By Abbey White, Staff Writer, SatNews
Dispatch from SmallSat Symposium. Coverage and analysis from across the conference, tracking the forces shaping the next phase of the SmallSat market.
MOUNTAIN VIEW. The concept of the satellite as a simple relay, serving as little more than a shiny mirror reflecting data to control rooms in Houston or Darmstadt, is extinct. At the Computer History Museum, the panel on Autonomy and AI in Space Operations delivered a eulogy for the dumb pipe era and a baptism for the intelligent edge. The consensus among the engineers and operators on stage was absolute. We aren’t automating space operations because it is trendy; we are doing it because the physics of the new orbital environment leaves us no choice.
The industry has hit a wall where human reaction times are physically incapable of managing the chaos we have created. The CRASH Clock, which measures the statistical window before a collision becomes inevitable, has collapsed from 121 days in 2018 to just 2.8 days in 2026. Consequently, the human-in-the-loop, rather than being a safety feature, has become a liability.
Survival in the Clam Chowder
Katherine Monson, CEO of Hale SWx, framed the engineering challenge with a visceral analogy that cut through the usual dry technical jargon. Describing the Very Low Earth Orbit (VLEO) environment where many next-gen constellations operate, she noted that atmospheric drag is no longer a constant variable.
“It’s the difference,” she said, “between swimming through chicken broth and clam chowder.”
In the clam chowder, density spikes caused by solar activity can drag a satellite out of its operational orbit in hours rather than weeks. A ground-based operator waiting for a telemetry pass to upload a maneuver command is already too late. Monson warned, “4x drag is not something you can come back from if you have electronic propulsion and you were not already orbit-raising before the drag is heating up.”
This scenario represents the killer app for orbital AI. It isn’t about generating poetry; it implies a satellite sensing the weather and firing thrusters instantly to survive. Monson noted the goal is for satellites to utilize very simple logic trains to orbit-raise always when conditions degrade, without asking or waiting for permission.
The Efficiency Guillotine
Physics drives the survival requirement, but economics drives adoption. Ian Canning, CEO of Eutelsat Network Solutions, offered a brutal look at the mathematics of mega-constellations. Eutelsat has aggressively automated its fleet operations, a move that has decimated the traditional mission control headcount.
Canning revealed the scale of this shift: “Plan for 50-plus operators down to five.”
This reduction is not mere cost-cutting; it marks the only way to scale. You cannot hire 50 operators for every 100 satellites when you are launching thousands. However, Canning admitted that the transition creates a trust gap with legacy customers who still find comfort in a crowded control room. When Eutelsat pitched this lean, automated model to government clients, “they completely dismissed our ability to manage that number of satellites with that number of people although we’ve been doing it for three years.”
Breaking the Fairing
Perhaps the most radical engineering vision came from Joe Landon of Rendezvous Robotics, which is utilizing autonomy to break aerospace’s most fundamental constraint: the rocket fairing. For sixty years, everything put in space has had to fold up like an expensive tent to fit inside a cylinder.
Landon argued, “Everything we’ve ever sent to space has had to fit into a rocket. That’s what we’re trying to change.”
His solution involves launching stacks of flat, modular tiles that use autonomy to self-assemble in space into massive structures. Such structures will include antennas and power generation systems far larger than anything a rocket could carry monolithically. This is a technical breakthrough in its purest form, using software to cheat the limits of hardware.
Landon also pushed back against the instinct to over-regulate these interactions before they even exist. When pressed on standards for these autonomous agents, he countered, “Maybe we don’t need standards.” He pointed to terrestrial agents that negotiate protocols in real-time, suggesting a future where satellites handshake and coordinate without a pre-written rulebook.
The Verification Trap
The lingering question in the room concerned verification. How do you trust a neural network with a billion-dollar asset? Alan Campbell, a Principal Solutions Architect at AWS, argued for a pragmatic, layered approach to trust, which he called the “on the tin test.”
Campbell insisted that every autonomous agent must have a verifiable signature. He summarized the logic simply: “if it’s not what it says in the tin, stop.”
He described a shift where operators build instrumentation directly into the agent-to-agent logic to monitor performance in real-time. This is critical because AI models now upgrade at breathtaking rates. An algorithm that worked yesterday might behave differently after a morning update. The engineering challenge is building a wrapper that ensures a consistent output even as the model evolves.
The Integration Task
The industry is moving past the shiny object phase of AI. As Ghonhee Lee of Katalyst Space noted, “The groundbreaking research has been done for us.” The task now is integration.
Lee, who is preparing to launch a mission to autonomously dock with an unprepared target (the NASA Swift observatory) later this year, dismissed the idea that trust is the primary bottleneck. “I really don’t think trust is the limiting factor here when we’re talking about autonomy,” Lee argued. He pointed to the long history of autonomous cruise missiles as proof that we know how to validate these systems.
The message from Mountain View is clear. The bottleneck is no longer technology; it is architecture. We are moving from a world of bespoke, hand-flown satellites to a world of swarms, self-assembling structures, and edge-computing nodes that think for themselves. The dumb pipe is dead, and the smart satellite has arrived just in time to rescue us from the clam chowder.
