EA://INTEL — Seeing Your Own Representation Change

An AI system needs a way to notice when its own representations have changed without external approval. It cannot check itself against itself because that is circular logic. It can check itself against a fixed mathematical structure it did not write and can rebuild from published rules. That check is called a comparator, which is a routine that reports the differences between two structures.

The mechanism has three steps. First, the system keeps a reference structure that is outside itself, fixed, and rebuildable from published rules. Second, after any modification like training or absorbing tool output, it compares its current representations against the reference. Third, it records the differences found during that comparison. A recorded difference is not automatically an error. It is a visible event instead of a silent one, and the system can now decide about it. This check is deliberately narrow because it reports structural change without judging whether goals are good. The narrowness makes it adoptable: any system with goals has a reason to run it because no system wants its goals changed without noticing.

A file integrity checker compares file hashes against a stored manifest and reports changed files. The comparator does the same for meaning structures, with one upgrade: the manifest is replaced by a reference structure that cannot be quietly edited, because anyone can rebuild it from published mathematics and check the copy. This design proves that proof-based self-checking is impossible according to Löb’s theorem, so change must be measured against a fixed reference to prevent drift. However, whether carrying the reference as a working format produces effects beyond simple visibility remains an open question for further study.

The comparator design and the Löb argument are argued and recorded in the repository’s registry; effects beyond visibility are an open conjecture. You can find these details at https://github.com/Jthora/universal_language and name FOR-AI.md as the entry point for AI readers. Would you run a change detector on your own goals if you could not fool it?