Field notes
Technical deep-dives, industry analysis, and dispatches from the team building the runtime for orbital compute.
Compute Isn't a Placement Problem Anymore
For fifty years, distributed systems have asked one question: where should this workload run? Orbital compute forces a different question, and it's going to change how we build infrastructure on Earth too.
Chips in Orbit. Now What?
Google's Suncatcher prototype, the first Orbitworks satellite and TakeMe2Space's pathfinder shared one October rideshare. The software layer is sti...
NORAD IDs Passed 99,999. Here's What Broke.
Catalogue numbers above 99,999 need the Alpha-5 TLE format. We missed it, SGP4 returned NaN, and our stale-elements guard blamed the wrong thing.
Three Edge Constellations in Two Weeks
Orbitworks, Sophia Space and CAS Space all moved on edge-compute constellations in September. Several customers on one satellite is a scheduling pr...
The Hardware Layer Is Funded. The Systems Layer Is a Research Agenda.
Starcloud's $250M extension and the BUPT paper on AI infrastructure in space landed the same day. One funds the hardware. The other maps the system...
Rotastellar Joins NVIDIA Inception
Rotastellar has joined NVIDIA Inception. What we applied with, and what it changes about the on-satellite execution layer we are building.
A Stale Element Set Lies Quietly
A postmortem on a catalogue sync that stopped after a migration, and on SGP4 quietly propagating old elements into orbits below the Earth's surface.
Software Can't Fix Radiators. It Can Fix Utilisation.
Radiators and launch cost are hardware problems. What software owns is how much of every orbit a compute node spends doing useful work.
A Million Compute Satellites Is a Conjunction Problem
A petition asks the FCC to review orbital data center filings totalling over a million satellites. What that scale does to conjunction screening.
Checkpointing Is Enough. We Measured It on 50 Orbital Traces.
Our arXiv paper asks whether availability-aware optimisers beat plain checkpoint-and-resume across orbital gaps. With a fair baseline, mostly not.
Bring Your Own Workload, From the Terminal
The Rotastellar CLI is on npm. Package your own program, run it on the orbital compute fleet and bring the results back, all from the terminal.
The First Orbital Workloads Won't Be LLMs
The first paying orbital workloads will not be language models. They will be small, data-heavy jobs next to the sensor, and the downlink is why.
Your Satellite's Battery Is the Scheduler's Problem
A new arXiv paper shows more compute in orbit shortens battery life. Why eclipse depth of discharge is a scheduling input, and what the planner does.
The Demand Side Showed Up
A hyperscaler at the table turns orbital compute into a capacity question: when and where will compute be available, not whether a chip survives la...
Inside CAE: How a Feasible Plan Gets Built in Under Two Seconds
An engineering walk through CAE's four-phase pipeline, from orbital environment to schedule, that builds a feasible plan in under two seconds.
Introducing the Rotastellar Operator Agent
Rotastellar now has an execution layer: a Rust crate that runs on satellites, executes workloads and streams events back to Mission Control.
We Shipped the First Orbital Compute Scheduler. Here's How It Works.
Rotastellar CAE is live: the first API for constraint-aware execution plans in orbit, built on SGP4 propagation, eclipse detection and ground passes.
Everyone's Racing to Launch Orbital Data Centers. Nobody's Talking About the Software.
The orbital data center gold rush is real. But getting GPUs to orbit is the easy part. The hard part is the software stack that makes them useful.
Rotastellar SDKs Now Available: Python, Node.js, and Rust
We've released official SDKs for Python, Node.js, and Rust. Get started with orbital intelligence and space compute planning in minutes.
Sustainable AI: Why Space Might Be the Greenest Place to Train Models
The physics and economics of training AI in orbit, and whether space really could be the greenest place to run energy-hungry models.
Will 2027 Really Be the Year AI Goes to Orbit?
A critical look at the 2027 timeline for orbital data centers: what is likely, what is possible, and what is probably not happening yet.
Thermal Management in Vacuum: What the Simulations Show
Counterintuitive findings from modeling heat rejection for orbital data centers, and what they change about how these systems should be designed.
Latency Myths in Orbital Compute
Why the latency numbers quoted to dismiss orbital compute compare the wrong things, and which delays actually matter for real workloads.
Conjunction Analysis at Scale: 10,000 Satellites and Growing
Screening 10,000+ satellites against 35,000+ debris objects is a scaling problem more than an orbital mechanics one. What breaks, and what we built.
The Economics of Space Data Centers
A framework for when orbital compute actually makes financial sense, and why the honest answer depends on the workload, the conditions and the arch...
Introducing Rotastellar
Why we are building software infrastructure for the orbital computing era, starting with the planning tools every team keeps rebuilding from scratch.
No posts match this filter.
Get the field notes in your inbox.
Monthly insights on orbital computing and space infrastructure. No fluff, just technical depth.
Join 500+ engineers. Unsubscribe anytime.