Using ChatGPT to generate example code

I ran this query in ChatGPT this morning:

I would like to show people how to fit a titration curve in a Jupyter Notebook using Python, pandas, numpy and scipy. Could you please generate a fake titration data set with noise and then provide code that does a fit and generates a graph?

and got the following answer. Let’s try running it in a new notebook in ctb.github.io/2026-bcp-101-stats!

(I’ll talk a bit more about ChatGPT and AI/LLM-based code generation in a bit.)

Some interim thoughts on teaching

Hopefully after tuesday and this morning you’ve seen some of why we use this stuff.

But it’s a lot, right??

Why are we teaching you this, and what should you teach your students, and why?

See the data science Venn diagram:

trefoil

I mostly teach the “hacking” because that’s what students in my grad programs are lacking the most.

I try to teach this at three levels simultaneously -

  • practical - gotta teach something, even if details will change quickly
  • conceptual - this is how computers think and work, even if details change, you’ll still do something like this
  • computational thinking - this is how you have to break down problems and structure your attempted solutions

Computer science is a completely distinct area, which contains (among other things):

  • programming and software engineering: how to effectively build and maintain software
  • data structures and algorithms: how to efficiently store and compute things

I usually don’t teach programming per se - I don’t teach what we did yesterday, usually, because it requires a whole quarter-long course all on its own! - but rather walk students through the interactions between computers and data and show them how to get results that they’re probably interested in. (But I also work with graduate students and upper division undergrads mostly, and they ahve different motivations.)

All of which is to say this is a lifetime learning endeavor and I honestly don’t know what you all should be teaching :)

My own soapbox: I focus on teaching and developing:

  • computational practice
  • computational thinking
  • how to build intuition by poking and prodding large data sets

because these are my areas of speciality.

BUT.

Computers and programming are really kind of terrible. Think about all the problems you’ve had in the last few days! We spend a LOT of our time teaching syntax, installing software, and tracking down errors

This may be changing, somewhat.

what has changed recently?

Large Language Models (AI/ChatGPT/Claude)

ChatGPT, Claude, and other LLMs make it easy to generate working code, and can also generate step-by-step installation instructions.

HOWEVER:

Computational precision/thinking is challenging and LLMs do not solve this. See this nice post.

Quote:

As Dijkstra says, “Formalism isn’t a burden.” Instead, it bridges the messy realm of human intent and the exact demands of machine execution.

Basically, English isn’t precise enough to tell the computer what to do exactly, so guesswork is involved!

There are also ethical challenges to consider: profit model; source of training data; and cost of running models.

If you’re going to use it, I suggest asking LLMs “how to do X” not “what should I do?”

Only ask it things where you can figure out if it’s right or not

And ask yourself: Is pulling apart a fully functioning example easy or hard for your audience? Can poke or prod…

Let’s go back to our learning goals:

  • we want to expose you to how we do things in the STEM data science research world
  • and show you a little bit of how the training community teaches those approaches,
  • and maybe see if some of those tools and approaches could be useful to you in your classroom settings.

Building and configuring your own custom site

Now we’re going to use GitHub to create our own custom JupyterLite cite.

Logging into GitHub.

First, let’s make sure you have a GitHub account and are logged in!

Go to github.com. Do you see “Dashboard”? If so, that’s all you need! If not, please create an account ;).

What is GitHub?

GitHub is a lot of things! It’s where a lot of programming code is shared and worked on, individually and collaboratively. It is also a great place to host code and Web sites for your own use.

The two key concepts for this section are (1) a GitHub repository, which is a collection of files, and (2) a GitHub pages Web site, which is built from the collection of files in a repository and then provided to everyone on the Internet.

Today we are going to make a copy of a repository by “forking” it, and then we are going to customize the repository by adding and removing files.

Previously, we’ve shown you three different “custom” JupyterLite deployments:

These all varied mostly in terms of what data sets and notebooks were “prepositioned” - readily available to people running that site. Next, we’re going to show you how to set up your own

(It is 100% possible to start with a “blank” site - just JupyterLite and basic software installed - and then have students upload their own data and notebooks. This just adds a lot of failure points the first time through any lesson.)

