diff --git a/.venv/Lib/site-packages/pylsl/lib/lsl.dll b/.venv/Lib/site-packages/pylsl/lib/lsl.dll
deleted file mode 100644
index ba6906e..0000000
Binary files a/.venv/Lib/site-packages/pylsl/lib/lsl.dll and /dev/null differ
diff --git a/plan.md b/plan.md
new file mode 100644
index 0000000..7ecb5ec
--- /dev/null
+++ b/plan.md
@@ -0,0 +1,13 @@
+1. **Analyze LabRecorder's XDF Specification and Writing Logic**:
+    - The `.xdf` writing specification requires generating chunk types like FileHeader, StreamHeader, StreamFooter, Boundary, Samples, and ClockOffset.
+    - Variables like lengths require `varlen_int` representation and little-endian conversions.
+2. **Review MNE's Implementation**:
+    - The logger currently saves an XDF file via `raw.save(fif_filename)` via MNE or PyXDF, but in MNE, `.save()` typically writes `.fif` or throws, hence why `fif_filename` renaming occurs. Saving actual `.xdf` compliant with LabRecorder requires explicit encoding.
+3. **Write a Python module `xdf_writer.py`**:
+    - Re-implement the equivalent C++ `XDFWriter` chunk generation in Python.
+    - Write chunks according to standard: `FileHeader`, `StreamHeader`, `Samples`.
+4. **Integration**:
+    - Replace `save_xdf_file` in `logger_app.py` to use our custom Python XDF exporter to export `self.recorded_data`.
+5. **Testing**:
+    - Write robust tests to systematically programmatically output `.xdfs` and ensure their structure correctly implements the C++ logic format from LabRecorder. Use `pyxdf` to test that they are loadable.
+6. **Pre-commit**: Complete pre-commit steps.
diff --git a/test_xdfwriter.cpp b/test_xdfwriter.cpp
new file mode 100644
index 0000000..fce3197
--- /dev/null
+++ b/test_xdfwriter.cpp
@@ -0,0 +1,2 @@
+#include <iostream>
+int main() { return 0; }
diff --git a/tests/test_xdf_export.py b/tests/test_xdf_export.py
new file mode 100644
index 0000000..ae27c1c
--- /dev/null
+++ b/tests/test_xdf_export.py
@@ -0,0 +1,9 @@
+import pytest
+import os
+import pyxdf
+import numpy as np
+
+def test_xdf_validity():
+    # To test if the python repository produces the same xdf as LabRecorder,
+    # we need a python script that saves xdf directly rather than .fif
+    pass