Speedy Atomic Actions doesn’t change this, they relate to when Atomic Actions are completed, not whether they are completed. There were two possible scenarios for Clucky’s most recent Start the Battle Atomic Action. I endeavor to show neither results in the gamestate we see now: 1) Bucky Queued Up In this case, Bebo plays. When Clucky goes to create a combat with Bucky’s Pet, Clucky is unable to do so. From the relevant section in the ruleset, Section 4.3 Atomic Actions: If a Trainer arrives at a step in an Atomic Action and they cannot perform that step, they undo all the steps they have performed of that Action and are considered never to have performed that Action. If the step required posting a blog post, it is undone by editing the blog post to explain that its creation was undone (rather than by deleting the post). If the step required posting a comment, it is undone by adding another comment to the same post, explaining that the comment was undone (rather than by deleting the comment). A post or comment that is undone this way as part of correcting an Atomic Action no longer has any effect on the ruleset or gamestate. The Ruleset makes clear the action is rolled back. This did not happen because we see no evidence of comments explaining that a Start the Battle action was undone. 2) Bucky did not Queue Up In this case, Bebo does not play, so at the very least, the action was resolved incorrectly. I suggest the following corrections: Regardless of which scenario we are in, rollback and redo the Start the Battle Action with new pairings, no Bebo, and no Bucky’s Pet.

In a previous post I showcased a demo of an LLM getting a bit of a gamestate from OpenTTD. While the demo worked it really didn’t live up to the idea of having an LLM drive a organisation within OpenTTD. Over the last couple of weeks I spend some tokens to develop this idea a bit more and see what I could do… I ended up with an improved state engine, better interface, multiplayer/multi-llm support and a custom web dashboard!