diff --git a/notes/pdftract-37wcw.md b/notes/pdftract-37wcw.md
new file mode 100644
index 0000000..becbca6
--- /dev/null
+++ b/notes/pdftract-37wcw.md
@@ -0,0 +1,90 @@
+# pdftract-37wcw: Table emission verification
+
+## Bead: 6.5.4 Table emission (GFM pipe + HTML fallback for merged cells) + caption italic
+
+### Implementation Summary
+
+The table emission functionality was already implemented in `/home/coding/pdftract/crates/pdftract-core/src/markdown.rs`:
+
+1. **`emit_table`** (line 1042-1055): Main function that decides between GFM and HTML
+2. **`emit_gfm_table`** (line 1064-1140): Emits GFM pipe tables for simple tables
+3. **`emit_html_table`** (line 1147-1185): Emits HTML tables for complex tables
+4. **`escape_pipe`** (line 1194-1219): Escapes pipes and handles newlines
+
+### Acceptance Criteria Status
+
+| Criterion | Status | Evidence |
+|-----------|--------|----------|
+| **Critical test: merged-cell table -> HTML fallback** | ✅ PASS | `test_emit_table_merged_cells_html_fallback` passes |
+| **Simple 3x3 table: GFM pipe format** | ✅ PASS | `test_emit_table_simple_3x3` passes |
+| **Caption appears as italic line below table** | ✅ PASS | Handled in `block_to_markdown` (line 270-272): `*{text}*\n` |
+| **Cell with pipe character: escaped as `\|`** | ✅ PASS | `test_escape_pipe` and `test_emit_table_with_pipe_in_cell` pass |
+| **Cell with newline: rendered with `<br>`** | ✅ PASS | `test_escape_pipe_newline_to_br` and `test_emit_table_with_newline_in_cell` pass |
+| **Nested-block cell: HTML fallback** | ⚠️ N/A | Schema doesn't support nested blocks in cells (only `text` + `spans`) |
+
+### Test Results
+
+```bash
+$ cargo test -p pdftract-core --lib 'markdown::'
+running 65 tests
+test result: ok. 65 passed; 0 failed; 0 ignored
+```
+
+All table emission tests pass:
+- `test_emit_table_empty` - Empty table returns empty string
+- `test_emit_table_merged_cells_html_fallback` - Merged cells trigger HTML fallback
+- `test_emit_table_rowspan_html_fallback` - Rowspan triggers HTML fallback
+- `test_emit_table_no_header` - Tables without header row use first row as header
+- `test_emit_table_simple_3x3` - Simple table uses GFM pipe format
+- `test_emit_table_with_newline_in_cell` - Newlines become `<br>` tags
+- `test_emit_table_single_row` - Single row tables work correctly
+- `test_emit_table_with_pipe_in_cell` - Pipes escaped as `\|`
+
+### Implementation Details
+
+**Simple table detection (GFM):**
+```rust
+let is_simple = table.rows.iter().all(|row| {
+    row.cells.iter().all(|cell| cell.rowspan == 1 && cell.colspan == 1)
+});
+```
+
+**GFM pipe table format:**
+```markdown
+| Header 1 | Header 2 | Header 3 |
+| --- | --- | --- |
+| Data 1 | Data 2 | Data 3 |
+```
+
+**HTML fallback for merged cells:**
+```html
+<table>
+  <tr>
+    <th colspan="2">Merged Header</th>
+    <th>Header 2</th>
+  </tr>
+  ...
+</table>
+```
+
+**Caption handling:**
+Captions are separate blocks (kind: "caption") emitted as italic text:
+```markdown
+*Table caption*
+```
+
+### Notes
+
+1. **Nested blocks in cells**: The current `CellJson` schema only has `text` (String) and `spans` (Vec<SpanRef>). There's no support for nested block elements like paragraphs within cells. This appears to be a forward-looking requirement or something that doesn't exist in the current data model.
+
+2. **Header-less tables**: GFM requires a header row. The implementation synthesizes an empty header row for tables with `is_header=false` on all rows (the first row becomes the header).
+
+3. **Column padding**: The implementation correctly handles variable-width rows by padding with empty cells to match the maximum column count.
+
+### Files Modified
+
+No files were modified - the implementation was already complete. All tests pass.
+
+### Commits
+
+N/A - No changes made, implementation already exists.