I'm the developer of JustDNS, a self-hosted filtering DNS resolver for your whole network. I built it because I wanted more than a block count. I wanted to see which companies my devices actually talk to.

What it does

  • Blocks ads, trackers and malware, like Pi-hole or AdGuard Home.
  • Privacy dashboard that attributes DNS traffic to the companies behind it, with a per-device breakdown and a world map of where answers resolve. You can block a company for a single device or the whole network.
  • Encrypted upstreams (DNS-over-TLS) by default, plus DNSSEC validation.
  • Per-device and per-group policies, user accounts with roles and TOTP.
  • Optional multi-node high availability.

Running it

  • Docker / docker compose, or a one-click Unraid Community Apps template.
  • AGPL-3.0, no cloud account, nothing phones home.

Repo: https://github.com/bonderaustriaAT/JustDNS

It's still young, so I'd genuinely like feedback from people running Pi-hole or AdGuard today: what would make you switch, and what's missing? Happy to answer questions.

EDIT:

AI Disclosure:

  • Design - No AI
  • Implementation - Hints
  • Testing - Assisted
  • Documentation - Pair
  • Review - Assisted
  • Deployment - Assisted
you are viewing a single comment's thread
view the rest of the comments
[–] 15 points 2 hours ago (5 children)

How much Claude is in this?

  • source
  • hideshow 5 child comments
  • [–] [S] 7 points 2 hours ago* (3 children)

    The design, architecture and decisions are mine. I use AI as a tool (mostly OpenCode, sometimes Claude) for structure and the occasional code snippet, and most of all for the UI, since design isn't my strength.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 1 point 24 minutes ago*

    Your commit patterns suggest you use AI for more than you share here. A feature lands, then 15-60 minutes later a fix commit occurs. This is pretty typical when a second agent reviews the work of the first. I run the same pattern myself.

    Here are some examples that I see just on Nov 14:

    • the HA cluster rolling update (22 files, about 2,500 lines at 17:11, review fixes for a start race, downgrade protection and worker hardening at 18:06),

    • GeoIP enrichment (19:30, review fixes at 19:43),

    • and the world map and geo-watch (1,650 lines at 22:00, review fixes at 22:14).

    These are backend features (cluster sync, crypto, update worker). Nobody writes that much Rust by hand that fast.

    Tons of comments like an AI would write. You're certainly using AI to write tests, too.

    It's okay to use AI for hobby projects, but when you're offering a product to the public, the public deserves to know the products provenance.

  • source
  • parent