Tilebox announced an update today aiming to make AI agents more effective geospatial-data analysts.
AI is supposed to make EO data much more accessible to non-technical analysts, but a problem persists: AI can often feel like a black box—submit your prompt, get an answer, and don’t worry about how we got there.
Tilebox’s update aims to solve this issue, by giving human analysts in-depth understanding of how a given AI workflow arrived at its results.
“Agents need context. As a human, if you get an answer from an agent, you want to know where that answer comes from, which data was used, which steps happened,” Tilebox cofounder Laura Costa told Payload. “You want to be able to be sure that the results that you’re given are certain, and you can trace it.”
Tilebox 2.0: Tilebox was founded in 2022 to make large quantities of disconnected satellite data easier to understand, but today’s geospatial intelligence analysts are increasingly artificial.
As a result, Tilebox’s update is focused on helping analysts use AI more effectively.
It’s not another chatbot. It’s a system that sits between the AI and large datasets, and teaches AI—including Anthropic’s Claude Code, OpenAI’s Codex, Amp, and others—how to best work with their human counterparts.
The update arms AI tools with skills to work more effectively with Tilebox’s APIs. It allows AI agents to continue using multiple data sources, but to do so while providing proof of how it arrived at certain answers.
“It’s a way for people to bring their own agents…and hook them into a system such that this agent can then develop…complex intricate workflows,” cofounder Stefan Amberger told Payload. “With the correct guidelines—and with the same context that a human would have, so it doesn’t have to make assumptions—it’ll get more complex workflows right.”
The bottom line: The update is the result of two major shifting forces in the geospatial intelligence market: the influx of AI, and buyers increasingly relying on multiple, disconnected sources of data. AI is good at analyzing a lot of data quickly, but with Tilebox those answers can be fact-checked and reliable.
With the new agent-support tool, Tilebox wants to help developers build applications where AI can use multiple, fragmented data sources to surface more advanced insights, show its work, and not hallucinate along the way.
“We are not for teams that just want to purchase an image. We are for teams that want an answer, a workflow, a product, anything that it’s built using this kind of data,” Costa said.
Symphony Space Unveils Adagio-XL ODC Sat
With Adagio-XL, Symphony Space is getting into the orbital data center game.
Orbital Raises $5M Pre-Seed to Build ODC Constellation
Orbital’s ODC concept instead relies on a network of relatively smaller satellites, which can together handle large AI inference workloads, and remain in constant contact with the ground.
Reflex Aerospace Taps Arkadia Space for 2027 Mission
Reflex Aerospace selected Spanish startup Arkadia Space to provide a propulsion system for its satellite mission, launching in 2027.
NewOrbit Raises $18.5M to Build VLEO Spacecraft at Scale
While NewOrbit has yet to fly in any orbit, the company is targeting a first flight of its VLEO-capable NEO-1 satellite in 2028.
SpaceX is set to go public on Friday at a $1.75T valuation.
Analysts are split over whether that’s the right number, and how the stock will perform in the short term—but everyone agrees the space industry is booming.
The market case: In a typical IPO, officials test out different prices ahead of the stock’s debut, and set pricing and valuation after some back-and-forth between underwriters and buyers. However, Howard Morgan, the chairman of B Capital, said this process didn’t happen during the lead-up to the SpaceX IPO, in which officials instead settled on a number without consulting the market.
Keep reading