Restructure/modify tutorial to encourage AI use

[fltfx]

I am working on a bigger project of adjusting or re-doing the Django Girls tutorial but in the age of AI. We should expect students to be tempted to plug all the questions or things they're stuck on into an agent instead of trying to figure it out for themselves. I want to restructure the tutorial (at a higher level) in a age where we expect the students to use the agent. In fact during our events, I have already seen students use an agent throughout the tutorial and I'm not sure they're actually using AI in the best way. Let's face this head-on and encourage usage.

I refer to this article as a guide (but it is aimed at more basic python than web development): https://ildarakhmetov.substack.com/p/cs1-in-the-age-of-ai

and this one is also interesting: https://www.theatlantic.com/ideas/2026/06/ai-open-ai-anthropic/687689/

I also think there is possibility that we keep the tutorial the same, but just include information every step of the way on how to use Cursor to ask questions when they get stuck and to not encourage them to just plug their problem into chatgpt and copy/paste code. I have been going back and forth with chatgpt and so far have settled on:

If I were redesigning Django Girls, I'd keep about 90% of the tutorial intact. One of Django Girls' greatest strengths is its approachable, step-by-step narrative, and I wouldn't want to lose that.

Instead, I'd weave in three new elements throughout:

Think first: Before each major coding step, ask learners to predict what they're about to build or why a concept exists.
AI coach: Provide example Cursor prompts that encourage explanations, debugging, and exploration rather than code generation.
Reflect afterward: After completing the step, ask learners to explain what changed, why it worked, or how they'd modify it.

That preserves the tutorial's accessibility while teaching an additional, modern skill: how to collaborate productively with AI without becoming dependent on it. Over time, students would internalize a workflow that mirrors experienced developers today—observe, hypothesize, ask focused questions, verify the answers, and only then make changes.

I will create some prototypes about perhaps @das-g , you and I can discuss the best path forward. I'll DM you.

[das-g]

There's two IMO somewhat orthogonal changes being proposed here:

1. giving some guidance to tutorial readers and workshop participant who'd use «AI» anyway for following the tutorial on how to do so somewhat more responsibly (for their learning and comprehension success as well as for getting the software done, as well as hopefully causing less learned helplessness than naïve usage would probably cause)
2. encouraging «AI» use

As there definitely will be a significant portion of participants who'd use «AI» no matter what (even if we were to discourage using it for the tutorial), (1) is probably a good idea.

(2) however would IMO be irresponsible, as «AI» in its current iteration has many problems:

• environmental impact (energy consumption, sprawling data centers, water usage, …)
• socioeconomic issues (Who owns and controls all those services? What's their business model.)
• copyright (and related intellectual property) infringement
• other content theft, putting creatives out of jobs
• overloading (sometimes inadvertently DDOS-ing) websites with scrapers
• non-determinism
• confabulation, including inventing fictitious sources
• omitting actual sources (thus plagiarizing)
• rug-pull-ish bait-and-switch pricing changes
• lacking explainability
• generates over-complicated code
• generates under-complicated (but often overly complex) explanations
• probably generally makes us dumber when we use it, even if we try to "use it right"

Conflating (1) and (2) is, with all due respect, kinda like suggesting the "use responsibly" labels on alcoholic drinks are there to encourage alcohol consumption.

If this project decides to go towards (2), I'll consider stepping down as a maintainer, as that isn't something I can stand behind pedagogically, didactically or ethically. (Not that I was doing all that much maintaining recently, anyway. It'd probably be good for the project to have some more maintainers.)

[das-g]

I'll DM you.

Where? (GitHub AFAIK doesn't have DMs, does it?)

Also, why? Are there aspects to be discussed in private rather than in public? 🤨

[das-g]

Where? (GitHub AFAIK doesn't have DMs, does it?)

Ah, I now see you've sent me an e-mail. 😄

[ekohl]

I think @das-g has expressed many of my concerns with AI in general, but want to also emphasize that I strongly doubt that you actually learn anything by following the tutorial using AI. You don't learn math by using a calculator: you start by hand and once you've built a mental model then you can use a calculator effectively.

