My C/C++ Optimizing Compiler and Simple AI Chat Program

I also created a program sometime in 2008-2009 that emits command-line data compression "programs" of whatever name you like and of varying compression efficiency. You can name it "zpaq" or "nanozip", even many versions of "zstd.exe" like the ones I emailed to Facebook.

Or maybe it was early 2010, I cannot fully recall.

You have a choice of high-compression or fast compression, or parallel/hyperthreading or single-thread programs.

I recalled the 1970s when we were distributing free compression and encryption/decryption programs.

The said program even outputs "self-extracting programs" or "self-extracting archives". 

And maybe I also created a Windows Win32 version of it.

The command line "program-emitting" program may appear smart and like an "AI" of today but actually not. "Like an AI" because it communicates to the user with puzzles.

Not very "smart AI" but it chats to the user and tests the user's knowledge of data compression and computer science in general for it to give varying degrees of compression performance in the programs it emits. Like a guessing game, and riddles and puzzles. 

The program's chat statements or prompts to the users are "pre-programmed" and the users have to play with the capabilities of the program. It may appear to the user as "smart AI" but actually just clever programming of answers and prompts. But isn't that actually how a smart AI is programmed?

If it "learns" and "changes and adapts its actions" for better future responses, then it is a "smart AI", which perhaps I may have programmed in my "data compression program-emitting program." Perhaps.

Just like a basic compiler but with conversation prompts and chats. Which is like "Programming Copilots" available already today in ChatGPT, Gemini, Claude, Groq, Visual Studio and GitHub etc. But we already had done that in 1970s and early 1980s again.

I said they should test my program to chat with many human programmers or developers to see the best compression programs it emits--and also run the program in many parallel threads, meaning, to test it with autonomous AI programs like ChatGPT, Gemini, Gemma, and Claude, Groq, Meta AI, etc.

Perhaps the human programmers can intelligently interact with my program and make it emit "a data compression program which performs on average up to 98% compression ratios"?

Or maybe I limited its performance to just 90% or 95% compression ratios (ave. % reduction). Performance up to just 90% or 95% compression ratios or 10%-5% entropy, but hard to reach for the ordinary user's chat statements or conversations. You have to impress its simple "AI" engine with your knowledge of data compression, and computer science in general.

You may need a genius scientist or an AI by your side. Many AI programs running in parallel to chat with it. Win32 app to Win32 app communications in Windows, or Linux. Yes, I suppose it should run in Linux. 

Or it was "a data compression program at first" wherein you can show it faster and very speedy and optimized code and programming operations and functions that it can use to compress your file. So it is always running and improving itself via new code it encounters. A new idea on how to build faster code but I guess that's how a better human programmer and compiler writer improves and can evolve his programming language compiler.

So my program has become like a compiler whose outputs are data compression programs. 

When I was developing my console-based "simple AI program" which outputs for the user a "data compression program", it clearly was like a "compiler" but the program may not require source code from the users. It like tests the user's knowledge of data compression and emits compression programs of varying efficiency and average compression ratios in performance. Some source code and functions can be supplied and learned by the AI program mainly for speed if it is shown faster program code. But the program's main function is to output a data compression program.

If the user is that much knowlegeable in data compression and computer science in general, maybe the program can output for them a very high-compression command-line data compression program, or Win32 program, like perhaps 98% compression ratios in performance? The chat interface in English was not good (should not deter the curious user especially if he's a programmer) but the program learns from inputs and sure can output "high-compression" and "very fast" compression programs like named "zstd.exe" or "nanozip.exe" in 2008 to 2010.

We had similar AI programs in 1970s to early 1980s (console-based and Win32) so when I remembered it, I tried developing a program of same purpose and functionality sometime in 2008-2010, and also simple Win32 app version.

Perhaps my Win32 app of the said program was a very small message box only, text input and like in CLI programming compiler. It should be very compact since it is about "data compression", after all. (?)

