diff --git a/notes/curr.md b/notes/curr.md
new file mode 100644
index 000000000..bb826a77c
--- /dev/null
+++ b/notes/curr.md
@@ -0,0 +1,12 @@
+v0.6.3
+
+- [x] better watch regen message
+- [x] disjunct ordered models
+- [x] --watch-templates & clear function
+- [ ] only do adhoc things if needed
+- [ ] --as-module flag (also --create-module) or --module-* flags? (including subgens from -G)
+- [ ] list / info flags (for available gens, or maybe just part of help?)
+- [ ] record demo
+  - [ ] adhoc to module, show new mode & flags
+	- [ ] generating types across languages (also as example for docs)
+
diff --git a/notes/datamodel.md b/notes/datamodel.md
new file mode 100644
index 000000000..ae02767a1
--- /dev/null
+++ b/notes/datamodel.md
@@ -0,0 +1,27 @@
+# datamodel
+
+next version (0.6.4) datamodel upgrades
+
+- [ ] load datamodel history into generators
+- [ ] diff3 tests, can we get an easy fix in? (timebox this)
+- [ ] add thema mode & support, can we bidirectionally convert between formats?
+- [ ] support using raw cue values with constraints as valid input and markup for the code engine
+- [ ] support translation methods between models / versions to be defined by user, parse refs in cue to something consumable by code gen
+- [ ] we're more focused on datamodel++ (the extra config/metadata needed for codegen of applications), than using history at runtime (?) (tradeoffs)
+- [ ] implement views / relns into info system
+- [ ] mock (faux) data (https://www.getsynth.com/docs/getting_started/core-concepts)
+- [ ] maybe we just need to add lacuna to our schemas? do they represent the cases where we cannot automate or the user wants to be explicit?
+- [ ] support with @rename() / @move() for fields (& models?), use `><` in the diff algo, helper #Defs for users in their CUE
+
+can we use the concepts from hof dm and thema
+to support code gen mods interoperability?
+
+first, generators need an explicit schema / datamodel version, indep of their release version
+this is like k8s resources having apiVersion or docker-compose / OpenAPI spec version when writing
+
+examples:
+
+- how to make client cli tools support newer API versions, and the client-side filler data
+- even if the server is not Go, can it send CUE to a Go CLI such that these can be pushed dynamically
+- what about a single CLI talking to many copies of the API, like kubectl -> clusters
+
diff --git a/notes/prototypes.md b/notes/prototypes.md
index a7724680c..e72f1af91 100644
--- a/notes/prototypes.md
+++ b/notes/prototypes.md
@@ -1,14 +1,3 @@
-v0.6.3
-
-- [x] better watch regen message
-- [x] disjunct ordered models
-- [x] --watch-templates & clear function
-- [ ] --as-module flag (also --create-module) or --module-* flags?
-- [ ] list / info flags (for available gens, or maybe just part of help?)
-- [ ] record demo
-
----
-
 Priorities:
 
 - [ ] datamodel history example
@@ -16,23 +5,7 @@ Priorities:
 	- [ ] server / client versioning support
 - [ ] deploy API updates to consumers
 
----
-
-next version, datamodel upgrades
-
-- [ ] load datamodel history into generators
-- [ ] diff3 tests, can we get an easy fix in? (timebox this)
-- [ ] add thema mode & support, can we bidirectionally convert between formats?
-- [ ] support using raw cue values with constraints as valid input and markup for the code engine
-- [ ] support translation methods between models / versions to be defined by user, parse refs in cue to something consumable by code gen
-- [ ] we're more focused on datamodel++ (the extra config/metadata needed for codegen of applications), than using history at runtime (?) (tradeoffs)
-- [ ] implement views / relns into info system
-- [ ] mock (faux) data (https://www.getsynth.com/docs/getting_started/core-concepts)
-- [ ] maybe we just need to add lacuna to our schemas? do they represent the cases where we cannot automate or the user wants to be explicit?
-- [ ] support with @rename() / @move() for fields (& models?), use `><` in the diff algo, helper #Defs for users in their CUE
-
-can we use the concepts from hof dm and thema
-to support code gen mods interoperability?
+- should we focus on tighter integrations with golang?
+  - if (for) most users (who) don't care what tech is used?
+  - via CUE validation, transformations, hof/flow, custom logic
 
-first, generators need an explicit schema / datamodel version, indep of their release version
-this is like k8s resources having apiVersion or docker-compose / OpenAPI spec version when writing