From my point of view the tutorial should discourage use of AI during the workshop. You have expert coaches around, which is the value IMHO. You should also learn to ask questions, discuss with others. Those are IMHO much more valuable than actually completing the tutorial.

[das-g]

Indeed, I originally wanted to focus my above message on the aspects @ekohl now mentioned in #1905 (comment), but didn't know how to phrase those well, then got carried away listing all the other downsides of «AI» (by which I also still stand) and then kinda only subsumed these central aspects under the kinda plump «probably makes us dumber» argument. 🙈

But yeah, I too believe «AI» usage hampers learning rather than helping it, and that it's doing that more so (and more fundamentally) than using a calculator when trying to learn math, and even we're trying to "use it right" for that learning purpose, even though certain techniques can probably mitigate the negative impact somewhat.

[fltfx]

Hey @das-g @ekohl !

@das-g About your thoughts:
(1) I agree with you that the goal of this is to provide some guidance (and mention of the downsides/risks) responsibly on how to use AI alongside the tutorial (even if we were to discourage using it for the tutorial).
(2) When I said "encourage", I didn't mean to come across as necessarily being "pro-AI" (as I am not) or that we should not mention the downsides of AI. When I used "encourage", I more mean that we know (I have seen it at multiple DG events in the past 2 years) that attendees use AI during the tutorial anyway -- so we might as well face it and provide guidance. I guess I should have wrote something more along the lines of "we should accept AI usage alongside the tutorial and provide guidance". However, I also think this is an issue of sort-of semantics --> if we decide to add content/guidance on AI usage in the tutorial, even if we have many disclaimers, are we not condoning ("encouraging"?) AI? I am not quite sure how to handle this. But I don't think we should maintain tutorial as is and never at least mention AI.
I also want to be clear that for (2), I am aware and in agreement with you and the problems of AI.

@ekohl Hi! Thanks for your thoughts. I agree with you that during the tutorial DG event, the attendees should be asking their coach (or other group members) for help -- we have always encouraged this (even for participants that start using AI on their own during the event). However, there are 2 scenarios not covered here:

1. Attendees who cannot finish the tutorial during the event (and often, the majority of attendees do not finish), they go home, try to continue, and get stuck on something -- and can't continue because they're stuck. In an ideal world, we want the attendees to reach out to their coach/group-members for further help, but it doesn't seem to work out that way. I am really doubtful that the younger generation would try to search through stackoverflow etc to try to continue. They would plug in their question into an AI agent anyway at this point.
2. People that try to follow the tutorial without attending an event (either they can't as it's not offered nearby or don't want to). If this person gets stuck, again, I think they're going to try to use an AI agent.

Thanks to you both for your thoughts.

Re: private vs public conversation 😄
My goal is to keep the amount of participants on this issue relatively small, so we can come to a conclusion quickly. I think if there's a lot of disagreement on the approach, perhaps we can bring the advisory board in to help make a decision.

[das-g]

However, I also think this is an issue of sort-of semantics --> if we decide to add content/guidance on AI usage in the tutorial, even if we have many disclaimers, are we not condoning ("encouraging"?) AI?

Eh, IMO not necessarily, if we phrase it carefully. It'd be like those recommendations along the lines of «Best is not to drink alcohol at all, but if you decide to drink some anyway, keep it at maximum one standard-size drink per day and do not combine it with medication.»

[ekohl]

Thank you for that much more nuanced reply @fltfx. When I thought about it this morning (before reading your reply) I also came to the conclusion that a section on AI guidance would be useful.

A chapter that discourages the use of AI during events doesn't seem controversial to me and something that can later be extended.

Also interesting you mention that 2 years timeframe. It's actually 2 years since the last time I was a coach since the group that used to organize it no longer does. Because of that, I didn't catch any of the AI changes at events myself.