Lots of examples of poorly written code done by AI out there. Here's a solution: https://bluelabs.com/fixing-the-leaning-tower-of-ai/
Since writing the whitepaper (https://doi.org/10.5281/zenodo.17968796), I've written several more applications using this method and trained a few orgs in the process. The results are encouraging, these apps are in production and are being measured with proper observability. You can see some on my github page (https://github.com/pacepace).




















