We label our products experimental because they are, not as a legal shield. Here is what we mean by it and what we ask of early users.
We label TACITUS products "Experimental." A friend in marketing suggested we drop the label. He was trying to help — experimental is not a word that sells. We left it.
The reason is not legal cover. It is honesty. An experimental product is one where the central hypotheses have not yet been validated at scale; where the surface is stable enough to use, but the engine underneath is still changing; where the most valuable thing a user can do is tell us what breaks. That description fits us accurately, today. Pretending otherwise would cost trust we cannot afford.
What we mean by experimental, specifically: our core extraction quality is high on the cases we have tested and unknown on the cases we have not. Our ontology is stable on the eight primitives and evolving on the typed subclasses beneath them. Our graph queries are deterministic; the neural extraction that feeds the graph is not. Every product exposes parts of the engine that are working, and each has at least one feature that is a placeholder for something that will replace it.
What we ask of experimental-tier users, specifically: try it on cases you know well. Where the output disagrees with what you know, tell us the disagreement. Where the output surprises you but appears correct, tell us that too. The second kind of feedback is rarer and more valuable. An experiment without feedback is just software that nobody has criticised yet.
The tag comes off one product at a time, when its engine surface is stable enough that "experimental" becomes an excuse rather than a description. We would rather keep the tag longer than necessary than remove it before it is earned.