Like the one we showed at Lagro Quirino Highway, near OLFU in late 1970s or early 80s. The bigger Win32 app of it was again probably sent to bigtech NASA and NSA etc.

My program, written maybe during 2008-2010, is an "interactive computer program" like a simple chat app but you may not know that it outputs actual or real console-based "data compression programs". But in my emails I most probably told them so.

The subject of my emails was data compression, after all.

One Nanozip version, for example, was programmed in such a way that its output compressed file is equal to that of Zstd's. You just named them Nanozip and Zstd separately. They themselves can be equal or using the same compression algorithms. Same outputs, or of equal program size outputs of my "compression program-emitting program" which I may even have named first as "nanozip.exe", I cannot fully recall now. Meaning, my nanozip from 2008-2010 can also emit or generate better or best "data compression programs" like even zstd.exe, exactly as the first zstd.exe that was released by Facebook. That is, if Facebook used it, knowing that it is very fast and having seen its next versions.

Or perhaps it was my Zstd from 2008-2010 (that can also emit/output better or best data compression programs) whose output was a compression program which I named "nanozip"? Maybe only later that I created a separate "compression-only" zstd.exe and nz.exe compression programs.

Perhaps the human programmers can intelligently interact with my program and make it emit "a data compression program which performs on average up to 98% compression ratios"?

Or maybe I limited its performance to just 90% or 95% compression ratios (ave. % reduction).

Maybe 90% only, and the better 95% to the likes of NASA and NSA/CIA but it showed that I have control of the program's outputs, via maths.

The "outputs are data compression programs" of varying compression performance and efficiency. Or it was "a data compression program at first" wherein you can show faster and very speedy and optimized code and programming operations and functions that it can use to compress your file. So it is always running and improving itself via new code it encounters.

A new idea on how to build faster code but I guess that's how a better human programmer compiler writer improves and can evolve his programming language compiler. That's how it is done, via the computer's ability to run autonomously, running in the background looking for solutions and then alerts you when some solutions are found. Generative AIs like today's AlphaGo, AlphaZero, MuZero, ChatGPT, Gemini, and AlphaCode can generate program code which we had already done in the 1970s and early 80s.

So my simple AI program has become like a compiler whose outputs are data compression programs but having learned some compression algorithms may need no program C/C++ source code from you.

Or was it my "zpaq.exe" from 2008-2010 that also emit/output better or best "data compression programs" which I named "nanozip.exe/nz.exe" or "zstd.exe"?

I knew there was a "zpaq.exe" in the 1970s and early 80s. The said program may had been brought to me sometime in 2008-2010 to look at it or optimize it based on C/C++ code optimizations I learned that time.

I can calculate or predict my simple AI program's average compression performance on most files but would not give away the formula, especially when it has ability to find for higher compression and faster code at runtime and when it uses advanced encoding/decoding algorithms already.

And then, to the user's surprise or amazement, the CLI program can actually output a Win32 app!

A data compression program and its Win32 app version!

And then after, when I started coding numerical analysis and numerical algorithms, my computer programs were very fast and were "quantum-like" or ideal for calculations that need quantum chips, quantum processors and quantum algorithms. They can be "single-thread" programs only.

I was remembering my pioneering work in data compression, AI and quantum computing in the 1970s to early 80s, and tried to do some actual programming when I hit upon the idea of what I now call the "RDC2 algorithm" and found solutions to it and "hash compression". I knew I own/co-own the Tech Giants Microsoft, Google, Apple, Facebook, IBM, Intel, AMD, TSMC, Samsung, Nvidia etc.

My "nanozip" compression program of 2008-2010 (not the one from us in 1970s) can emit a data compression program exactly like "zstd.exe" byte by byte, or same compression output. Same "zstd.exe" algorithm in nanozip.exe's array of algorithms in its evolved solutions to data compression.

Or, as I've said, they're the same one program: nanozip.exe, zpaq.exe and zstd.exe from me which are simple "generative AI chat programs" and "polymorphic". Just re-coded to be "compression-only". 

