i made <a href="https://github.com/xloem/multiplayertimetravel/tree/main/ftlblink">https://github.com/xloem/multiplayertimetravel/tree/main/ftlblink</a><div>some concern around &quot;expected death&quot; being near research fields instead of consenting prototype humans</div><div>but it&#39;s not for humans now anyway</div><div><br></div><div>:s</div><div><br></div><div>--</div><div><br></div><div>so p piles are hard for me cause of thing</div><div><br></div><div>one concern is: where could we model uncertainty coming from, and how?</div><div><br></div><div>possible sources?</div><div>- agent tick function</div><div>- world tick function</div><div><br></div><div>also, conflation around functions and models and representations</div><div><br></div><div>what if tick and world can both return a P structure over their possible outputs</div><div><br></div><div>ok, now for prototype we probably want only 1 of those to be probabilistic</div><div><br></div><div>how about agent. that moves toward human control of an agent. sims can control their simulated environments but not human-controlled characters. good for software.</div><div><br></div><div>_buuut_ for a bot prototype ... and since we&#39;re not actually teleporting humans here .... i think actually region&#39;s tick is the correct interpretation of the logic.</div><div><br></div><div>because if an agent is influenced by a human, and then teleports, it then has different humans influencing it. the humans are the environment here.</div><div><br></div><div>so region is where prototuped uncertainty goes.</div>