> ## Content Index > Fetch the complete content index at: https://www.urgentmatter.press/llms.txt > Use this file to discover other available public pages before exploring further. # Geolava CEO says its AI model could forecast failures at historic sites - URL: https://www.urgentmatter.press/geolava-ceo-says-its-ai-model-could-forecast-failures-at-historic-sites/ - Published: 2026-11-07T15:22:55.000Z - Updated: 2026-11-07T15:22:55.000Z - Description: Museums and historic sites could use AI to forecast decay and structural failures before they happen. - Author: Adam Schrader - Tags: Science and Technology, #public-access Hantz Févry, chief executive of the San Francisco artificial intelligence company Geolava, told Urgent Matter on Friday that museums and historic sites could use the company's model to forecast decay and structural failures before they happen, though Geolava has not yet worked with either. "We tend to react more after something breaks and we're almost in reactive mode. We basically optimize how fast and how efficiently we react," Févry said. "But what if we could be more proactive?" Févry said Geolava trains its model on imagery of properties, sidewalks, roads and bridges, the assets it is focused on for now. Applying the model to historic sites would require testing against the materials, construction methods and conservation standards specific to those structures, such as the National Park Service’s [standards for treating historic properties](https://nps.gov/dscw/cr-treatment-of-historic-properties.htm?ref=urgentmatter.press). [Add Urgent Matter on Google](https://www.google.com/preferences/source?q=urgentmatter.press&ref=urgentmatter.press) "For historical sites, there's more at stake," Févry said. "If something collapses, there is more at stake than just making a repair because we want to keep that historical value. We don't want to get into that failure mode." Févry said some sites rely on regular inspections. "Regular inspections help but are not the final solution," he said. "If they were, we would never see problems." Févry said he had considered landmarks before Urgent Matter asked. One of his first companies built an earthquake early-warning system. He described a system called Microtremor, placed in buildings to test whether they would collapse in a quake. "For example, in the Caribbean in Haiti you have the Citadel on top of a big mountain. An earthquake can happen at any time in the north," Févry said. "What happens with that building?" ## Sign up for Urgent Matter This article is provided free with the support of paying subscribers. Not ready for a paid subscription? Sign up for a free account today. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime. The Citadel is part of the National History Park – Citadel, Sans Souci, Ramiers in northern Haiti, which UNESCO added to its World Heritage List in 1982. Urgent Matter used [satellite imagery](https://www.urgentmatter.press/satellite-imagery-shows-quake-destruction-around-venezuelas-coastal-landmarks/) in June to check landmarks on Venezuela's central coast after twin earthquakes there, and reported in March on a [drone survey](https://www.urgentmatter.press/drone-survey-launched-map-war-damage-syria-palmyra-ruins/) Syrian antiquities officials launched to map war damage at Palmyra. "Unfortunately, I haven't seen much demand from that industry to say, ‘Hey, we need to maintain, forecast, and avoid any disaster,’" Févry said. "But I feel like right now it might be changing." Févry also raised the Eiffel Tower. "It's tall, but is it safe? Should we be concerned or not?" he said. Asked whether Geolava had run its model on the tower, he said it had not but added, "that's something we should try." "You're giving me a good project," he said. ![](https://storage.ghost.io/c/3c/26/3c26cf7a-2384-4a8d-82e1-f44702177698/content/images/2026/11/Geolava---Team.png) Hantz Févry, the chief executive of Geolava, is pictured far left. Photo courtesy of Geolava He said the company could run the model on "key infrastructures" to identify risks and potential failures or confirm they are safe. "I'll ask my team to prioritize this. We can start looking at different sites," Févry said. Févry previously worked as a product manager at Google DeepMind. Geolava's blog lists a post announcing a $4.3 million funding round. The [company's website](https://www.geolava.com/platform?ref=urgentmatter.press) calls its model Terra "a world model for the built world." Févry said Geolava's biggest challenge has been that people assume all AI is a large language model predicting the next piece of text. "The world doesn't run on words, it runs on physics and time," Févry said. "So you need to have a different type of model." "When it comes to the physical world where we have the most important assets, it can be government infrastructures, roads, buildings where we live, offices, even historical sites. Those are completely offline," he said. "We're hyperfocused on the built world and we want to make sure that people anticipate failures even before it happens," Févry said. Terra starts with the current condition of a physical asset, applies an action such as heavier traffic, maintenance or no maintenance at all and forecasts the asset's future condition, Févry said. A sample forecast on Geolava's website shows a property with maintenance delayed for 24 months. The model predicts 8.4 millimeters of foundation displacement and facade cracking. "When it comes to failure for a property, it's not an isolation. It's a chain of events," Févry said. For example, a tree can damage a gutter, change water drainage and cause pooling that leads to another problem. Terra has been about 86% accurate, Févry said. The company sometimes tests the model by treating a past point as the present and making a forecast. He said that level of accuracy is "amazing to see the model do that type of forecasting into the future." Févry told Tom's Guide in an interview published October 25 that Geolava's flood-risk forecasts for the Florida Department of Transportation's roads were about 87% accurate. According to the company, it “recently signed a multi-million, multi-year contract" with FDOT. Government clients use the model to plan maintenance of roads and sidewalks and stretch their budgets, Févry said. Asset managers use it to monitor properties beyond standard condition assessments, he said. "We're not a point solution. We're an infrastructure," Févry said. He said clients build their own tools on top of Geolava. Some check the model daily to monitor a site, while others use it once or quarterly. Févry said a museum without technical staff could use Geolava's web app, where users type a question and get answers. More technical clients can use an API. Geolava can integrate with platforms clients already use. "Museums, historical sites, if you want to really monitor them, especially their condition, how they're decaying, what maintenance needs to be prioritized or any type of failure that we can anticipate, then yes, I think Geolava is the perfect tool for that," Févry said. Urgent Matter previously reported on [Octave](https://www.urgentmatter.press/after-louvre-break-in-lidar-surveillance-enters-the-museum-security-debate/), which markets lidar software that maps galleries in 3D and tracks intruders, and on [Art Guard](https://www.urgentmatter.press/art-security-firms-warns-thieves-daytime-gap-advantage/), which sells magnetic sensors to protect artworks directly. Asked if Terra could work with such security systems, Févry said some clients already feed video into the model to help it make predictions. "When I mean failure I don't mean only environmental ones," Févry said. He said failures can also come from people, either through direct action or maintenance that should have been done but wasn't. Geolava can connect to security systems and other safety systems, including those meant to keep sites compliant with the Americans with Disabilities Act, he said. "So the operator is no longer guessing and can always stay on top of everything to avoid a disaster," Févry said. Asked if artists and others wary of AI should worry about surveillance, Févry said AI speeds up scientific discovery and can work through problems too complex for a person to calculate. "I do understand sometimes there are some extremely powerful arguments to be worried,” he said. “What if the system goes rogue? What about surveillance?" But he said AI also democratizes intelligence and that it “is no longer in a silo for a small number of people where it's expensive and hard to access.” “I believe for any technology there are pros and cons,” he said. “It's a matter of how it's leveraged." Geolava operates only in North America for now and is talking with organizations outside the United States and Canada, Févry said. "Our system doesn't look at an asset in isolation," he said. Févry said the company wants to better account for how a neighboring building or road vibration affects an asset. "If you're building a road you can have a better understanding what is the best optimization because you can already see into the future what your decision would lead to," he said. *Urgent Matter is reader-supported. Please* [*consider subscribing*](https://www.urgentmatter.press/mta-first-amendment-museum-of-sex-bus-ads/#/portal) *to support independent arts journalism.*