You might've heard about "caveman optimization" that presumably allows you to save tokens when talking to LLMs. I don't know whether it works or not, I don't use LLMs much, but it makes sense.
Here ChatGPT itself explains it(taken from reddit):
I was really struggling to write this update for some reason. I'm not a native speaker and I'm also a certified introvert, so public writing was always hard for me, but for this particular update I just couldn't write more than 2 sentences in one sitting.
You probably can already guess where this is going...
Less words, pictures more, game update EA.0.3.1 good.
New Cores
Tasty bird, crazy person, and a bad whether guy. All early, all give Actions, all fun.
Move up - good. Move down - bad.
Loki strong. Many codegates open.
Was too strong for Depth-1. Depth-2 now it lives.
Depth/Difficulty widgets
Alex and Andrew make window pretty.
Depth and Hard now easy to track.
Big core defeat now can see.
Locations
2 stones for 1 shiny.
Many gold for 4 green never took. 5 green much better.
New data place sells big rare books.
Book's heavy. Hits 4 times.
Legacy Code can't be upgraded.
Andrew thinks it's funny. I don't get it.
Corruption is spreading
Animals didn't wash hands. All animals sick now.
Miscellaneous
Demo fresh. Demo deep again.
Bandwidth removed from Mirror copy and Police report. Barry scared.
Instant Message singleton now. 1 Data Express OP.
Depth-3 me wants to make soon. Unga-bunga boss lives there. Tough battle, worthy opponent. Hard to make right.
Conclusion
Thanks for reading. I apologize for low-effort post. And, of course, I didn't mean to offend any cavemen among readers (my ancestors were cavemen too).
If you have any questions - you can always ask them here on Steam or on our Discord. There's also a Trello board with a roadmap here.