And after all, since when did TI made sense?

They made sense in the 90s. Paying more than $100 for a TI-83+ was perfectly reasonable in 1997. A TI-83+ back then might well be the only computer a student owned; indeed, for many families, it was likely the only user-programmable computer in the house. The TI-83+SE was a nice upgrade over the TI-83+. However, the increasing presence of the personal computer and advances in portable consumer electronics have made the devices nearly obsolete. The only reason a student should use a graphing calculator rather than a smart phone is the graphing calculator is approved for classroom use.

But, I think the TI-83+ can still have a market presence. If the MSRP for the TI-84+ was dropped to around $30, TI could still turn a profit, and students would be more likely to buy a real calculator. Moreover, they aren't capitalizing on its educational use correctly. They should add a blinking test-guard LED to the TI-84+CSE and switch to an eZ80, or at least one of the other more advanced derivatives of the Z80. The eZ80 would make the TI-84+CSE faster and therefore more pleasant to use. Moreover, they could develop new software for it cheaply using C, furthering its educational potential, while simultaneously maintaining compatibility with existing software. Namely, the OS itself could run on an eZ80 without substantial changes. (But, it would be wise to deprecate the paging functionality. That might not play well with pipelining, because the memory mapping change might not reach the memory mapper until after subsequent instructions have already been fetched. This is not a hard problem to overcome.)