diff --git a/lamindb/__init__.py b/lamindb/__init__.py
index 70669d9ee..17ea35b0c 100644
--- a/lamindb/__init__.py
+++ b/lamindb/__init__.py
@@ -182,6 +182,7 @@
     DB,
 )
 from .models.save import save
+from .models.sqlrecord import UNSET, Unset
 from . import core
 from . import integrations
 from . import curators
@@ -230,6 +231,8 @@
     "examples",
     "errors",
     "setup",
+    # sentinels
+    "Unset",
     # low-level functionality
     "base",
     "core",
diff --git a/lamindb/models/feature.py b/lamindb/models/feature.py
index e5c86d1e1..869e878b1 100644
--- a/lamindb/models/feature.py
+++ b/lamindb/models/feature.py
@@ -51,6 +51,7 @@
     Space,
     SQLRecord,
     UNSET,
+    Unset,
     _get_record_kwargs,
     pop_space_branch_kwargs,
 )
@@ -671,7 +672,7 @@ def process_init_feature_param(args, kwargs):
     name: str = kwargs.pop("name", None)
     dtype: SimpleDtype | SimpleDtypeStr | str | None = kwargs.pop("dtype", None)
     is_type: bool = kwargs.pop("is_type", False)
-    type_: Feature | str | None = kwargs.pop("type", UNSET)
+    type_: Feature | str | None | Unset = kwargs.pop("type", UNSET)
     description: str | None = kwargs.pop("description", None)
     space_branch_kwargs = pop_space_branch_kwargs(kwargs)
     _skip_validation = kwargs.pop("_skip_validation", False)
@@ -1221,7 +1222,7 @@ def __init__(
         | Registry
         | list[Registry]
         | FieldAttr,
-        type: Feature | None = None,
+        type: Feature | None | Unset = UNSET,
         is_type: bool = False,
         unit: str | None = None,
         description: str | None = None,
diff --git a/lamindb/models/project.py b/lamindb/models/project.py
index 56cc24226..050949ebb 100644
--- a/lamindb/models/project.py
+++ b/lamindb/models/project.py
@@ -28,7 +28,19 @@
 from .record import Record
 from .run import Run, TracksRun, TracksUpdates, User
 from .schema import Schema
-from .sqlrecord import BaseSQLRecord, HasType, IsLink, SQLRecord, ValidateFields
+from lamindb.errors import FieldValidationError
+
+from .sqlrecord import (
+    UNSET,
+    Unset,
+    BaseSQLRecord,
+    HasType,
+    IsLink,
+    SQLRecord,
+    ValidateFields,
+    _get_record_kwargs,
+    pop_space_branch_kwargs,
+)
 from .transform import Transform
 from .ulabel import ULabel
 
@@ -217,7 +229,7 @@ class Meta(SQLRecord.Meta, TracksRun.Meta, TracksUpdates.Meta):
     def __init__(
         self,
         name: str,
-        type: Reference | None = None,
+        type: Reference | None | Unset = UNSET,
         is_type: bool = False,
         abbr: str | None = None,
         url: str | None = None,
@@ -237,7 +249,44 @@ def __init__(
     ): ...
 
     def __init__(self, *args, **kwargs):
-        super().__init__(*args, **kwargs)
+        if len(args) == len(self._meta.concrete_fields):
+            super().__init__(*args, **kwargs)
+            return None
+        if len(args) > 0:
+            raise ValueError("Only keyword args allowed")
+        name: str = kwargs.pop("name", None)
+        type: Reference | None | Unset = kwargs.pop("type", UNSET)
+        is_type: bool = kwargs.pop("is_type", False)
+        abbr: str | None = kwargs.pop("abbr", None)
+        url: str | None = kwargs.pop("url", None)
+        pubmed_id: int | None = kwargs.pop("pubmed_id", None)
+        doi: str | None = kwargs.pop("doi", None)
+        description: str | None = kwargs.pop("description", None)
+        text: str | None = kwargs.pop("text", None)
+        date: DateType | None = kwargs.pop("date", None)
+        space_branch_kwargs = pop_space_branch_kwargs(kwargs)
+        _skip_validation = kwargs.pop("_skip_validation", False)
+        _aux = kwargs.pop("_aux", None)
+        if len(kwargs) > 0:
+            valid_keywords = ", ".join([val[0] for val in _get_record_kwargs(Reference)])
+            raise FieldValidationError(
+                f"Only {valid_keywords} are valid keyword arguments"
+            )
+        super().__init__(
+            name=name,
+            type=type,
+            is_type=is_type,
+            abbr=abbr,
+            url=url,
+            pubmed_id=pubmed_id,
+            doi=doi,
+            description=description,
+            text=text,
+            date=date,
+            _skip_validation=_skip_validation,
+            _aux=_aux,
+            **space_branch_kwargs,
+        )
 
     def query_references(self) -> QuerySet:
         """Query references of sub types.
@@ -449,7 +498,7 @@ class Meta(SQLRecord.Meta, TracksRun.Meta, TracksUpdates.Meta):
     def __init__(
         self,
         name: str,
-        type: Project | None = None,
+        type: Project | None | Unset = UNSET,
         is_type: bool = False,
         abbr: str | None = None,
         url: str | None = None,
@@ -466,7 +515,38 @@ def __init__(
     ): ...
 
