Home

The wiki documents my findings and observations attempting to condense World Info (WI) entries down as compact as possible, down to the last character. This is a fork of the Futureman format, but has deviated away from it more and more each day.

What do you need?

• This format, like Futureman, is designed to work with EWIJSON.
  • New here? What the heck is EWIJSON?! It's a script that provide more control over how AID uses WIs. Normally, AID puts a triggered WI up at the very top (hidden to viewer), but EWI lets you place it anywhere. Because of distance weakening (the higher up something is, the weaker its importance to the AI's writing direction), combined with concepts that are inherently weak already (unusual fantasy races, weird behaviors, concepts the AI doesn't understand well), EWI is a useful tool to move these down lower to counter this problem.
• Time. This format is tedious and time consuming. You'll need to have a token viewer handy, like this one. There are others out there. Ask around on the discord for other options.

Strengths and Weaknesses

• It's primary strength, as the name suggests, is reducing WIs to as few characters as possible. This allows more characters to be reserved for the actual story content instead, and/or being able to trigger more WIs without overwhelming the memory.
• This format does not make any promises on being the most accurate or strongest of the formats. There is some expectation that you will need to occasionally reroll an output, but not so much as to frustrate the player into giving up on a concept.
• This format normally does not concern itself with an overbearing amount of entity definition. It's more about pointing the AI in the right direction and letting it do its thing.
• Due to its nature (explained later), this format is unlikely to work well on anything other than dragon. I do not use griffin enough to comment on how well it compares to other formats.
• Neither a strength nor a weakness, but full disclaimer: I primarily play in 3rd person mode, and only using story input (as opposed to Do or Say input), and with a 1.1 randomness. My bread and butter genre is high fantasy medieval in a world with a wide variety of colorful races.

No categories beyond this point

Best I just teach by example, and explain why things were done the way they were.

< Cheeps♀AvianPink feathers obnoxious social noisyhappy sing song>

• First, the encapsulation. Those familiar with futureman may already be aware of << >>>> encapsulation.
  • Futureman encapsulation is more resistant to the WI format leaking. Most of the time I do not encounter leaking, but the few times I have typically are when I'm breaking the fourth wall and prompting the AI to try and test out a WI for accuracy.
  • More niche, I've seen the format leak when talking to characters whom intentionally have bad grammar for whatever reason.
  • Monky, author of futureman, states < > is the second best encapsulation according to his testing. It should be noted, however, there are a wide variety of tests and scenarios to account for, and reflects in one set of variables may not reflect in another. In my experience, < > has suited me well. This format is all about shaving off every last little character, and little by little, it adds up.
• Next, the leading space before the name. While this format hacks off leading spaces left 'n' right, special care is given for the subject noun. We'd like an exact token match for how it will most often be seen in content. Predicting that it will most likely be used mid-sentence, we leave the leading space. But there are exceptions, explained later.
• The name itself. If you have the luxury of picking any name you'd like, try to pick one that is token-friendly. Here are some thoughts on this:
  • If you use an irl name that's somewhat common, chances are it'll be a single token, and the corpus may already include gender bias to it. But it may also include some other undesirable biases, you never know. Single token names are very strong with keeping the AI's facts lined up.
  • Single token names-that-aren't-names seem to work fairly well, especially for animals. For example Lumin or Snow.
  • When it comes to multi-token names, I aim for two tokens if possible. But what's more important is token length.