Use agents to review lidar for possible ancient sites
A Google engineer had 48 agents inspect NASA lidar and says they identified one unregistered Maya hilltop site in under three days.
- What they did
- They first worked out the approach with several chatbots, then opened a token on NASA’s public lidar portal. Next, they turned about 1,200 km² of Maya-area lidar strips into images and sent 48 agents through them, using a written field guide instead of a model classifier. The agents also cross-checked publications and archives, prepared a Spanish write-up, and looked back at the raw point cloud before the material was forwarded to INAH.
- How it went
- The review surfaced one hilltop place that seemed to include a plaza, an oval pyramid, and a possible reservoir, and INAH’s registry search did not turn up a listed site there. The author stressed that the finding is not official, and any on-the-ground work still depends on INAH approval. The exact coordinates were placed in a confidential annex.
- Tips
- Use agents to generate candidate leads, but keep a human or institutional review step for anything archaeological or otherwise sensitive. If you need to share a location, hold back precise coordinates until the proper authority has cleared the next step.
Try it yourself with Claude Code
I want to examine public [lidar dataset] for possible [archaeological features] in [region]. Split the coverage into tiles, run several agents in parallel to inspect the tiles, and have them note shapes and terrain patterns that might indicate built structures. Then check the strongest candidates against [public papers, archives, or site registries], and ask the agents to review the raw point cloud for those spots. Before you send, publish, delete, or submit anything, stop and ask for my approval, especially for any coordinates or report text. Return a ranked list of the top [number] candidates with a brief confidence note for each, and clearly label them as leads that still need expert verification.
Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.
Get 3 agent uses like this every morning
With the day's top AI news. Free, and one click to unsubscribe.
Related agent uses
Set up scheduled research on a stock watchlist
They used scheduled agent runs to scan watchlists for fresh company news, events, and insider activity.
Use parallel agents to run a research workflow
Sakana’s Agentic Scientist coordinates several agents at once to carry a project from idea to write-up.
Use a research agent for systematic literature reviews
A Reddit thread points to tools that can help with systematic reviews, paper notes, and keeping a project organized over time.
Build a dashboard from a semantic data layer
Use a coding agent with Graphene to define metrics, query your warehouse, and publish a report with charts and screenshots.
Turn Xiaohongshu searches into reusable skills
An agent browsed Xiaohongshu notes, compared the most-liked results, and saved the shared patterns as a knowledge-base skill.
Check how long open roles have been live at target companies
An agent sampled jobs on five company boards and measured how long each posting had been open.