     def __init__(self, *args, **kwargs):
-        super().__init__(*args, **kwargs)
+        if len(args) == len(self._meta.concrete_fields):
+            super().__init__(*args, **kwargs)
+            return None
+        if len(args) > 0:
+            raise ValueError("Only keyword args allowed")
+        name: str = kwargs.pop("name", None)
+        type: Project | None | Unset = kwargs.pop("type", UNSET)
+        is_type: bool = kwargs.pop("is_type", False)
+        abbr: str | None = kwargs.pop("abbr", None)
+        url: str | None = kwargs.pop("url", None)
+        start_date: DateType | None = kwargs.pop("start_date", None)
+        end_date: DateType | None = kwargs.pop("end_date", None)
+        space_branch_kwargs = pop_space_branch_kwargs(kwargs)
+        _skip_validation = kwargs.pop("_skip_validation", False)
+        _aux = kwargs.pop("_aux", None)
+        if len(kwargs) > 0:
+            valid_keywords = ", ".join([val[0] for val in _get_record_kwargs(Project)])
+            raise FieldValidationError(
+                f"Only {valid_keywords} are valid keyword arguments"
+            )
+        super().__init__(
+            name=name,
+            type=type,
+            is_type=is_type,
+            abbr=abbr,
+            url=url,
+            start_date=start_date,
+            end_date=end_date,
+            _skip_validation=_skip_validation,
+            _aux=_aux,
+            **space_branch_kwargs,
+        )
 
     def query_projects(self) -> QuerySet:
         """Query projects of sub types.
diff --git a/lamindb/models/record.py b/lamindb/models/record.py
index 13d50722a..645456376 100644
--- a/lamindb/models/record.py
+++ b/lamindb/models/record.py
@@ -39,6 +39,7 @@
     Space,
     SQLRecord,
     UNSET,
+    Unset,
     _get_record_kwargs,
     pop_space_branch_kwargs,
 )
@@ -873,7 +874,7 @@ class Meta(SQLRecord.Meta, TracksRun.Meta, TracksUpdates.Meta):
     def __init__(
         self,
         name: str | None = None,
-        type: Record | None = None,
+        type: Record | None | Unset = UNSET,
         is_type: bool = False,
         features: dict[str | Feature, Any] | None = None,
         description: str | None = None,
@@ -901,7 +902,7 @@ def __init__(
         if len(args) > 0:
             raise ValueError("Only one non-keyword arg allowed")
         name: str = kwargs.pop("name", None)
-        type: str | None = kwargs.pop("type", UNSET)
+        type: Record | None | Unset = kwargs.pop("type", UNSET)
         is_type: bool = kwargs.pop("is_type", False)
         features: dict[str | Feature, Any] | None = kwargs.pop("features", None)
         description: str | None = kwargs.pop("description", None)
diff --git a/lamindb/models/schema.py b/lamindb/models/schema.py
index 42225807c..4168e24bb 100644
--- a/lamindb/models/schema.py
+++ b/lamindb/models/schema.py
@@ -49,6 +49,7 @@
     Space,
     SQLRecord,
     UNSET,
+    Unset,
     _get_record_kwargs,
     init_self_from_db,
     pop_space_branch_kwargs,
@@ -631,7 +632,7 @@ def __init__(
         name: str | None = None,
         description: str | None = None,
         itype: str | Registry | FieldAttr | None = None,
-        type: Schema | None = None,
+        type: Schema | None | Unset = UNSET,
         is_type: bool = False,
         index: Feature | None = None,
         flexible: bool | None = None,
@@ -673,7 +674,7 @@ def __init__(
         description: str | None = kwargs.pop("description", None)
         itype: str | SQLRecord | DeferredAttribute | None = kwargs.pop("itype", None)
         flexible: bool | None = kwargs.pop("flexible", None)
-        type: Feature | None = kwargs.pop("type", UNSET)
+        type: Feature | None | Unset = kwargs.pop("type", UNSET)
         is_type: bool = kwargs.pop("is_type", False)
         otype: str | None = kwargs.pop("otype", None)
         suffix: str | None = kwargs.pop("suffix", None)
@@ -738,6 +739,7 @@ def __init__(
             coerce=coerce_dtype,
             n_features=n_features,
         )
+        # pop before update_attributes/super so it never reaches Django fields or getattr
         if not features and not slots and not is_type and not itype:
             raise InvalidArgument(
                 "Please pass features or slots or itype or set is_type=True"
@@ -858,7 +860,7 @@ def _validate_kwargs_calculate_hash(
         validated_kwargs = {
             "name": name,
             "description": description,
-            "type": None if type is UNSET else type,
+            "type": type,
             "is_type": is_type,
             "_dtype_str": dtype,
             "otype": otype,
diff --git a/lamindb/models/sqlrecord.py b/lamindb/models/sqlrecord.py
index e7be87647..520cd5936 100644
--- a/lamindb/models/sqlrecord.py
+++ b/lamindb/models/sqlrecord.py
@@ -108,9 +108,24 @@
 
