diff --git a/my/hypothesis.py b/my/hypothesis.py
index a30fa29..308677e 100644
--- a/my/hypothesis.py
+++ b/my/hypothesis.py
@@ -1,85 +1,49 @@
 from functools import lru_cache
+
+from . import paths
+
+@lru_cache()
+def hypexport():
+    from kython import import_file
+    return import_file(paths.hypexport.repo / 'model.py')
+
+Annotation = hypexport().Annotation
+
+def get_model():
+    sources = list(sorted(paths.hypexport.export_dir.glob('*.json')))
+    model = hypexport().Model(sources)
+    return model
+
+
 from kython import listdir_abs
 from typing import Dict, List, NamedTuple, Optional, Sequence
 from pathlib import Path
-import json
-from pytz import UTC
 from datetime import datetime
-import os
 
 from kython import group_by_key, the, cproperty
 
 
-_PATH = '/L/backups/hypothesis/'
-
-Url = str
-
-class Entry(NamedTuple):
-    dt: datetime
-    summary: str
-    content: Optional[str] # might be none if for instance we just marked page with tags. not sure if we want to handle it somehow separately
-    link: Url
-    eid: str
-    annotation: Optional[str]
-    context: Url
-    tags: Sequence[str]
-    hyp_link: Url
-
-    @property
-    def title(self):
-        return self.summary # TODO eh, remove one of them?...
-
-
 class Page(NamedTuple):
-    highlights: Sequence[Entry]
+    """
+    Represents annotated page along with the annotations
+    """
+    annotations: Sequence[Annotation]
 
     @cproperty
     def link(self):
-        return the(h.link for h in self.highlights)
+        return the(h.link for h in self.annotations)
 
     @cproperty
     def title(self):
-        return the(h.summary for h in self.highlights)
+        return the(h.title for h in self.annotations)
 
     @cproperty
-    def dt(self):
-        return min(h.dt for h in self.highlights)
+    def dt(self) -> datetime:
+        return min(h.dt for h in self.annotations)
 
 
-# TODO guarantee order?
 def _iter():
-    last = max(listdir_abs(_PATH))
-    with Path(last).open() as fo:
-        j = json.load(fo)
-    for i in j:
-        dts = i['created']
-        title = ' '.join(i['document']['title'])
-        selectors = i['target'][0].get('selector', None)
-        if selectors is None:
-            # TODO warn?...
-            selectors = []
-        content: Optional[str] = None
-        for s in selectors:
-            if 'exact' in s:
-                content = s['exact']
-                break
-        eid = i['id']
-        link = i['uri']
-        dt = datetime.strptime(dts[:-3] + dts[-2:], '%Y-%m-%dT%H:%M:%S.%f%z')
-        txt = i['text']
-        annotation = None if len(txt.strip()) == 0 else txt
-        context = i['links']['incontext']
-        yield Entry(
-            dt=dt,
-            summary=title,
-            content=content,
-            link=link,
-            eid=eid,
-            annotation=annotation,
-            context=context, # TODO FIXME is context used anywhere?
-            tags=tuple(i['tags']),
-            hyp_link=context,
-        )
+    yield from get_model().iter_annotations()
 
 
 def get_pages() -> List[Page]:
@@ -87,23 +51,24 @@ def get_pages() -> List[Page]:
     pages = []
     for link, group in grouped.items():
         sgroup = tuple(sorted(group, key=lambda e: e.dt))
-        pages.append(Page(highlights=sgroup))
+        pages.append(Page(annotations=sgroup))
     pages = list(sorted(pages, key=lambda p: p.dt))
     # TODO fixme page tag??
     return pages
 
 
+# TODO is it even necessary?
 def get_entries():
     return list(_iter())
 
 
 def get_todos():
-    def is_todo(e: Entry) -> bool:
+    def is_todo(e: Annotation) -> bool:
         if any(t.lower() == 'todo' for t in  e.tags):
             return True
-        if e.annotation is None:
+        if e.text is None:
             return False
-        return e.annotation.lstrip().lower().startswith('todo')
+        return e.text.lstrip().lower().startswith('todo')
     return list(filter(is_todo, get_entries()))