Why I Started Godel Space

After two decades of building systems at Capital One, T-Mobile, Thomson Reuters, and Trilogy, I found myself drawn to a question that wouldn't let go: why do we launch billion-dollar satellites into orbit and then treat them like dumb cameras?

The Problem I Couldn't Unsee

Today's earth observation satellites capture terabytes of imagery every day. But nearly all of it follows the same pipeline: capture, downlink, store, process on the ground, deliver insights hours or days later. For a wildfire spreading at 80 mph, those hours cost lives.

The technology exists to do better. Foundation models for remote sensing have matured. Edge compute hardware like NVIDIA's Jetson platform can run these models. Satellite buses increasingly support third-party compute payloads. The pieces were all there, but nobody was assembling them into an autonomous system.

Why Agents, Not Just Models

Early on, I realized that running a model on orbit isn't enough. A model can classify a pixel. But an agent can decide what to observe next, prioritize which detections matter, and choose the most efficient way to communicate findings back to Earth. The difference between inference and autonomy is the difference between a camera and an observer.

This is what drew me to the agentic paradigm. Not just software that processes data, but autonomous systems that orchestrate perception, triage, and communication. Agents that replace the entire observe-downlink-analyze pipeline with on-orbit intelligence.

Why Now

Three things converged to make this the right moment:

• Foundation models: Remote sensing foundation models like Prithvi have reached 300M+ parameters with genuine multi-spectral, multi-temporal understanding.
• Edge hardware: NVIDIA Jetson AGX Orin delivers the compute needed to run these models at the power envelope satellites can support.
• Satellite operators: Companies like D-Orbit and Loft Orbital are building platforms specifically designed for third-party AI payloads.

What We're Building

Godel Space is building autonomous AI agents that deploy to satellite payloads. Our dual-tier architecture pairs on-orbit agents (for speed) with cloud-based Remote Sensing Foundation Model validation (for accuracy). The edge makes fast decisions; the cloud provides ground truth.

Our first applications focus on disaster response: wildfire detection, flood mapping, and structural damage assessment. These are use cases where minutes matter, where the current hours-long pipeline isn't just slow but is actively costing lives and property.

The Name

Kurt Godel proved that any sufficiently powerful formal system contains truths that the system itself cannot prove. It's a profound insight about the limits of closed systems. We named our company after him because we believe the same principle applies to earth observation: the most valuable insights about our planet won't come from the ground looking up at data. They'll come from systems intelligent enough to understand what they're seeing, on orbit, in real time.

This is what I'm building. If you're working on similar problems, or if your satellites need to think for themselves, I'd like to hear from you.