diff --git a/.gitignore b/.gitignore
index bd295eb..8bfbf1a 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,6 +1,7 @@
 data/*.parquet
 __pycache__/
 *.pyc
+*.egg-info/
 .venv/
 venv/
 data/dispensing_partitioned/
diff --git a/README.md b/README.md
index 5cb8f51..07892c3 100644
--- a/README.md
+++ b/README.md
@@ -79,19 +79,28 @@ python3 -m venv .venv
 
 ### Step 4: Activate the virtual environment
 
+On macOS and Linux:
+
 ```bash
 source .venv/bin/activate
 ```
 
+On Windows (Git Bash), the venv puts its scripts in `Scripts/` instead of
+`bin/`:
+
+```bash
+source .venv/Scripts/activate
+```
+
 You should now see `(.venv)` at the start of your terminal prompt.
 
 ### Step 5: Install dependencies
 
 ```bash
-pip install -r requirements.txt
+pip install .
 ```
 
-This installs a single package: `polars==1.42.1`.
+This reads `pyproject.toml` and installs a single package: `polars==1.42.1`.
 
 ### Step 6: Generate the data
 
@@ -142,7 +151,7 @@ If you see `SSL: CERTIFICATE_VERIFY_FAILED` due to a corporate firewall that
 intercepts HTTPS traffic:
 
 ```bash
-pip install --trusted-host pypi.org --trusted-host files.pythonhosted.org -r requirements.txt
+pip install --trusted-host pypi.org --trusted-host files.pythonhosted.org .
 ```
 
