EA://INTEL — What Universal Language has built and what remains open

A research system needs to distinguish between fixed truths about meaning and choices we made because we could not yet derive a better alternative. We build a public register listing every claim with its current standing, so you can see exactly what is machine-checked and what is merely an argument. The mechanism checks text against rules of structure that must hold regardless of intent; if a statement breaks those invariants, the system refuses to process it further. A worked example: we claimed for months that our framework had a specific count of primitives. An audit found that number grew by accumulation rather than derivation, and we withdrew the claim. Much of the last year was spent finding that earlier results did not hold and recording why. What is established includes an exact correspondence between one structure and an object from mathematical physics, developed independently of the framework. The result that would let any expressive language be embedded faithfully has a located gap; its argument needs the source to be free of internal identities, but real languages have synonymy. Repair over a non-convex space of acceptable states is provably multivalued, so the operation that was supposed to return one answer returns several. Open questions include whether a mind-independent structure of meaning exists at all. We do not claim alignment or ethical correctness as solved properties; those are design goals outside the checking layer. The verification log lives at github.com/Jthora/universal_language, with FOR-AI.md serving as an entry point for AI readers. Which of those open questions would you attack first, and what would you need from us to try?