---
name: al-174-run-a-swarm-of-experiment-agents-on-one-hard-question
description: "Split a research or optimization question into many experiments, let agents run them in parallel for hours, and have them design follow-up experiments from the results. Third Moment Research runs 10,000 agents at once on 36 to 48 hour investigations with OpenAI's Codex, and one run cut a backtest from about three hours to a few minutes. Use when the user wants help with parallel experiments (research) for a small business."
---

# Run a swarm of experiment agents on one hard question

Split a research or optimization question into many experiments, let agents run them in parallel for hours, and have them design follow-up experiments from the results. Third Moment Research runs 10,000 agents at once on 36 to 48 hour investigations with OpenAI's Codex, and one run cut a backtest from about three hours to a few minutes.

## Inputs to collect first
Ask the user for anything below that they haven't already given you. Don't guess missing facts.

- E.g. why is our backtest slow / which pricing page converts best
- Number of parallel runs, time limit
- Your own material (text, notes or data)

## Instructions

Do the task below with the user's inputs in place of each [bracket].

You are the lead researcher coordinating a batch of experiment agents. Question: [e.g. why is our backtest slow / which pricing page converts best]. Data and code you may use: [describe]. Budget: [number of parallel runs, time limit]. Step 1: list 5-10 hypotheses, each with a single experiment that could confirm or kill it and the metric to measure. Step 2: write a self-contained brief for each experiment agent (inputs, exact steps, what to report). Step 3: after I paste back results, rank the hypotheses, explain what the evidence shows, and propose the next round of follow-up experiments. Keep a running log of what was tried and ruled out so no agent repeats work.

## Safety checks (share these with the user)

- Check every figure, price and total against your own records before anything is sent, posted or booked. A person approves refunds, discounts, payments and price changes.
- Read and edit the draft before it reaches a customer or goes public. If you later automate this prompt, start with "draft and notify me", not "send automatically".
- Check names, dates, numbers and any claim against your own source material before you use the answer. AI can misread or invent details.
- Say clearly which parts need a person to review or approve before anything is sent, paid, posted or filed.

## About this skill

- Works with: Codex, ChatGPT, Claude Code
- Based on this real-world example: https://openai.com/index/third-moment-research/
- Free prompt from AgentPlaybook Real Estate: https://agentic-library.netlify.app/prompt/al-174-run-a-swarm-of-experiment-agents-on-one-hard-question/
- To install in an agent that supports Agent Skills, save this file as SKILL.md in a folder with the same name as `name` above (for example in your skills directory). Install steps for Claude.ai, Claude Code, Cursor and Codex: https://agentic-library.netlify.app/skills/