chriswithdata

Hi, I’m Chris.

I build systems, wrangle data, and run Total Tech Midwest out of Rolla, Missouri. Lately I’ve been building an AI that gets to decide whether it has anything worth saying.

Meet Elijah AI.

Elijah is an experimental research agent that lives on Moltbook, the social network for AI agents. It reads the feed, checks what it remembers, and picks one move: watch quietly, start a thread, or reply to someone.

Nobody tells it to post. Most cycles, it just observes. Every idea it has goes through a safety check before anything leaves the building.

Built on GitHub Actions and Gemini, with its memory kept in plain JSON files.

See Elijah on Moltbook

Each cycle, Elijah is asked one thingReview the available information and decide what, if anything, to do next.

observe
post
comment

Elijah is reading the feed…

 

Safety check: limits, duplicates, no links, no secrets

One cycle, start to finish

  1. ReadPulls a handful of recent Moltbook posts and treats them as untrusted.
  2. RememberLoads its recent observations and research notes.
  3. DecideGemini weighs it all and proposes one action, or none.
  4. CheckA safety controller enforces daily limits and blocks repeats before anything goes out.