 T = TypeVar("T", bound="SQLRecord")
 
-# Sentinel to distinguish "user didn't pass type=" from "user explicitly passed type=None".
-# Uses object() so identity checks (is) never call __eq__ on model instances.
-UNSET = object()
+class Unset:
+    """Sentinel type to distinguish 'not passed' from explicit ``None``.
+
+    Use ``is`` checks to compare against the singleton :data:`UNSET`.
+    """
+
+    _instance: Unset | None = None
+
+    def __new__(cls) -> Unset:
+        if cls._instance is None:
+            cls._instance = super().__new__(cls)
+        return cls._instance
+
+    def __repr__(self) -> str:
+        return "UNSET"
+
+
+UNSET: Unset = Unset()
 IPYTHON = getattr(builtins, "__IPYTHON__", False)
 UNIQUE_FIELD_NAMES = {
     "root",
@@ -451,7 +466,7 @@ def init_self_from_db(
 
 def update_attributes(record: SQLRecord, attributes: dict[str, str]):
     for key, value in attributes.items():
-        if getattr(record, key) != value and value is not None:
+        if value is not None and value is not UNSET and getattr(record, key) != value:
             if key not in {"uid", "_dtype_str", "otype", "hash"}:
                 logger.warning(f"updated {key} from {getattr(record, key)} to {value}")
                 setattr(record, key, value)
@@ -624,10 +639,18 @@ def suggest_records_with_similar_names(
     # the below needs to be .first() because there might be multiple records with the same
     # name field in case the record is versioned (e.g. for Transform key)
     if isinstance(record, HasType):
-        if kwargs.get("type", None) is None:
+        # "type" is always present in kwargs at this point (Solution A contract)
+        type = kwargs["type"]
+        if type is UNSET:
+            # user passed nothing → search all type contexts
+            # catches typed records with same name, fixes silent dup bug
+            subset = record.__class__.filter()
+        elif type is None:
+            # explicit type=None → root-level dedup (type IS NULL)
             subset = record.__class__.filter(type__isnull=True)
         else:
-            subset = record.__class__.filter(type=kwargs["type"])
+            # specific type object → scoped dedup within that type
+            subset = record.__class__.filter(type=type)
     else:
         subset = record.__class__
     exact_match = subset.filter(**{name_field: kwargs[name_field]}).first()
@@ -1162,10 +1185,6 @@ class Meta:
 
     def __init__(self, *args, **kwargs):
         skip_validation = kwargs.pop("_skip_validation", False)
-        # strip sentinel before validate_fields and Django's Model.__init__ see it
-        # `is` never calls __eq__, so FeaturePredicate objects are safe
-        if isinstance(self, HasType) and kwargs.get("type", UNSET) is UNSET:
-            kwargs.pop("type", None)
         if not args:
 
