Scan NASA Lidar data for unmapped Maya sites with parallel agents
A Google engineer pointed 48 AI agents at NASA Lidar data and reportedly found an unmapped Maya pyramid within 72 hours. ·
Done with a custom-built or unnamed agent
- What they did
- After brainstorming with several chatbots, the author created a token on NASA's open lidar portal and had agents convert about 1,200 km² of Maya-region lidar strips into images. 48 agents inspected them against a written guide, with no classifier, and flagged structures. Agents then checked the literature and archives, drafted a Spanish report, and verified the raw point cloud. The author sent the report to INAH.
- How it went
- One hilltop site with a plaza, an oval pyramid and a possible reservoir appeared unregistered, and INAH's records check found no registered site there. The author says it is not official, and fieldwork is up to INAH.
- Worth knowing
- The author is not an archaeologist and had to trust the agents' Spanish translation. Any fieldwork needs INAH authorisation, and exact coordinates went in a confidential annex.
Try it yourself with Claude Code
I want to scan public [Lidar dataset, e.g. NASA or USGS Lidar tiles for a region I name] for possible unmapped archaeological features such as [mounds, pyramids, or earthworks]. Split the area into tiles, process them in parallel, and flag candidates by their elevation patterns. Finish with a ranked list of the [10] most promising locations with coordinates, a hillshade image of each, and a note on how confident you are, and be clear these are leads that need expert verification.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Discussion on HN · Nov 11, 2026
Related use cases
Find connections across 100 non-fiction books
Claude Code browses a personal library with CLI tools and surfaces trails of related excerpts across books.
Search years of research files in an Obsidian vault
Claude Code, guided by a CLAUDE.md, searches research papers, scans and PDFs via Spotlight, extracts text, and files daily notes in dated folders.
Save and search links via a Telegram second brain
An n8n Telegram agent reads links sent to it, summarises and categorises them into Google Sheets, and answers questions from the library using Supabase semantic search.
Generate updated research reports on followed companies
The poster built and open-sourced a custom stock-research agent that produces updated reports on companies they follow.
Custom agent
Research US stablecoin adoption with an agent
A Manus agent connected to Heurist's mesh ran deep research on US stablecoin adoption in 2026 after one approval step, finishing in two minutes.
Research potential rental properties with deep research
An investor uses ChatGPT's deep research feature to produce a full report on a property before running exact numbers on a rental investment.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.