On TI-68k/AMS, TI provided some documentation about the math system, and it was used (and expanded) by some people to actually build native code math programs which can be called from the Home screen (whose equivalent in Nspire documents is a calculator page). I'm both one of the users, and one of the reverse-engineerers expanding third-party knowledge of the CAS ( as I told in the meeting in Paris, even nowadays, see https://github.com/debrouxl/gcc4ti/commits/next ).
And we did math programs in C/ASM precisely because it allowed faster and more powerful math programs, or even things that cannot be done in BASIC, even with built-in functions like part().
[...]

By using functions at a lower level, my C implementation of the Aitken delta^2 algorithm on TI-68k/AMS requires a number of operations close to the element count of in the input list (O(n) complexity), while the original TI-BASIC implementation requires a number of operations proportional to the half of the square of the input element count
(O(n^2) complexity). My C program does about 50 operations for a 50-element list, the TI-BASIC program requires more than 1000 op[eration]s...

But on Nspire/Phoenix, there's no official documentation about the Nspire document system, and it's hard to reverse-engineer by ourselves (which is, by the way, "reverse-engineering for interoperability purposes", exempted from DMCA prosecution), because the Nspire OS is so huge. We don't know how to do _interactive_ programs, where users can enter arbitrary math functions and read/write values in per-document variables.
Sure, I have performed enough reverse-engineering on OS 1.7.2741 to show that the Nspire series' CAS is the same as the TI-68k series' CAS (same data structures, same function names - derived from my own disassembly of AMS 3.10 for 89T, which itself contains many names that were publicly documented by TI and a number of names I made up and have little chance of matching TI's), and to make a non-interactive port of my aforementioned Aitken delta^2 program (with a hard-coded function, so it's useless for practical purposes): http://www.ticalc.org/archives/files/fileinfo/437/43727.html .
But I'm not motivated to go further...

[...]

If TI wants more math & science programs (and the faster, more powerful math manipulation which native code can provide), at a minimum, we need to be able to integrate both ways with Calculator screens (similar to what Nspire BASIC libraries can do: take arguments, return values), how to read/write per-document variables, and how to make a native
code program well-formed for the purposes of containing per-document variables (current Ndless programs are not well-formed).

OSLauncher can be blocked in mere seconds if they felt like it.

They just don't care, that's all.

(Then again, have we even showed it to them? )

What games we write will have no effect on the market. Add to the fact that the majority of textbooks are written for the TI-8x calculators, and there you have it: monopoly. Unfortunately, Casio didn't have a chance from the start.** You would need an antitrust lawsuit to fix things.

Again, I suggest we take the road where iOS jailbreaking has done - "new OS update? Great! Let's hack it again!" The JB community there doesn't even bother to convince Apple.

obviously they're going to exploit us as much as possible

3.  A few people have gone over to the dark side and paint a rosey picture of the future for nspire programming which doesn't seem to have any basis in fact and needs to be filtered by reality checks.

So where are the opportunities for the very talented people who post here? What do they tackle next? Where can they make their mark?