Building your own custom site

FIRST, Titus is going to demonstrate this process so you know what it looks like.

THEN we are going to go through it together! This may involve some troubleshooting!

Step 1 - make your own copy of the repo.

Start by going to the repository you want to copy. We’ll start with the JupyterLite demo repository, which is here:

https://github.com/jupyterlite/demo

On the upper right, there should be a green button labeled “Use this template.” Click on it, and select “Create a new repository.”

Now name it. You can name it anything you like. I have hundreds of these so I tend to put the year in, but you can use something short, like “bcp-101-demo”. Note that spaces and “weird” symbols aren’t allowed - just letters, numbers, and dashes.

You can leave Description blank, if you like, or enter something to help you and others know what it is - “Demo of jupyterlite”, maybe?

Last but not least, choose the visibility of the new repository. You have to make this public for later steps to work.

Now, click “Create Repository”.

Wait a few seconds and click “Refresh”.

Now you should have your very own copy under your own account!

Step 2 - Enable GitHub Pages.

Go to “Settings”, on the upper right.

Scroll down to Pages.

Under “Build and Deployment”, “Branch” - select “main”. Click “Save.”

Now, go to “Actions” in the top middle. You should see an orange “pages build and deployment” entry - it will turn green eventually (you may have to reload the page). This means everything is working! (The red entry on the bottom is because GitHub fails the first time it tries to build a Web site, since Pages is not yet configured!)

Step 3 - Make it easy to go to the Web site.

Go to your “Code” button (upper left) and click on it

On the right side, you’ll see a little gear icon next to “About”. Select that, and check the box that says “Use your GitHub pages web site.”

Then “Save Changes.”

On the right side under About you should now see a blue link. If you click on this blue link, that is the JupyterLite Web site that you can use yourself and also share with others! You can try opening it (maybe in an incognito window…) if you like!

Step 4 - customize your Web site!

In the GitHub repository, you will see a bunch of files. Anything in “content” will end up in your deployed GitHub Pages Web site and hence in your JupyterLite site.

You can upload any file you want here, as long as it’s under ~50 MB.

Let’s start by doing two things: first, deleting some files; and second, adding our own file.

To delete a file, click on it to view it. Then go to the three dots on the upper right, and select “Delete” from the menu. Then select “commit changes”, and then “commit changes” again.

Try deleting javascript.ipynb!

To add a file, go to ‘content’ and click “Add file”, and then select “Upload files”. You can now select or drag/drop files to add to your site.

Let’s try adding a notebook!

The question you might ask is … where do I get a notebook?? The answer is, anywhere! You can download one from your existing JupyterLite demo sites from day 1 or day 2, OR you can find one on GitHub, or write a new one yourself in a running JupyterLite and save it locally for later upload to your new GitHub repository.

If you need one, try going to this link and clicking on markdown-examples.ipynb. Now click the little download link on the right side, and it will bring it to your computer (where the file lands will differ by computer and browser, sorry…) Now go find it and select it to upload.

Step 5 - test your site

Each time you change your repository, GitHub runs the “Actions” to build the new Web site. This takes between 30 seconds and 5 minutes. You can check on this by going to Actions and waiting for it to turn green.

Now, go to your “Code” view and select the blue link, and open it in incognito mode. You should see your site with all your adjusted content!

Step 6 - Celebrate!

OK, really, iterate :). You can keep on adjusting and editing your site as you wish. Keep on testing it in incognito mode, tho!

Post-workflow notes

Making your own template site!

It is possible to make your own template repository out of any repository you own. This lets you develop your own “base” of files that you use and then make copies of it.

To do so, go to “Settings” again and select “Template repository”.

Now go back to the “code” view and you will see “Use this template.” Voila!

There is no limit to the number of repositories or sites in your GitHub account

…I have about 400 repos in my GitHub account :)

This site uses Just the Docs, a documentation theme for Jekyll.