Mark Stephen Sobkow's Code Factory - The Roots of the Fractal   

Nowadays some companies are preaching the LLM as the latest "silver bullet" solution to the dilemmas of programmer productivity and errors, but they won't succeed, because LLMs aren't actually intelligent. They don't know what you're talking about or understand it. What they do is apply some incredibly complex statistics to guess what you probably mean based on other people who've asked similar things over the years of data that were scraped and ground into the LLM's statistical data core. And LLMs are proving to be just as error prone as their human counterparts, just in different but equally frustrating and problematic ways.

There has to be a solution, and I think I've found it by merging together aspects of compiler technology, data analysis logic, expert systems approaches, functional programming method search techniques, and decades upon decades of experience in front of a keyboard to come up with something "new." New to the world, at least, but not new to me - this technology began existence around Java 1.6... it got up to 2.13 so far, and now 3.1 is being worked on, which refreshes the whole underlying approach to take advantage of much newer programming technologies.

In a very big way, when you use 3.1 after it's initial release, you're using me to write code for you! You're using me to precisely follow your design to produce a code fractal - an expression of your model so predictable that it is fractal-like in it's nature, but without the vagaries and inconsistencies of LLM-generated code that try to do the same thing through mere statistics.

The key concepts that come together are functional method searching combined with a reverse LALR grammar specification done through XML documents and a sufficiently robust Business Application Model that provides the "design" the expansions follow. For performance reasons, many aspects are implemented in "xxxcust" packages of custom code that enhance the manufactured fractal code.

It's worth noting that each earlier generation of MSS Code Factory wrote the fractal code for the generation to follow. What you are about to witness in a few short months with 3.1 is the culmination of 14 generations of code evolution and design over nearly 30 years time.

Despite it's age, I don't think the idea is obsolete. I think the core tenet of ISO9000 feedback is crucial to solving the software reliability problem, and I think this tool is going to prove very useful in finally providing a means of implementing such feedback corrections.