"Using AI in Development" content

[adamblanchard]

Since we previously agreed on an outline for the module, this PR actually fills out the content in further detail, specifically the exercises, prep and assignment.

What I need reviewing:
While I have added a lot of content here, it's all up for feedback. If you have better ideas for any of the prep, exercises, assignment etc. just share! The goal would be to get this into a good enough state to run the session with, and then act on feedback/learnings after running it for the first time.

Preview changes:
You can preview all of the changes easily here.

Remaining changes:
The exercise code should actually live in the hyf-assignment-repo, so trainees have it to hand during the session. But, i will leave them all in here for the review, and move them/update links afterwards.

[AThraen]

Just a bit of feedback from me here - use it if it seems relevant.
Curriculum looks good - but I am missing some parts:

• What AI/LLMS should I use for what.
• How does the context window work, what should I be concerned about there (context rot).
• Whats the difference between the different kinds of ways developers experience AI (code completion, chat-prompt, agent mode, PR-agents on github, CLI).
• How can you use CLI agents most effeciently (plan mode, subagents, etc).
• What are AI agent skills and how do I use them?
• What are MCP servers and which ones makes sense to use as a developer and so on.
• Why does RAG matter
• What is a Ralph Wiggum loop and how do I use it

[hosseinzahed]

Thanks @adamblanchard and @AThraen,
You've pointed out the latest capabilities and it looks great to me. Let's try it in practice to see the outcome.

[adamblanchard]

Thanks @AThraen for the pointers. I don't know all the things you've mentioned, but i pull out two themes: deeper understanding of AI and how it works (context behaviour, llm specifics, ralph wiggum) and agent-type dev workflows (skills, cli, mcp).

In general, those things feel quite far out of the scope that this module is currently aiming for. I'm quite for the idea of that being in scope in the future though, since they are both themes we are clearly going quite light on right now.

However, I've discussed with @saloumeh-67 , and i think you have too, about testing deeper AI training with some kind of masterclass or additional sessions first. I would be super keen to test this soon in it's own feedback loop, and bring learnings back into the curriculum for the next teams as appropriate.

For now, @hosseinzahed has lots of experience running a session like this, and comparing it to the last curriculum, the trainees will still leave with a much better understanding of using AI than they did before. We can get feedback from that, as well as having lots more scope to explore, all ready to improve the next run of this module.

Thanks for all the ideas.