Glad to see some thought going into this, as it’s definitely very important. What you’ve lined out for mutation strategies seems like a good base to get started on making things look interesting.
One thing I want to say right from the start though. I have mainly stated this in the multicellular concepts thread, so many people will have missed it:
Our current pattern of “enter Multicellular Stage with one cell, then build, differentiate and design from there” is a very poor fit for the real-life development of multicellular life and is an awkward transition from potentially living several generations mainly in colonies. For both of these reasons, it would be in my opinion much preferable to start the stage with a little blob of identical cells that you then start differentiating.
So it would be good for auto-evo concepts to have some notion of adapting cells already on the grid into different cell types.
For me as well. Because you haven’t mentioned it, but these selection pressures also apply to the player species, that’s how the population calculations work. So the scoring systems need to handle player-made cells decently well, or we’re going to get the same problem as what I have been trying to eliminate from Microbe Stage: The player very well designing an effective organism, but the auto-evo system saying “Nah, that’s terrible, you’re going extinct.”
One option is to go a bit mild on geometry-related scoring so that players don’t get destroyed and use mutation strategies that can only make “good” geometry like you said. But I think there are limits to that.
Microbe stage auto-evo does take positioning into account for some of the organelles, like movement tools (and in my ongoing PR I’ve added pili to this). Could we do something similar with cell placement in macroscopic? “Mouth cells” should go towards the front, etc.?
I also have had plans for this in the Microbe Stage that I still want to do. But the only thing holding me back from this is auto-evo calculation times. And I don’t think that’s going to get any easier for Multicellular.
Other assorted thoughts:
Cell differentiation
You briefly referenced it, but we functionally have two layers of editing here: cell level and organism level. So I’ve been wondering if we can have a two-layer auto-evo.
I think Deus and I have settled on the idea that creating specialised cell types is a big theme of the Stage. So what I would want to do is have the organism-level auto-evo decide “We want to have mouth cells on the front”, then makes a new cell-type, names it “Mouth” and slaps it on the front.
It then passes the buck to the cell-specific auto-evo which uses the mechanics of our Microbe Stage auto-evo with a special set of miches to evaluate how good of a mouth this mouth cell is, and evolve it in that direction. The organism-level auto-evo only gets fed back that score and decides when it wants to improve some cell-type.
To add another example:
This would be a case of organism-scale auto-evo wanting to maintain a perimeter of “skin” cells, and then instructing that cell type to use the “skin” miche nodes.
Time-scale split
I want to actively advocate for having Multicellular Stage on a separate time-scale from Microbe Stage. Meaning that once you enter it, auto-evo spends most of it time on Multicellular Auto-evo, only occasionally doing a round on Microbe-stage species. Auto-evo calculating times are already limiting for just Microbe Stage. So I think we need to do this to have any depth in Multicellular auto-evo.