If I remember it right, perhaps in some of my later updates to my simple AI program you can actually control the compression efficiency of the program it outputs in the command-line and not through the chat, like if the AI program is named "zpaq", then the command is "zpaq <Huffman/LZ77/arithmetic coding>" to emit a <Huffman/LZ77/arithmetic coding> EXE program, and "zpaq zstd.exe 85%" to emit or create a compression program or EXE file named "ztsd.exe" of 85% average compression ratios in efficiency, or "zpaq zstd.exe 90%" or "zpaq nanozip.exe 90%" or "zpaq nz.exe 90%" or "zpaq nz.exe 95%". Although most probably in the wide distribution of my simple AI program, users can do that only via the chat interface.

It is entertaining and exciting through the chat...?

They were like computer games which would not easily bring the user to a better or perfect "score" or "prize" wherein it outputs a high-compression and very fast data compression program according to the user's knowledge of data structures, algorithms analysis, data compression algorithms, and computer science in general. How efficient the data compression program it outputs is like the "score" or "grade" of the user but it can output "zstd-quality" or "nanozip-quality" programs or even better in compression ratios and speed.

Artificial intelligence (AI) scientists and experts can teach my AI program "better English" statements although I may had done that already in my updates but it can revert back or others can revert it back to its "non-perfect English language" ability, or may not distribute the better updates to my AI program.

If its English language ability is improved or perfected via "supervised or non-supervised learning", then it can learn and learn more encyclopedia contents (science and mathematics) and due to more computer memory or RAM (it's a highly-compressed data structure actually) and big files it can read. I reckoned it can do that in just hours according to the speed of computers running it like workstations, supercomputers, high-performance computers (HPCs), and quantum computers. Yes, most probably.

I developed my zstd.exes and nanozip.exe and nz.exe algorithms and implementations when I recalled and solved the "RDC2 algorithm" and "Zstd of Facebook" and solved "hash compression".

My advanced data compression algorithms and program implementations development period is during 2008-2010 (many emails sent) which clearly preceded the official release of Zstd, Gemini, Gemma, and ChatGPT and Windows10 which have become top software products now. I probably improved in the end my AI chat program's English conversation ability, continuing throughout its Win32 and Win64 development while emailing mostly the PKZIP team that time, and then the PeaZip and WinRAR development teams again.

So my simple AI chat program is an "engine" to optimize C/C++ source code and can "generate" actual data compression programs of very high-compression ratios performance and very fast speeds.

New Breakthrough Data Compression Algorithms in 2008-2009

I have a Podcast on YouTube. Here it is:

New Breakthrough Data Compression Algorithms in 2008-2009 

On each of the podcasts, I wrote comments. I asked Gemini AI to write a blog article out of them, to just re-arrange my sentences and statements in those comments:

"The following are my comments on my YouTube podcasts on my advanced and breakthrough compression algorithms.

Edit them for a post on Blogger or YouTube. Correct spellings and grammars, then collate the sentences and paragraphs into cohesive and coherent statements and paragraphs, into one cohesive whole, like a good book or well-written novel. Also arrange them chronologically and delete unnecessary repetitions of sentences and thoughts. 

Use my same sentences and statements. Do not try to add new sentences, Gemini. Same sentences from me but you can omit unnecessary sentences. Correct spellings too and grammar. Same contents in my DRAFT/draft, like 98% do not add sentences of your own."

Gemini's Response:

Here is your text, edited to correct spelling and grammar while removing repetitive thoughts, organized into a chronological narrative using only your original sentences and statements.

Text for Blogger blog post:

I had discovered or invented and implemented in 2008–2009 advanced data compression algorithms that attain on average 75%+ compression ratios, and even 85%+ and 94%+ compression ratios, as shown in emails sent in 2008 and then 2009.

