>>/3680/
во-перых мы сохнраним номер группы внутри которой действие происходит.

вот что чатжпт предлагает

Here are some elements that you might need to consider when creating this minimal encoding:

    Action Parameters: Besides the action name, it is crucial to retain the parameters of the action. For instance, the action 'move' would require additional information about the direction and distance.

    Temporal Ordering: You'll also need to preserve the temporal ordering of the actions. The sequence in which actions occur can significantly impact the final outcome.

    Action Dependencies: Some actions may be dependent on the completion of other actions. These dependencies need to be encoded. In other words, the fact that Action B cannot be executed until Action A has been completed.

    Conditional Actions: Actions that only occur under specific circumstances (If X, then Y) should be marked as such. Although you've mentioned not storing specific conditions, there might be inherent conditions within the action sequence that need to be retained.

    Repetitions and Loops: If certain actions or sequences of actions are repeated, it might be useful to encode this repetition rather than repeating the entire sequence in the encoding.

    State Changes: Actions often change the state of the world in some way, which could impact future actions. Even if you're not storing the specific conditions of the world, you might need to keep track of these state changes in some way.

    Action Outcomes: Some actions might have multiple potential outcomes. Even if the action name is the same, the effect of that action could vary depending on other factors.

    Intensity/Scale: Depending on the action, you might need to specify the scale or intensity of the action. For instance, for an action like 'throw', the force behind the throw could significantly affect the result.

Remember, this is a high-level, theoretical framework. The actual implementation would depend heavily on the specifics of your use case, the actions you're working with, and the world model you're using to interpret the actions.