Blog

News, deep dives, and the life of the project.

The longer story behind what we build. For the terse, per-month changelog see Dev updates. Prefer a reader? Subscribe via RSS.

We asked an AI to tear Axiom apart

  • A stranger gave me a tip: feed your project to a capable AI and tell it to attack your assumptions, not flatter them. So we did.
  • The uncomfortable finding: our 'deterministic firewall' guards the game-state numbers, not the story text, so the prose can still fib for a turn.
  • Nothing here is fatal. Two fixes plus our existing test-harness and actor-model plans just moved to the top of the roadmap.
Read →

Under the hood: how saves and universes work

  • We ran a full quality pass on the save and universe system, so here's a tour of how the thing actually works.
  • Your universe is plain text you can read and version, the database is just a throwaway cache, and each save is one self-contained, shareable file.
  • The QC caught a sneaky rewind bug, we fixed it, and we added a test so it can never sneak back.
Read →

What we took from Hindsight

  • Axiom's memory got a serious rework, inspired by an open-source project called Hindsight.
  • NPCs can now keep facts, form opinions that shift over time, and hold a living profile of who they are.
  • It's all optional, and the offline mode that never touches an AI still works exactly like before.
Read →

Axiom has a blog now

  • We finally have a real place to talk about the project, not just a dry changelog.
  • Expect deep dives, the thinking behind the big changes, and the messes we walk into.
  • There's an RSS feed too, so you can follow along without ever opening the site.
Read →