cuda-the-spire VS sm64

I found this seed primarily using StS CUDA code to sieve for seeds that get Neow's Lament on floor 1 and can buy Juzu Bracelet at the first shop because of the Sssserpent event while also avoiding event node combats. I accounted for unlock manipulation (changing the size of the common relic pool), along with details of the Java random number generator's shuffle method to increase the throughput of seeds.

GPU-accelerated seed searching for high-performance searches such as the first mathematically proven unwinnable seed, and

Depending on the player's path, the probability of hitting a shop or treasure is usually between 2% and 8%, and in this one, we get lucky 8 times in a row! We found this seed by running a GPU-accelerated search. The run also minimizes the number of rest sites visited, because they have a fixed-time animation that cannot be avoided. It is theoretically possible to get one more shop/treasure room instead of the 1st rest site, with enough luck. Perhaps some other day!

I agree that Neow options on floor 0 also make it extremely difficult to find an unwinnable seed with this criterion. If I dug deeper into the Jaw Worm AI, I could probably put together a GPU-accelerated sieve for candidate seeds in the CUDA code I've been playing with.

The SM64 is still going strong https://github.com/n64decomp/sm64

Nintendo won't allow any binaries floating around though.

Romhacks are typically modifying the compiled binary ROM image. Kaze' work is based on the painstakingly disassembled code from the n64decomp project[1]. He's working in C, modifying the game and compiling it again for the original hardware. Not sure I'd call that a "romhack".

Great videos though!

[1] https://github.com/n64decomp/sm64

A team or something reverse engineered super mario. I think that's the repo and it's mostly in C

From course_table.h and course_defines.h we can find that: COURSE_MIN == 0 and COURSE_MAX == 25 - so it looks like every secret stage is a full-fledged "course" for the purposes of counting stars, bringing us to 25 courses + castle stars (COURSE_NONE). So, the maximum possible value for starCount should be 7 * 26 = 182.

Multi-year decompilation project: https://github.com/n64decomp/sm64

The port source code of the decompilation is still up on github and Nintendo hasn't taken it down in years, cause they can't since everything has been done with legal methods

This is false, the source code never leaked; it was meticulously and painstakingly decompiled by hand.

It sounds like, rather than a ground-up rewrite, COBOL should be treated as an object-code language, and "hand-decompiled" (ala efforts like https://github.com/n64decomp/sm64) into an HLL that can, at every point, be losslessly transpiled back into the original COBOL.

I know the tooling for doing that doesn't currently exist... but paying someone to develop it would be cheaper than any one of these ground-up rewrite projects!