@Tim Fairley ICM stands for Interpreted Context Methodology, it's a system of building orchestration and routing applying the basic principle of sofware engineerig. It's basically folders and text files that tells any ai agent what you want to do and some guardrails it should work within. With the attched workspace as sample, It's a set of plain text files (identity, rules, examples, reference docs) that tell the AI exactly how to review a project, instead of one long prompt trying to do everything at once with a good chance of the ai later hallucinating and going offrails. For Titan, the workspace has a rules file with 17 numbered rules, a fixed list of flag types and severity levels, and an output schema the AI has to follow every time. So the review comes out the same structure every time, not different depending on the day or ai model used. One point to mention though: it's not analyzing drawings. It's reviewing text documents, the supervisor's brief, the geotech report, and trade scope notes, and checking them against each other for missing info, scope overlaps, and schedule risk. The reason this beats a typical AI tool: most tools are a black box, you don't know why it flagged something. Here every flag has to quote the source document. You can open the rules file and see the exact logic. No orchestration software, no hidden steps, just files you can read and edit yourself. check out this video, he explains the icm framework better https://youtu.be/KPVaUuBkPz8?si=U-JDygQMU9H4sj0X