• 11 Posts
  • 309 Comments
Joined 2 years ago
cake
Cake day: June 7th, 2024

help-circle



  • I looked into CubeSats in a previous job. Basically, there are four questions:

    • What’s the purpose?
    • How to design and build one?
    • How to launch it?
    • How to collect the data?

    Part 1: this is the back of the napkin sketch. What are you trying to do? Weather, water, fire, radiation, or air data? Imaging? Has anyone already done this? What’s the plan?

    Part 2: you can DIY the whole thing, starting with the CalPoly CubeSat workshops: https://www.cubesat.org/. They’re the folks that started the whole thing.

    There are also kits and services out there. One example is Pumpkin: https://www.pumpkinspace.com/, but there are a lot of others like it out there. You want to figure out what sensors you need, mechanisms to orient the sensors, radios, power management, etc. Also, what’s the lifespan before it descends into the atmosphere and burns out.

    Part 3: The big problem is launch. You need to eventually get it up into space. There are commercial services, but you’re looking at $10K-$50K and up to get into the queue. Another option is to go through NASA’s Launch Intitiative: https://www.nasa.gov/kennedy/launch-services-program/cubesat-launch-initiative/ or ESA’s Fly Your Satellite program: https://www.esa.int/Education/CubeSats_-_Fly_Your_Satellite

    These require being part of a non-profit or educational institution. And the waiting list is long. Like, years.

    Part 4: OK, now that you got it up in space, what do you do with the data? It’s circling the globe and there’s a narrow window where the radio can connect to an earth station, send the data, and maybe receive instructions like where to point the sensors. Forget about OTA. You won’t have a large enough connection window or bandwidth to do that.

    You can roll your own comms, or you can see about using an existing service, like AWS Ground Station: https://aws.amazon.com/ground-station/. Microsoft had a similar service called Azure Orbital, but they retired it last year.

    After all is said and done, you now have some cool data. You’ll want to process it and use it for something. This goes back to step 1. Figure out what’s the purpose, what you want to get out of it, and work backward. You can use the AWS service, pipe it into an S3 bucket or store it in a database, then run analytics and visualizations on it. If you want realtime, it’ll cost extra.

    It won’t be cheap, but it will likely be a lot of fun. I proposed several projects in a past life. We got pretty far, but the launch window was years away and by then I was heading out. All this is an infodump of what I learned back then. Hope this helps.





  • On the AI coding IDE side, VSCode has pretty much hoovered up everyone, mainly because JetBrains offered their own AI option, which kept competitors away. On the server side, though, integrating with AI is still wide open.

    You eventually have to hit Python because of all the ML libraries. But you can run that as a separate microservice or process. Here’s a chance to do something whacky, like let JS invoke Python-ML inline, or port the main ML libraries to JS, or cross-compile JS to CUDA (just spit-balling here). It’ll be a lot easier to try these experiments than trying to push it upstream into Node.

    Plus, Bun is used by a bunch of cross-platform CLI tools, including Claude Code, so they can make sure there are no breaking changes.

    TBH, I’m surprised nobody’s snapped up Mojo (and Chris Lattner). They have a lot more advanced, AI-relevant, cross-platform tech.



  • It’s not a bad outcome. Bun is cool but has $0 revenue and some hand-wavy thing about future paid cloud services. This way, larger companies will give them a more serious shot than they would a small startup.

    It still doesn’t have a revenue story, but it’s now strapped onto the side of one of the few AI companies with a decent chance of surviving the next AI Winter. And if Anthropic goes sideways, the Bun engineers can fork the code and keep going.