generative AI in Economic History (NBER WP 35374)https://www.nber.org/papers/w353742026-06-01T00:00:00ZAndreas FerraraAndreas Ferrara's step-by-step workflow for using LLMs in research: model selection, prompting, cost control, and validating LLM-generated measures.academic-humanizer: Strip AI-Writing Tells from Papers and Grant Proposalshttps://github.com/AIScientists-Dev/academic-humanizer2026-06-01T00:00:00ZJesse Lastunen (curator)A Claude Code/Codex skill that removes AI-writing tells from papers and NSF/NIH proposals while keeping scholarly voice, citations, and numbers intact.Scholarahttps://www.scholara.ai/2026-06-01T00:00:00ZJesse Lastunen (curator)AI systematic-review tool: PICO/MeSH search building, triple-agent abstract screening, PDF data extraction, forest plots, PRISMA-ready audit trails.ChatDiDhttps://chatgpt.com/g/g-6a20206ca0508191bcbb14e28b35a5b6-chatdid2026-06-01T00:00:00ZClément de ChaisemartinA custom GPT by Clément de Chaisemartin for difference-in-differences—pick the right estimator and use the de Chaisemartin–D'Haultfoeuille packages.AI Economist Agent: An Agentic Framework for Model-Grounded Economic Analysis (arXiv)https://arxiv.org/abs/2606.200412026-06-01T00:00:00ZMasahiro KatoA RAG-based AI economist agent that grounds LLM narratives in economic theory and data, shown on U.S. inflation/Fed policy and bank stress-test reports.Using Claude Code for academic workhttps://psantanna.com/claude-code-my-workflow/2024-01-01T00:00:00ZPedro H. C. Sant'AnnaAn end-to-end Claude Code workflow for academic research: 18 specialized agents, an orchestrator pattern, simulated peer review, and 52+ slash commands.Theorist Toolbox: Tools for Agent-Based LLM-Assisted Economic Theory Research (arXiv)https://arxiv.org/abs/2606.223372026-06-01T00:00:00ZMoran KorenA verification-first protocol for LLM-assisted economic theory, using an adversarial prover-verifier loop—external checking, not model capability, is key.Writing Code vs. Shipping Code: Productivity Effects Across Generations of AI Coding Tools (SSRN)https://papers.ssrn.com/sol3/papers.cfm?abstract_id=68431182026-06-01T00:00:00ZMert Demirer, Leon Musolff, Liyuan YangMIT Sloan event-study of 100k+ GitHub devs: AI coding agents lift commits 180% but only 30% of that reaches actual releases—human bottlenecks bind.