💼 Work & career

Let AI agents dig through public data on a hobby question (and try to prove you wrong)

Two non-scientists just used teams of AI agents on free public data: one found a Maya ruin hidden under jungle in NASA lidar scans overnight, the other turned up two planet candidates that NASA's TESS team has now listed. You pick the question and the rules; the agents do the hours of looking, checking and writing up.

AdvancedUpdated · By BotsDoThat · Added Nov 11, 2026♥ 8,727 · ↻ 836 on X · measured Nov 19, 2026

Seen on X · Hacker News · Official docs

What the bot does

  1. Helps you turn a curious question into a clear plan: which free public data to use (NASA, government maps, archives) and what would count as a real find.
  2. Downloads the data and splits the big job across several agents that work in parallel, for example 48 agents each checking a batch of lidar images against the same written guide.
  3. Runs overnight and leaves you a report in the morning with its best leads and the evidence for each.
  4. Tries to knock its own finding down: hides part of the data and predicts it, checks for look-alikes and false alarms, and has a second, separate AI review the work.
  5. Searches papers, catalogs and official registers to see whether someone already found it, then drafts a short report for the right experts.

What you end up with

A shortlist of real leads from public data, each with the evidence and the checks it passed, plus a draft note for the experts who can confirm it.

Copy-paste prompt

Fill in the [brackets]. Then paste it into any tool below.

Copy this prompt
I'm curious whether [your question, e.g. there are unmapped old ruins in this region] using free public data from [source, e.g. NASA lidar / TESS]. First, write a plan: which data to download, what a real find would look like, and 5 tests that could prove a lead is NOT real. Then do the work in parallel batches and give me a morning report with your top 3 leads and the evidence for each. For every lead, search papers, catalogs and official registers to check whether it's already known, and tell me how sure you are. Don't contact anyone or submit anything without asking me.

New to this? Start with what an AI agent is, in simple terms, then how to build an AI agent without coding, or set up a bot in 5 plain steps →

Want one new job like this each week, with a prompt to copy? Get the free BotsDoThat Weekly email and the free Starter Pack: 7 ready-to-paste AI agent setups, plus a power-user bonus.

Setup

  1. Pick a question where good free data already exists (NASA's TESS star data, public lidar scans, census or government records).
  2. Use a coding agent that can download files and run code on its own, such as Claude Code or Codex, ideally on a spare computer or a cloud workspace.
  3. Write down the pass/fail rules before you look at results, and ask the agent to try to disprove its own finding, not confirm it.
  4. Open a fresh session (or a second AI) with read-only access to review the work and look for mistakes.
  5. Before you announce anything, have the agent check the literature and official registers, then send a short, careful report to the experts who own that field.

Tools that can do it today

Plans and limits change; check each tool's page.

Watch out

  • Agents can miss earlier work. The Maya site turned out to be already mapped in a 2025 paper; the agents missed it during their literature search, so it was a rediscovery (author's Nov 11 update).
  • Ask for ways it's wrong, not whether it's right. The planet hunter's lesson was to keep giving the agent tests that could disprove the signal, and an audit still caught one claim that looked stronger than it was.
  • A candidate is not a discovery. TOI 7930 is listed as 'planet candidate' until follow-up observations confirm it, and the TESS team flags it as a possible multi-planet system around a variable star.
  • Archaeological sites are protected by law in many countries. Report finds to the authority (in Mexico, INAH) and keep exact locations private.
  • Long agent runs use a lot of compute. Set a spending limit on your AI account before leaving it running overnight.

What this costs to run

In a chat app it runs inside your plan. If you run it as a bot on an AI API, the bill depends on how often it runs and how much it reads.

Estimate it in the cost calculator →

Tip: Use a small AI model for simple jobs, a big one for hard ones. More in Run your bots for less and how much an AI agent costs.

Questions people ask

Which AI tools can do this?

Claude, ChatGPT or Gemini can do this today. Paste the copy-paste prompt on this page into any of them and fill in the [brackets].

How hard is it to set up?

It's rated Advanced. Start with the steps under “What the bot does”.

What should I watch out for?

Agents can miss earlier work. The Maya site turned out to be already mapped in a 2025 paper; the agents missed it during their literature search, so it was a rediscovery (author's Nov 11 update).

Sources

Last checked Nov 12, 2026. Self-reported results from posts are the poster's own claims.

BotsDoThat Weekly

Get the free Starter Pack: 7 AI agents you can put on a schedule

Join the BotsDoThat Weekly and we'll send you 7 ready-to-paste setups for ChatGPT, Claude, OpenClaw and Hermes Agent, plus a bonus for power users. Each one asks you before it sends, buys or deletes anything.

We'll email you a confirm button. Click it to confirm and get your Starter Pack. Unsubscribe any time. Privacy

Know someone who'd like this? Forward it, or send them to botsdothat.com/newsletter