diff --git a/README.md b/README.md
index 1f613974d..efd413aee 100644
--- a/README.md
+++ b/README.md
@@ -2,7 +2,7 @@
 
 A Rust implementation of the [GeoArrow](https://github.com/geoarrow/geoarrow) specification and bindings to [GeoRust algorithms](https://github.com/georust/geo) for efficient spatial operations on GeoArrow memory.
 
-This repository also includes [JavaScript (WebAssembly) bindings](https://github.com/kylebarron/geoarrow-rs/blob/main/js/README.md), wrapping the GeoArrow memory layout and offering vectorized geometry operations.
+This repository also includes [Python bindings](https://github.com/kylebarron/geoarrow-rs/blob/main/python/core/README.md) and [JavaScript (WebAssembly) bindings](https://github.com/kylebarron/geoarrow-rs/blob/main/js/README.md), wrapping the GeoArrow memory layout and offering vectorized geometry operations.
 
 ## Documentation
 
diff --git a/python/README.md b/python/README.md
new file mode 100644
index 000000000..df6d1f365
--- /dev/null
+++ b/python/README.md
@@ -0,0 +1,20 @@
+# `geoarrow.rust`: Python bindings to `geoarrow-rs`
+
+This folder contains Python bindings to the [GeoArrow Rust implementation](https://github.com/geoarrow/geoarrow-rs).
+
+`geoarrow.rust` is distributed with [namespace packaging](https://packaging.python.org/en/latest/guides/packaging-namespace-packages/), meaning that each python package `geoarrow-rust-[submodule-name]` (imported as `geoarrow.rust.[submodule-name]`) can be published to PyPI independently. The benefit of this approach is that complex C dependencies can be built and packaged independently.
+
+Modules so far:
+
+- [`core`](./core/README.md): All algorithms and data structures implemented in pure Rust without any C dependencies. Having a pure Rust dependency tree means it's trivial to build binary wheels for many operating system architectures that might not be possible with C dependencies.
+
+Future modules:
+
+- `geos`: [GEOS](https://libgeos.org/)-based algorithms on GeoArrow memory.
+- `proj`: [PROJ](https://proj.org/en/9.3/)-based coordinate reprojection on GeoArrow memory.
+
+In order to obtain relevant modules, you should install them from PyPI directly, e.g.:
+
+```
+pip install geoarrow-rust-core
+```
diff --git a/python/core/README.md b/python/core/README.md
index 206b43be0..0c1eff938 100644
--- a/python/core/README.md
+++ b/python/core/README.md
@@ -1,13 +1,25 @@
-# geoarrow.rust.core
+# `geoarrow.rust.core`: Python bindings to `geoarrow-rs`
 
-This library contains:
+This library contains Python bindings to the [GeoArrow Rust implementation](https://github.com/geoarrow/geoarrow-rs).
 
-- standalone classes to represent GeoArrow arrays in rust: `PointArray`, `LineStringArray`, etc
-- pure-rust algorithms (from [`georust/geo`](https://github.com/georust/geo)) that don't require a C extension module and can statically link on every platform.
+- **Fast**: Connects to algorithms implemented in [GeoRust](https://georust.org/), which compile to native code.
+- **Parallel**: Multi-threading is enabled out-of-the-box for all operations on chunked data structures.
+- **Self-contained**: Zero Python dependencies.
+- **Easy to install**: Distributed as static binary wheels with zero C dependencies.
+- **Strong, static typing**: geometry arrays have a known type
+- **Interoperable ecosystem**: Data can be shared at zero cost with other libraries in the burgeoning [GeoArrow ecosystem](https://geoarrow.org/), such as [geoarrow-c](https://github.com/geoarrow/geoarrow-c/tree/main/python) or [lightning-fast map rendering](https://github.com/developmentseed/lonboard).
 
-In the future, this will also contain:
+More specifically, it contains:
 
-- Chunked classes: `ChunkedPointArray`, `ChunkedLineStringArray`, etc
-- Table representations: `GeoTable`, where one of the columns is a geospatial type. This will support e.g. geospatial joins using Arrow memory.
+- Classes to represent GeoArrow arrays: `PointArray`, `LineStringArray`, etc.
+- Classes to represent _chunked_ GeoArrow arrays: `ChunkedPointArray`, `ChunkedLineStringArray`, etc.
+- A spatial table representation, `GeoTable`, where one column is a geospatial type, to enable future support for geospatial joins.
+- Rust-based algorithms for computations on GeoArrow memory.
+- Rust-based parsers for various geospatial file formats.
 
-Refer to the [GeoArrow Python module proposal](https://github.com/geoarrow/geoarrow-python/issues/38) for more information.
+Future work:
+
+- [ ] 3D coordinates. Only 2D geometries are supported at this time.
+- [ ] CRS management. This currently loses the CRS information in the [GeoArrow metadata](https://geoarrow.org/extension-types#extension-metadata).
+
+Refer to the [GeoArrow Python module proposal](https://github.com/geoarrow/geoarrow-python/issues/38) for more background information.