Remove basalt reference; clarify org has no public artifacts yet

README.md

@@ -18,18 +18,10 @@ it, and keeps documentation aligned with the actual state of the
 controls. The people behind it are engineers and compliance
 practitioners who got tired of treating audits as ceremony.
 
-This Hugging Face org is the open side of that work. We publish the
-ML pieces we can release: model adapters today, datasets later when
-licensing allows.
-
-## What's here
-
-[basalt](https://huggingface.co/deepfathom/basalt) is the only public
-artifact right now. It is a LoRA adapter for CMMC, NIST SP 800-171,
-and DFARS practitioner Q&A, layered on `ibm-granite/granite-4.1-8b`.
-The repo is private through M-Train v1 validation. Public release
-depends on eval results and a license review of the base model and
-corpus terms.
+This Hugging Face org is a placeholder for the open side of that
+work. We don't have public artifacts here today. If licensing and
+eval results allow, we may publish model adapters in the future, and
+datasets later still.
 
 ## How we build the models
 
@@ -42,10 +34,11 @@ the data version, seed, hyperparameters, and judge configuration so a
 later release can be checked against the artifact it claims to
 supersede.
 
-We publish LoRA adapters rather than merged weights. PEFT and vLLM
-load the base and the adapter together at runtime. The result is the
-latest upstream base, our specialty layer composed on top of it, and
-the option to swap each layer independently.
+When we publish models, we ship LoRA adapters rather than merged
+weights. PEFT and vLLM load the base and the adapter together at
+runtime. The result is the latest upstream base, our specialty layer
+composed on top of it, and the option to swap each layer
+independently.
 
 ## Principles
 
@@ -63,5 +56,4 @@ want shared evidence instead of screenshots.
 
 ## Reach
 
-[deepfathom.ai](https://www.deepfathom.ai). Use model repo
-discussions for technical questions.
+[deepfathom.ai](https://www.deepfathom.ai).