Maybe I already had a breakthrough data compression algorithm in 1999 to 2003 (Quezon City), or 2005 (here at SJDM, Bulacan), and 2006 or 2007, and then 2008 to 2009 (at Antipolo City). Years 2008 to 2009 were 17–18 years ago, when I was 33 to 34 years old. Made possible by a new internet connection at home and a new computer in late 2007 (October/December), you could email anytime, even every hour. Meaning, so many people and groups were emailed, providing more evidence of the truth of the existence of the advanced breakthrough compression algorithms.

My breakthrough data compression algorithms in 2000–2003 and 2005–2006 I am not sure are saved or intact up to now. I had deleted them in my HDDs, unless they were recovered illegally via UNDELETE/RECOVER programs by authorities who somehow perhaps entered my home without a search warrant. Only the compression programs emailed in 2008 to 2009 we can say are real and evidence of the advanced data compression algorithms' existence—because they were sent to many individuals and groups mostly with C source codes and a few C++ source codes. Or unless there were copies of them in public internet cafes' HDDs that time when I was still paying for internet rental services in public internet shops or internet cafes.

I emailed NASA, two or more NASA email boxes, maintainers of NASA mailboxes, the tech companies Microsoft, Google, Facebook, IBM, Intel, etc., the House of Representatives of the Philippines, the Senate of the Philippines, eBay, the PeaZip development team, the AVG Antivirus team, and the computer data storage companies. In 2009, the purpose of my emails again was to message a few corrections and solutions and improvements, how to decrypt my encrypted files, and also to state my concerns about the consequences and implications of those discoveries.

There was really a compressed ZIP file named "ebay.zip" attached in my emails to eBay company. Maybe I also uploaded it on the Internet Archive (archive.org). I used a different username perhaps and the password I somehow forgot. One upload on the Internet Archive I couldn't delete anymore. I wonder now if the files were encrypted. I think YES, standard Zip file encryption only, or possibly also with my own encryption. I have sent encrypted ZIP files in my emails in 2008–2012, but me no encryption expert. If I have mentioned Tensor Processing Units (TPUs) of Google and some cryptanalysts still didn't know about TPUs when I was "emailing and emailing", then they must have been surprised about TPUs.

I tried my best in making my own encryption secure, and was aware of quantum chips and quantum processors speed. Perhaps I was considering Shor's algorithm and Grover's algorithm speeds that even if they are used in cracking my encrypted files, it would still be hard decryption. But of course, government scientists, cryptographers, and cryptanalysts have their own very fast microprocessors, powerful workstations, or they have access to government or private supercomputers, quantum chips, and quantum processors by that time in the 2000s, or even cloud networks or "The Cloud" which I also mentioned in my emails.

In the end, my breakthrough data compression algorithms and programs were highly optimized already and the EXE files were very small in size. They were "recursive compression algorithms", "hash data compression algorithms", and then "integer compression algorithms"—solved and fully implemented. I think it was in 2007 that I solved and fully implemented the RDC2 compression algorithm, with improvements and upgrades. The data structure I used for the "Big Integers (BigInts)" or Arbitrary-Precision Integers in implementing the RDC2 algorithm is a "singly-linked list". I created a good enough simple library for BigInts.

I used the Turbo C compiler in solving and implementing the RDC2 algorithm, so I must have used first the "long" 32-bit integers and "long long" integers and "double" for the 64-bit integers, along with implementations using the "BigInts" integer library. RDC2 solution is largely based on "sorting frequencies" or any transforms on the "frequency table"; its compression performance is due to "context" and "finite state entropy" (FSE). I intended to fully implement and improve on my RDC2 programs in just one week but it became more than seven days. Then you add some "hash compression solutions" to a few RDC2 compression programs in the "RDC2 solutions ZIP file". And then, I revealed it later as "hash data compression" algorithm (solved) with many separate compression programs, improvements and updates, in another ZIP file.

I also had an "lzhash.exe" sent to them or maybe it was "lzhash2.exe", and an emailed "zhash.exe" program—both implementations of "hash data compression" where the outputs are only just integers or hashes of hash functions. I remember my "lzhash.c" before, in 2008–2009, was a data compression program perhaps with "lzhashd.c" as decompressor. I created the "lzhash.c, lzhash.h" include library files to hide the fact that there was a previous "lzhash.c" compressor with decompressor in another email, or perhaps it was a single-file "lzhash.c" compressor/decompressor (codec).