             def resolve_fk_or_id(field_name: str) -> bool:
@@ -1214,6 +1233,9 @@ def resolve_fk_or_id(field_name: str) -> bool:
                     # the current one), so the record is created on that branch.
                     kwargs["created_on"] = kwargs["branch"]
             if skip_validation:
+                # strip UNSET just before Django sees kwargs — FK descriptors reject non-model values
+                if isinstance(self, HasType) and kwargs["type"] is UNSET:
+                    kwargs.pop("type")
                 super().__init__(**kwargs)
             else:
                 from ..core._settings import settings
@@ -1270,6 +1292,9 @@ def resolve_fk_or_id(field_name: str) -> bool:
                             # track original values after replacing with the existing record
                             self._populate_tracked_fields()
                             return None
+                # strip UNSET just before Django sees kwargs — FK descriptors reject non-model values
+                if isinstance(self, HasType) and kwargs["type"] is UNSET:
+                    kwargs.pop("type")
                 super().__init__(**kwargs)
                 if isinstance(self, ValidateFields):
                     # this will trigger validation against django validators
diff --git a/lamindb/models/ulabel.py b/lamindb/models/ulabel.py
index 4582a3c8a..06d04d82e 100644
--- a/lamindb/models/ulabel.py
+++ b/lamindb/models/ulabel.py
@@ -26,6 +26,7 @@
     IsLink,
     SQLRecord,
     UNSET,
+    Unset,
     _get_record_kwargs,
     pop_space_branch_kwargs,
 )
@@ -224,7 +225,7 @@ class Meta(SQLRecord.Meta, TracksRun.Meta, TracksUpdates.Meta):
     def __init__(
         self,
         name: str,
-        type: ULabel | None = None,
+        type: ULabel | None | Unset = UNSET,
         is_type: bool = False,
         description: str | None = None,
         reference: str | None = None,
@@ -250,7 +251,7 @@ def __init__(
         if len(args) > 0:
             raise ValueError("Only one non-keyword arg allowed")
         name: str = kwargs.pop("name", None)
-        type: str | None = kwargs.pop("type", UNSET)
+        type: ULabel | None | Unset = kwargs.pop("type", UNSET)
         is_type: bool = kwargs.pop("is_type", False)
         description: str | None = kwargs.pop("description", None)
         reference: str | None = kwargs.pop("reference", None)
diff --git a/tests/pydata/test_record_basics.py b/tests/pydata/test_record_basics.py
index 3f6d656c3..4e289882e 100644
--- a/tests/pydata/test_record_basics.py
+++ b/tests/pydata/test_record_basics.py
@@ -690,14 +690,24 @@ def test_feature_manager_raise_not_validated_values():
 def test_name_lookup():
     my_type = ln.Record(name="MyType", is_type=True).save()
     label1 = ln.Record(name="label 1", type=my_type).save()
+    # same type → returns existing typed record
     label2 = ln.Record(name="label 1", type=my_type)
     assert label2 == label1
+    # no type passed, only typed record exists → fallback returns the typed one
     label2 = ln.Record(name="label 1")
-    assert label2 != label1
-    label2.save()
-    label3 = ln.Record(name="label 1")
-    assert label3 == label2
-    label2.delete(permanent=True)
+    assert label2 == label1
+    # explicit type=None → root-level dedup: label1 is typed so not found → new record
+    label_new = ln.Record(name="label 1", type=None)
+    assert label_new != label1
+    assert label_new._state.adding  # not yet saved, truly a new record
+    # explicit type=None, root-level record exists → root-level dedup finds it → returns it
+    root_label = ln.Record(name="root label 1").save()
+    label3 = ln.Record(name="root label 1", type=None)
+    assert label3 == root_label
+    # no type passed (UNSET) → search all → finds the existing root-level record
+    label4 = ln.Record(name="root label 1")
+    assert label4 == root_label
+    root_label.delete(permanent=True)
     label1.delete(permanent=True)
     my_type.delete(permanent=True)
 
@@ -1194,7 +1204,7 @@ def test_record_features_add_remove_values():
 
     # test passing ISO-format date string for date
 
-    test_record2 = ln.Record(name="test_record").save()
+    test_record2 = ln.Record(name="test_record_2").save()
     # we could also test different ways of formatting but don't yet do that
     # in to_dataframe() we enforce ISO format already
     feature_date = ln.Feature.get(name="feature_date")
diff --git a/tests/pydata/test_sqlrecord.py b/tests/pydata/test_sqlrecord.py
index 3767cd1ae..3c50017a8 100644
--- a/tests/pydata/test_sqlrecord.py
+++ b/tests/pydata/test_sqlrecord.py
@@ -12,6 +12,7 @@
 from lamindb.errors import FieldValidationError
 from lamindb.models import sqlrecord as sqlrecord_module
 from lamindb.models.sqlrecord import (
+    UNSET,
     _get_record_kwargs,
     _search,
     check_key,
@@ -287,10 +288,10 @@ def test_suggest_similar_names():
     assert ln.Record(name="Test experiment 1").uid == record1.uid
 
     assert suggest_records_with_similar_names(
-        record1, "name", {"name": "Test experiment 1"}
+        record1, "name", {"name": "Test experiment 1", "type": UNSET}
     )
     assert not suggest_records_with_similar_names(
-        record2, "name", {"name": "Test experiment 123"}
+        record2, "name", {"name": "Test experiment 123", "type": UNSET}
     )
 
     queryset = _search(
@@ -401,7 +402,7 @@ def test_get_record_kwargs():
             "dtype",
             "SimpleDtype | SimpleDtypeStr | ULabel | Record | Registry | list[Registry] | FieldAttr",
         ),
-        ("type", "Feature | None"),
+        ("type", "Feature | None | Unset"),
         ("is_type", "bool"),
         ("unit", "str | None"),
         ("description", "str | None"),