It takes more kernels toward the end because its' random. It can never be linear if the researched module selection is random. If the random results that are already researched are ignored, it would actually be faster because your removing results from the possiblity pool. A linear approach could be created, but that would mean all items in the branch would be researched at the same rate. The result would be that, instead of completeling a few modules as you go, they would all finish at about the same time. Functionally that would mean you would go weeks or months without being able to build anything, then all at once you would know the whole branch of that tier.

I think research should be changed in a way that you can achieve single goals faster without beeing forced to research everything, but there still need to be something to research after a year playtime (for researcher enforced by large corp it will be always quite faster, but it should no be usual for soloplayer to have 100% research in first year). Nevertheless it should not be possible to focus on one item only and have this on T4 after ~1 week research.

I would propose, that future research is even a bit slower than today, but you should be allowed to focus on item group (eg. weapon, electronic equipment, nexus) Kernel-color should direct what to research in a manner more clear to researcher by eg. giving Nexus or electronic equipment also color. This would lead to fact, that a specialist researcher is earlier in a position to deliver something to market (selling T2 small autocannon is not really effective) and to the way that a corp could decide wheter to have on main prototyper or several specialised

THe usage of kernels for prototyping is on one hand a good way to ensure, that there are still kernel needed after prototyper finished research, but I think it is on other hand to expensive for prototyper who did not finish research to handle double the kernel demand and would prefer to have no need for kernels in Prototypes.

hmm, nice diagram,

except: i would make the direct built items "non prototypes" (=usable, but not Reverseable) and the KB route creating CTs for mass production - if any restriction in that direction necessary.

hunters suggestion also included kernels when building CT only from KB. Basic idea is: kernel never lose value for the one that creates a CT.