And then I tried understanding and implementing fractal compression in which I succeeded again. Did I have a "fractal compression" program in 2008–2009? Yes, I think so. Most possibly, iirc, gave them one fractal data compression program. Three fractal compression programs eventually, in 2008 or 2009, one correction to the one sent first and another one improved and higher compression version. Year 2010 na siguro the correction to the first fractal compression program sent to NASA and with a new improved fractal compression program version. I think I also created a fractal version of my "hash data compression" programs, the ones whose outputs are only hash integers or integer hashes for the input blocks; so it's a "fractal hash" compression program. They're almost the same in compression ratios performance, iirc.

I created and developed a data compression program named "zstd.exe". I sent a "zstd.exe" to Facebook, which I possibly also sent to NSA or CIA (as a second party) without telling Facebook, but they knew it was "for Facebook". Most possibly I sent a "corrected bzip2 version" to the Bzip team and a bzip3 version with advanced compression performance similar to Zstd and then a bzip3 version with a "hidden" compress function in the "command-line options" in its EXE file. Very simply, I did not add that option in the command-line, but it's also possible that it is there in the command-line options with its unusual, suspicious, or lack of usage statements or displays on the computer screen for the curiosity of the programmers and testers.

My other data compression programs were also encrypted, not just Nanozip. Nanozip (nz.exe) in 2008–2009 or maybe to 2010, I recalled the nanozip command-line program or saw it on the Internet. The nanozip program cannot be easily decrypted or deciphered in a Hex editor. It was encrypted, but yes perhaps easy to expert programmers who usually or routinely decompile binary EXE files. If that was my Nanozip (nz.exe), it is slow due to encryption at the start and decryption after compressing the input file, and then encrypting and decrypting in between the compression process. Then I have to delete the encryption and decryption pair, so that my nz.exe would be very fast. One compress function in my Nanozip program is "same as Zstd compression", meaning, the output compressed file is exactly the same.

In 2008 to 2009, my Win32 GUI app which features Algorithm FGK I think still used "gtbitio.c", though I had a better and way faster BITIO library named differently, also included in few Win32 GUI apps or archives sent to just a very few tech companies or NASA/NSA. The Algorithm FGK source codes were possibly hidden via steganography or encrypted within the Win32 GUI program, I think, so when they decrypt it, it is just C/C++ source codes but my authorship. In two days perhaps, I emailed them that it's just source code of a basic compression algorithm so that they may not have to decrypt it. The "encryption" might even be "compression" or a simple image bitmap. Maybe one to two source codes and to be decrypted by them; if three, then it was an advanced algorithm. The "C/C++ source codes" appear in a "black and white" bitmapped image of a text file which were perhaps in "strike-through" fonts, iirc, and even added with some "text degradation functions" in the image to test the visual acuity and recognition of humans and OCR systems and visual recognition systems. 

I had my own simple AI program possibly named "zpaq" that time (or also "zstd.exe"/"nanozip") which emits "data compression programs" that are faster, so they brought me the original "zpaq.cpp" to optimize. In the end, one of the outputs of my AI program I named "zpaq.exe" (compression program-only) and sent to Facebook as well as its new versions and upgrades. Also a "nanozip.exe" and then to "nz.exe". Later, I added that option to "optimize" some C/C++ code that you show it; or if the C/C++ code you supplied is usable and faster, it is simply "learned" by my AI program. Yes, I think later I made it learn C++, and even simple machine code or assembly code via "inline assembly functions in C/C++".

Maybe I will now in 2026 develop a Win32 GUI app for my static and dynamic Huffman compression programs, my Algorithm FGK implementation, and then also for my LZ77 and LZW and PPP/lzpgt compression programs in one GUI Windows program.