Hi, thanks for the feedback!
There's one diclaimer I have to get out first: the funnel mode is auto-computed by the algorithm, and it's based on the inversed depth. This means, that even if atomic-bomb, silo, or other tech are the "last one" you research on that branch, since the previous tech are shared between all tech deep down to automation, it could happen that that particular last tech of the branch is actually located at (LAST_LEVEL - N) looking at the complete tech tree, so it gets multiplied by N (again, even if you see it as last tech of that branch).
That said:
-
I reckon 20k science packs for automation is not fun and I probably should exclude automation from the computation or give it a max value (e.g. 10-100-1k packs?) that cannot be bypassed, since it's handcrafted. This for sure is something that I care to fix!
-
Float multipliers: using integers has prevented many bugs and deadlocks in the code since it's safer. I'll see if I can make it work with floats too, but I cannot guarantee anything upfront since there are many modalities and mod compatibility to take in account, but I'll try. Allowing float multipliers will allow a better finr tuning of the difficulty, as well as making "reverse exponents for all techs" as you requested since the costs get actually divided :)
I can't promise much on the final values of "Atomic bomb" foe example, since it's based on the absolute depth of the research in the tech tree, but I'll look into it to see if there are any tweaks to make it more user-friendly :)
If there are other concerns or you belive I skipped something, please let me know again, thanks