 #### Wrong interpreter selected in VS Code
@@ -323,7 +332,7 @@ perform a lazy aggregation across the partitioned dataset.
 ```
 parquet-polars-training/
 ├── README.md                   ← you are here
-├── requirements.txt            ← polars==1.42.1
+├── pyproject.toml              ← project metadata + dependencies (polars==1.42.1)
 ├── .gitignore
 ├── data/
 │   ├── generate_data.py        ← run this to create the Parquet files
diff --git a/exercises/ex01_vscode_explorer.md b/exercises/ex01_vscode_explorer.md
index 1c1dddf..5ff6433 100644
--- a/exercises/ex01_vscode_explorer.md
+++ b/exercises/ex01_vscode_explorer.md
@@ -2,9 +2,9 @@
 
 > **SAS framing:** This is the replacement for double-clicking a `.sas7bdat` file
 > in SAS Enterprise Guide — look first, code second. Before writing any Python,
-> you should be able to open a data file and visually inspect its contents. You
-> can also run SQL queries against Parquet files directly in VS Code, just like
-> `PROC SQL` against a SAS dataset.
+> you should be able to open a data file and visually inspect its contents.
+> Data Wrangler gives you a point-and-click data grid with column statistics,
+> much like the SAS viewtable window combined with `PROC FREQ` output.
 
 ---
 
@@ -25,78 +25,77 @@ In VS Code, open the Extensions panel (Ctrl+Shift+X), then search for and instal
 
 After generating data (Exercise 0), you should have three files in `data/`:
 
-1. Open `data/demographic.parquet` — click it in the Explorer sidebar.
+1. Open `data/demographic.parquet` — right-click it in the Explorer sidebar
+   and choose **Open in Data Wrangler**.
 2. Open `data/enrollment.parquet`
 3. Open `data/dispensing.parquet`
 
 If the extension is working, you will see a data grid (not raw bytes). Take a
 moment to scroll through each file and get a feel for the columns and values.
 
-## Step 3: Query Parquet files with SQL
-
-Both Data Wrangler and Data Preview let you run SQL queries against opened
-Parquet files. This is the same concept as `PROC SQL` against a SAS dataset —
-the table name is the file name (without the `.parquet` extension).
-
-Here is a worked example. Open `demographic.parquet`, then run this query:
-
-```sql
--- PROC SQL equivalent: SELECT COUNT(*) AS n FROM demographic;
-SELECT COUNT(*) AS row_count
-FROM demographic;
-```
-
-You should get a result like `10000` — one row per patient.
-
-Try another one. This query counts how many patients are missing a
-`PostalCode` value (the `IS NULL` check works just like in PROC SQL):
-
-```sql
--- PROC SQL equivalent: SELECT COUNT(*) FROM demographic WHERE PostalCode IS NULL;
-SELECT COUNT(*) AS missing_postal
-FROM demographic
-WHERE PostalCode IS NULL;
-```
-
-> **Tip:** If you are not sure where to type the query, look for a "Query" or
-> "SQL" button in the Data Wrangler panel, or open a new SQL file
-> (`File → New File`) and save it with a `.sql` extension, then select the
-> Parquet file as the connection target.
+## Step 3: Explore with column statistics
+
+Data Wrangler is a **low-code/no-code tool** — there is no SQL query panel.
+Instead, everything you would ask with a small `PROC SQL` or `PROC FREQ` step
+is answered by clicking around the interface:
+
+- **Row count** — shown in the summary bar (e.g., "10,000 rows"). This is your
+  `SELECT COUNT(*)`.
+- **Column dtype** — shown in each column header next to the column name. This
+  is your `PROC CONTENTS`.
+- **Column statistics panel** — click a column header to open a summary of that
+  column: value distribution, unique values, and missing-value count. This is
+  your `PROC FREQ` / `PROC MEANS`.
+- **Filter and sort** — use the column header menus to filter rows or sort,
+  the way you would with a `WHERE` clause or `PROC SORT`.
+
+Here is a worked example. Open `demographic.parquet` in Data Wrangler and look
+at the summary bar — you should see **10,000 rows**, one row per patient. Now
+click the `PostalCode` column header: the statistics panel shows the count of
+missing values, which answers "how many patients have no postal code?" without
+writing a line of code.
+
+> **Note:** Data Wrangler can also generate Python (pandas) code for the
+> operations you perform in the grid. Feel free to peek at it, but do not copy
+> it into the exercises — this workshop uses Polars, and the two APIs do not
+> mix. Treat Data Wrangler as a viewer here, not a code generator.
 
 ## Step 4: Answer the following questions
 
-Use SQL queries (and the data grid view) to answer each question. Write your
-answers below each question or in a separate document.
+Use the data grid and column statistics panel to answer each question. Write
+your answers below each question or in a separate document.
 
 ### Q1: How many rows are in `dispensing.parquet`?
 
-Hint: `SELECT COUNT(*) FROM dispensing;`
+Hint: Open the file in Data Wrangler and read the row count from the summary
+bar.
 
 Your answer:
 
 ### Q2: What is the dtype of the `RxDate` column?
 
 Hint: Look at the column header in the data grid — the dtype is shown next to
-the column name. You can also query the schema if the extension supports it.
+the column name.
 
 Your answer:
 
 ### Q3: Which columns are in `enrollment.parquet`?
 
-Hint: `SELECT * FROM enrollment LIMIT 5;` — then look at the column headers
-in the result.
+Hint: Open the file and read the column headers across the top of the grid.
 
 Your answer:
 
 ### Q4: What are the unique values of `Sex` in `demographic.parquet`?
 
-Hint: `SELECT DISTINCT Sex FROM demographic;`
+Hint: Click the `Sex` column header — the statistics panel lists the unique
+values and their counts (like `PROC FREQ`).
 
 Your answer:
 
 ### Q5: What is the null count for `PostalCode` in `demographic.parquet`?
 
-Hint: `SELECT COUNT(*) FROM demographic WHERE PostalCode IS NULL;`
+Hint: Click the `PostalCode` column header and read the missing-value count
+from the statistics panel.
 
 Your answer:
 
@@ -105,7 +104,7 @@ Your answer:
 ## Why this matters
 
 In SAS, you would run `PROC CONTENTS` to see column metadata, `PROC PRINT` to
-eyeball the data, and `PROC SQL` to query specific values. The VS Code
-Parquet viewer gives you all three capabilities without writing a DATA step or
-PROC. Use it to confirm your expectations before you start writing Polars
-queries in the exercises that follow.
+eyeball the data, and `PROC FREQ` to check value distributions. Data Wrangler
+gives you all three capabilities without writing a DATA step or PROC. Use it to
+confirm your expectations before you start writing Polars queries in the
+exercises that follow.
diff --git a/pyproject.toml b/pyproject.toml
new file mode 100644
index 0000000..5ea003f
--- /dev/null
+++ b/pyproject.toml
@@ -0,0 +1,15 @@
+[build-system]
+requires = ["setuptools>=68"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "parquet-polars-training"
+version = "0.1.0"
+description = "Parquet + Polars workshop for SAS analysts"
+requires-python = ">=3.11"
+dependencies = [
+    "polars==1.42.1",
+]
+
+[tool.setuptools]
+packages = []
diff --git a/requirements.txt b/requirements.txt
deleted file mode 100644
index 02463aa..0000000
--- a/requirements.txt
+++ /dev/null
@@ -1 +0,0 @@
-polars==1.42.1