AmpBenzScientist @AmpBenzScientist@qoto.org

@PawelK @HN I misread that and revealed how much faith I have in ML. It's powerful but it presents its own unique problems. It's bleeding edge in that field.

@PawelK @HN I also had plans for that but it didn't fit the blind flight idea. You will get better results than ML nearly every time. There's nothing that would stop such a tool from using a double tap analysis.

@PawelK @HN I called it Blind Flight after the way some of the more advanced helicopters navigate in darkness with limited or no visibility. They use instruments only.

The ML portion was named Black Dolphin after a stealth interceptor helicopter. They were all named after helicopters. There was Big Sexy(CH-47F) or Chinook, Little Bird and Mil-17. The Chinook was for trying to document as much as possible. Mil-17 was essentially a database of learned data. Little Bird was training sets and detection of ML crippling data. Little Bird was meant to reinforce Black Dolphin's ML so it wouldn't be so easily compromised.

There's a lot more to the story than that but the amount of work never stopped growing. It would need to be ML like I've never seen it before or it would have required far more projects.

This is something that I suspected that the original Ghidra had some capability of. The VM, odd choice of programing language, P-code and the way that processors were implemented. That wasn't a list of assembly or even based on a processor. It could have easily made use of ML because of how everything was abstract (Mathematics Definition).

I do believe that Ghidra had or was planning on using ML similar to what I was working on. What I was planning would have been inferior to what I suspected they had years before I started.

That's effectively all of the remaining information about that project and the additional reasons for why I stopped.

@PawelK @HN It's a cool concept.

@PawelK @colinsmatt11 @HN I wanted to be able to have a tool that could work on foreign military equipment. There's nothing that would really be known and the best that could be done is looking for certain patterns. That's why it took a different approach. There's more to it and that's another reason why I stopped. It never stopped growing. If it could get some information, that would be knowledge gained about something not known.

@PawelK @HN That's the best part, it wouldn't have any real advantage over other Microkernels. With microkernels controlling much of computer hardware, it could be useful experience.

@PawelK @HN It was a little different than that. I wanted to use Machine Learning algorithms to try and deal with the unknowns. The goal was to get it to recognize common instructions but without knowing what it was working on. It should have been able to deal with Multiclet and other architectures that the world wants to forget.

The project grew into 4 different projects and I realized that it was too much work for something that will have worse results than traditional tools. It could function with little knowledge and get some results. That was the only benefit of that idea. I stopped right before the ML part and realized it wasn't feasible to continue down that path.

It would be like making a portion of Ghidra just for the ML and none of the supporting code. It was doomed from the start. A team could have achieved results but why bother for something that would need to run on a server and run an indefinite number of iterations to get the code to potentially be accurate.

It was all ML and it was designed to work with the unknown. It was far too large of a project and it would have been of no real use. That's the reason why it was stopped. It would have lived up to being revolutionary and that's all it would be capable of because of the design.

So that's the story of what happened to the project and there was a disk failure during the early stages.

@PawelK @HN I'm considering making a microkernel. There are other projects I might decide are more important.

I'm open to suggestions.