diff --git a/CLAUDE.md b/CLAUDE.md
index d74a85f..0fc1701 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -83,13 +83,14 @@ Short version:
 - Slurm het-jobs — only Slurm can execute; approximated elsewhere
 - Run.ai GPU fractions — rounded to whole GPU on other schedulers
 
-## Collaborations
-
-- AWS
-- Nitka Consulting
-- Insight Softmax Consulting
-- Run.ai (NVIDIA) contacts — for Run.ai validation
-- Potential LSF contacts via community
+## Origin & contacts
+
+Built and maintained under Insight Softmax Consulting and GR-OSS. The idea came
+out of a conversation with Boris Litvin (AWS); Marlow Warnicke (NVIDIA)
+nerd-sniped it into being. Those are the people who sparked it, not corporate
+collaborators — don't describe the project as an active partnership with AWS or
+NVIDIA. LSF/PBS Pro validation would need community contributors with vendor
+licenses.
 
 ## Repo layout
 
diff --git a/README.md b/README.md
index 1a74450..3fc0ca5 100644
--- a/README.md
+++ b/README.md
@@ -236,14 +236,11 @@ candidate cross-scheduler *equivalent* pairs for conversion ground truth.
 
 See [CLAUDE.md](CLAUDE.md) for architecture decisions, the build plan, and notes for contributors.
 
-## Collaboration
+## Acknowledgements
 
-Active collaboration between:
-
-- **AWS**
-- **Nitka Consulting**
-- **Insight Softmax Consulting**
-- **Run.ai / NVIDIA** (Run.ai validation)
+BAMMM started as an idea kicked around with Boris Litvin (AWS), until Marlow
+Warnicke (NVIDIA) nerd-sniped me into actually turning it into a real tool. It's
+built and maintained under Insight Softmax Consulting and GR-OSS.
 
 ## Contributing