[–] 4 points 1 month ago

Didn't they default to slower speeds the rest of the series and only increased during emergencies? It's been a while, but I seem to remember it actually spanning many episodes, even if they didn't harp on it. Pretty good for an episodic series, really.

  • source
  • parent
  • context
  • [–] 19 points 2 months ago* (last edited 2 months ago) (2 children)

    Lots of overcomplicated answers, imo.

    Here's what I did:

    1. Set up a DNS server (I just used Pi Hole docker)
    2. Tell router to use the DNS server
    3. Register domain
    4. Set up Caddy with https to use DNS challenge with registrar's API key for Let's Encrypt or similar.
    5. Add subdomain DNS entries to the DNS server for each service. The records point to Caddy's IP (e.g. jellyfin..com)
    6. Add each container to Caddy's list

    Now every device at home can hit all services by domain name over https. No need for any manual configuration on any client device (certs, hosts, etc.)

    Bonus: Set up split DNS and subnet routing with Tailscale so that it uses the DNS server for that domain. Now any device connecting to the tailnet can connect to the services just like at home. No need to reconfigure apps, etc.

    Simple and it works like magic.

  • source
  • [–] 1 point 8 months ago

    You can have calibre auto import from a folder. Though be careful because it deletes them from that folder (you might want to do single direction sync into that folder).

    And if you share the db with calibre-web or have some other sync method that works, you should be good to go.

  • source
  • parent
  • context
  • [–] 2 points 8 months ago

    Yes, they share the database. Yes, I add the books to calibre. I think I have it as read only for calibre-web.

    Regarding overcomplication, I admit the initial setup took a while. But I've had it running for a couple years now with no major issues.

  • source
  • parent
  • context
  • [–] 4 points 8 months ago

    Funny enough one of the best use cases I have found has been for information retrievel then generation.

    Spot on. And not just baked into the model but also extracted from the code base.

    Roo + Claude can find the code I want, make flow charts, etc.

    Similarly, I use it as a super search tool for my notes I take in Markdown.

    And lastly, meeting transcription + custom AI summary is very helpful as a starting point.

    All of these are similar in that they don't need to be 100% accurate. They're a starting point to save me time and find stuff I may have missed or forgotten. The actual decisions and brain work are done by me.

    As for actual code generation, I'm not fully sold yet. I use it situationally, but no PR review will be able to tell because I make sure anything it generates is exactly how I would have written it. That sounds weird, but what I mean is I either use it as "super auto complete" for small stuff or I have it build a draft and heavily edit it until it is good code that I understand completely.

    (Senior dev with over 10 years of experience.)

  • source
  • parent
  • context
  • [–] 15 points 9 months ago* (2 children)

    I've been looking into this as well and just bought my first components.

    I'm trying Meshtastic first and then will try Meshcore.

    What does everybody think of Reticulum Network and RNode? It honestly seems superior conceptually to Meshtastic/Meshcore, but I'm not sure how good it is in practice or if anybody is actually using it.

  • source
  •  

    There's a few 3D printable corner cut jigs out there. I designed my own for three main reasons:

    1. There's a little ruler on the side to mark the width of the fold-over section of the cloth
    2. It's parametric (using OpenSCAD), so you can customize various sizes (spacing from corner, various thicknesses, size of ruler, etc.)
    3. I like to model stuff myself because it's fun.

    https://makerworld.com/en/models/2000894-corner-cut-jig-with-ruler-for-bookbinding#profileId-2154412

    view more: next ›