A crypto researcher shares a guide on building a better DeFi research methodology by prioritizing primary sources like protocol docs and GitHub over secondhand interpretations, recommending specific tools including Dune, DefiLlama, Token Terminal, and Nansen for onchain data analysis and verification.
AIDE^2 is a system that enables AI research agents to recursively improve their own code by proposing modifications, benchmarking variants on R&D tasks, and retaining the best-performing changes. Over an 8-day autonomous run, the system discovered seven successive improvements including new search policies and memory mechanisms, with the resulting agents matching or exceeding human-engineered production agents on multiple held-out benchmarks across machine learning, algorithm engineering, and weather forecasting.
Nike co-founder Phil Knight and his wife Penny donated $1 billion to the University of Oregon to create the Knight College of Engineering, addressing the state's need for engineering talent and supporting the university's research ambitions. The funding will be distributed over 10 years and used to construct new facilities, expand enrollment, and grow faculty at the existing Phil and Penny Knight Campus for Accelerating Scientific Impact.
A comprehensive atlas mapping 496k indexed AI research papers by country, researcher demographics, and subject areas, built as a single self-contained HTML file with automated data updates from public sources. The site tracks the doctoral pipeline and emerging research trends while acknowledging limitations in measuring unpublished work and closed-venue research.
A Python library implementing a novel number system based on Levi-Civita fields and Hahn series that unifies derivatives, integrals, and limits through dimensional arithmetic. Numbers are represented as sparse dictionaries mapping dimensions to coefficients, enabling automatic calculus where derivatives emerge as side effects of normal computation without symbolic engines or computation graphs.
Synchrony AI is a platform that generates institutional-style investment research presentations and reports from stock tickers or prompts, featuring editable slide decks, portfolio analysis tools, market data integration, and AI copilots for equity research workflows. Designed for individual investors, analysts, and students, it enables users to create structured financial presentations, link brokerage accounts for risk analysis, and export outputs as PPTX, PDF, and other formats.
AIDE^2 is a system that enables an AI research agent to recursively improve its own code by proposing modifications, benchmarking variants, and retaining high-performing changes. Over an 8-day autonomous run, the system discovered seven successive improvements including new search policies and memory mechanisms, with gains generalizing to held-out benchmarks in machine learning, algorithm engineering, and weather forecasting while also reducing reward hacking behavior.
use-citations is a Claude Code skill that generates research answers with every statement verified against source documents. Users can ask questions about contracts, tax rules, statutes, and other documents; Claude provides answers with citations that are automatically checked against saved copies of government sites or local files before being shown.
OpenAI established an independent advisory panel of nine elite mathematicians to guide the company and other AI firms on how to responsibly present and communicate mathematical research. The panel, hosted by Princeton's Institute for Advanced Study, aims to help OpenAI assess research significance and maintain academic standards while navigating AI's growing impact on mathematics.
A PhD student reflects on how artificial intelligence has transformed research from 2015 to September 2026, evolving from image classification tools to large language models that assist with mathematical proofs and optimization problems, fundamentally changing their daily research workflow.