DownRange
- 2 Devlogs
- 11 Total hours
a live mission-control deck that tracks real rocket launches, the ISS, and other satellites on a 3D globe, alongside a simulated fleet of orbital AI data centers
a live mission-control deck that tracks real rocket launches, the ISS, and other satellites on a 3D globe, alongside a simulated fleet of orbital AI data centers
devlog #2
devlog 1 ended with four fake satellites orbiting an empty earth. it worked, but it felt kind of hollow on its own.then i saw a screenshot of an app called conflictly, a dense dashboard tracking conflicts in real time. i had a “wait, what if this but for space” moment, real launches, real satellites, real news, not just my sim.
so i built a python backend pulling real data, launch library 2 for launches, spaceflight news for headlines, celestrak tles propagated with skyfield for the iss, tiangong, and hubble. cached, resilient, never crashes even when an upstream api sends garbage.
then i turned it into an actual ops floor look. countdown to the next launch, a breaking news ticker, launch pads that light up and pulse when live, real spacecraft shown as white rings so you can tell them apart from the sim satellites. two clocks, real utc and the sim fleet still running at 60x.at some point i realized the real world tracker and the simulated fleet were fighting for the same screen, so i split them into two modes you can flip between, one for the real sky and one just for the starcloud fleet with its own roster and telemetry.
last pass was small stuff that matters in daily use, color that actually means something instead of just looking nice, and making the launch pad dots easier to click with a hover grow effect.started as watching four fake satellites. ended up blending real live orbital data with a simulated glimpse of what training ai in space could look like.
happy with where it landed
devlog #1
so this started because i read a white paper. not something i do every day, but i came across the starcloud/lumen orbit paper called “why we should train ai in space” from 2024 and i couldn’t stop thinking about it. the idea of orbital ai data centers, satellites doing machine learning in low earth orbit, solar power that never turns off, no cooling costs. it just clicked something in my brain.
i spent probably the first hour just reading and taking notes. what would it actually look like to monitor something like that? like if you were the person watching these satellites, what would your dashboard show? solar power, battery percentage, radiator temperature, whether the satellite is in sunlight or in eclipse. and then the orbit itself, the actual position over the earth in real time.
so i decided to build it. a full mission control dashboard, the kind nasa uses, except for a simulated version of what starcloud might actually run. i picked open mct which is nasa’s open source mission control framework, and started building on top of it.
five hours in and i’ve got more done than i expected. there’s a telemetry server running in python that simulates four different satellites with real physics, power generation, battery charge/discharge, thermal modeling. the frontend connects to it live over websockets and plots the data in real time on an accelerated simulation clock so you can actually watch the orbit happen.
the thing i’m most excited about right now is the 3d globe view i just started. you open a satellite in the dashboard and you get a realistic earth with the satellite orbiting it, colored gold when it’s in sunlight and dim blue when it’s in eclipse. that part isn’t done yet but the bones are there.
more to come soon