48 agents, ~1,200 km², no INAH site code: lidar blog is not a Maya registration

Miotti, writing as a Google software engineer and not an archaeologist, describes a 72-hour sequence. Agents pulled public lidar strips, rendered bare-earth images, and looked for plazas, platforms, and pyramids. He reports one cliff-edge hilltop with straight edges, a leveled plaza, an oval mound he estimates at 8–12 m, a courtyard, and a possible reservoir. He says agents drafted a Spanish note to INAH and that a government reply arrived inside 72 hours. The post does not quote that reply, publish a site code, or show a field check.

The method is picture inspection by general chat models, not a trained detector. No precision, recall, agreement score, or negative-control set is reported. The post dates the bare-earth layer to 2013 and attributes it to NASA, but publishes no product or flight ID. NASA’s public airborne archive for this class of data is G-LiHT. Published Maya lidar set a higher bar: Canuto et al., Science 361 (2018), mapped 61,480 structures across 2,144 km² and ground-checked a sample. Chase et al., Journal of Archaeological Science (2011), did the same at Caracol. Karst, milpas, and modern clearings also leave rectilinear signatures.

Archaeological monuments in Mexico are federal property under the 1972 Ley Federal sobre Monumentos y Zonas Arqueológicos, Artísticos e Históricos. INAH registers sites; a private blog does not. Airborne geometry is a hypothesis until survey. Releasing enough spatial detail to walk to an unregistered cliff-edge mound also raises looting risk, which is why professional papers often withhold coordinates. Employment at Google is not evidence. The post cites no Google research release, model card, or benchmark.

The operational test is an INAH technical note or registry entry, plus a pedestrian survey, with public coordinates withheld. Until that document exists, the output is an unverified visual flag on a 2013 raster. A replicable result would name the lidar product, tile list, prompt, agent count, and false-positive rate against already-registered sites in the same strips, and would quote INAH’s written disposition. Absent those artifacts, 72 hours measures drafting speed, not archaeological yield.