More doc improvements: simplify getting started page (#629)

* Simplify getting started page

* Move slides to articles page

docs/articles.md

@@ -1,5 +1,5 @@
-# Articles
+# Articles and Presentations
 
-[Cubed: Bounded-memory serverless array processing in xarray](https://xarray.dev/blog/cubed-xarray)
-
-[Optimizing Cubed](https://medium.com/pangeo/optimizing-cubed-7a0b8f65f5b7)
+- [Cubed: an introduction](https://cubed-dev.github.io/cubed/cubed-intro.slides.html) by Tom White, November 2024
+- [Optimizing Cubed](https://medium.com/pangeo/optimizing-cubed-7a0b8f65f5b7) by Tom White, April 3, 2024
+- [Cubed: Bounded-memory serverless array processing in xarray](https://xarray.dev/blog/cubed-xarray) by Tom Nicholas and Tom White, June 28th, 2023

docs/getting-started.md

@@ -0,0 +1,61 @@
+# Getting Started
+
+We'll start with a simple example that runs locally.
+
+## Installation
+
+Install Cubed with pip:
+
+```shell
+pip install cubed
+```
+
+This installs a minimal set of dependencies for running Cubed, which is sufficient for the demo below. You can also install the `diagnostics` extra package, which is needed for later examples to provide things like progress bars and visualizations of the computation:
+
+```shell
+pip install "cubed[diagnostics]"
+```
+
+Alternatively, you can install Cubed with Conda (note that this doesn't include the packages for diagnostics):
+
+```shell
+conda install -c conda-forge cubed
+```
+
+## Demo
+
+First, we'll create a small array `a`:
+
+```python
+import cubed
+import cubed.array_api as xp
+spec = cubed.Spec(work_dir="tmp", allowed_mem="100kB")
+a = xp.asarray([[1, 2, 3], [4, 5, 6], [7, 8, 9]], chunks=(2, 2), spec=spec)
+```
+
+Cubed implements the [Python Array API standard](https://data-apis.org/array-api/latest/), which is essentially a subset of NumPy, and is imported as `xp` by convention.
+
+Notice that we also specify chunks, just like in Dask Array, and a {py:class}`Spec <cubed.Spec>` object that describes the resources available to run computations.
+
+Next we create another array `b` and add to two array together to get `c`.
+
+```python
+b = xp.asarray([[1, 1, 1], [1, 1, 1], [1, 1, 1]], chunks=(2, 2), spec=spec)
+c = xp.add(a, b)
+```
+
+Cubed uses lazy evaluation, so nothing has been computed yet.
+
+```python
+c.compute()
+```
+
+This runs the computation using the (default) local Python executor and prints the result (if run interactively):
+
+```
+array([[ 2,  3,  4],
+       [ 5,  6,  7],
+       [ 8,  9, 10]])
+```
+
+That's it! For your next step you can read the [user guide](user-guide/index.md), have a look at [configuration](configuration.md) options, or see more [examples](https://github.com/cubed-dev/cubed/blob/main/examples/README.md) to run locally or in the cloud.

docs/index.md

@@ -29,9 +29,8 @@ Cubed is horizontally scalable and stateless, and can scale to multi-TB datasets
 :hidden:
 :maxdepth: 2
 :caption: For users
-getting-started/index
+getting-started
 user-guide/index
-Intro slides <https://cubed-dev.github.io/cubed/cubed-intro.slides.html>
 Examples <https://github.com/tomwhite/cubed/tree/main/examples/README.md>
 api
 array-api

docs/user-guide/diagnostics.md

@@ -5,7 +5,7 @@ Cubed provides a variety of tools to understand a computation before running it,
 To use these features ensure that the optional dependencies for diagnostics have been installed:
 
 ```shell
-python -m pip install "cubed[diagnostics]"
+pip install "cubed[diagnostics]"
 ```
 
 ## Visualize the computation plan