Research automation
Agents and pipelines that run parts of the research workflow, from data pull to replication.
Agent-first coding: use the frontier tools and pay for them
Claude Code or OpenAI Codex, desktop app or command line. Generic and budget setups hide what the new tools can do.
Make your data pull something an agent can run
Automation starts with the stage that has the clearest pass or fail. Turn the data pull into one command with checks and a manifest.
What you may do with an AI agent on licensed data
Most data licenses forbid putting the content into AI systems. Here is what that covers, what it leaves open, and how to keep working without risking your campus's access.
API, MCP, CLI or computer use: how to connect an AI agent to your research tools
Use the lowest-level, most structured interface the job allows. What that means for data pulls, vendor platforms and software with no API.
Fully automated research pipelines: what exists and how to point one at finance
Machine learning, mathematics, biomedicine and finance each have a system that runs most of the research loop. Two of them you can install today.
Give your agent a memory: the instruction file and your first skill
Two plain text files turn an agent from a stranger into a research assistant who knows the project. How to write both, for Claude Code and OpenAI Codex.
Automate SEC filings research with a coding agent, from EDGAR to a finished panel
The SEC's data service needs no key. Here are the endpoints, the instruction file, a skill and a scheduled run, plus linking to CRSP and Compustat through the wrds library.
Token management for researchers, and why to buy usage before you hire help
What a token costs, the three settings that cut the bill, and the case for spending on usage ahead of spending on support staff.
Academic research is far behind the AI frontier, in every field we checked
A paper about a model is out of date before it clears review. In mathematics, the newest results come from a model no academic can run.
AI has arrived for finance research, and data vendors cannot police it
Agents now operate a computer better than the human baseline on the standard test. Licenses forbid most AI use of data, and almost none of it can be detected.
If the platform forbids agents, build the dataset yourself from the source
Filings, financial statements and earnings calls were public before any vendor packaged them. An agent can go back to the source, and the tools to do it are free.
Researchers are using AI against the rules, because being left behind is worse
At a major 2026 conference, more than one in five reviewers who were told not to use an LLM used one anyway. The fear driving that is rational.
OpenAI releases machine-produced math results with Lean proofs and compute figures
The results come from an internal model. OpenAI says the average result used compute equal to roughly three hours of ChatGPT Pro thinking.
A benchmark of nine factor-mining methods finds no approach consistently wins
FactorBench compares roughly five thousand machine-mined factors across five equity markets, from genetic programming to LLM agents.
Self-evolving research agents show no consistent gain from accumulated skills in a factor test
Across 18 long-horizon alpha-research runs and 48 continuation branches, evolved capabilities did not reliably beat the starting set.
LLM literature reviews hold up for broad claims and break down for paper-level ones
A test on economics papers that use rainfall as an instrument finds accuracy falls as the reading task needs more context.
Agents can propose investment factors, but a frozen referee has to judge them
A working paper splits factor research in two and finds that a referee the agent cannot touch admits 5 to 11 times fewer false factors.
Scientists report saving nearly 7 hours a week with AI, and the bottleneck moves to verification
A study of 15 million Gemini interactions, over 2,600 specialized models and a survey of over 600 scientists maps how AI is used in research.