Teaching AI Pair Programming at Hackathons: A Look Inside the Nebula Fog Aider Workshop
Hook
The best way to teach developers about AI coding assistants isn’t through slides alone—it’s through reproducible terminal recordings they can replay at their own pace.
Context
Hackathons move fast. Participants have hours, not days, to learn new tools, form teams, and ship projects. When you’re trying to introduce an AI pair programming tool like Aider—which fundamentally changes how developers write code—traditional presentation formats fall short. You can’t just throw up bullet points about commands and expect people to internalize a new workflow.
Max Rousseau faced this challenge at the Nebula Fog hackathon. Rather than creating yet another slide deck that would be forgotten five minutes after the presentation, he built a teaching package that combines interactive markdown slides with terminal session recordings. The result is a reusable workshop format that lets developers see Aider in action, pause when they need to, and reference the material later when they’re actually coding. This repository represents a small but important pattern: how we document and teach AI-assisted development workflows in time-constrained environments.
Technical Insight
The technical architecture here is deliberately minimal, which is the entire point. The repository uses two complementary tools that each solve a specific presentation problem.
First, there’s Slidev—a presentation framework built on Vue.js that renders Markdown as slides. Unlike PowerPoint or Google Slides, Slidev lets you write presentations in pure Markdown with embedded code blocks that actually execute. For a technical audience, this means no more screenshots of code that become outdated the moment you take them. Here’s what a typical Slidev slide structure looks like:
# What is Aider?
An AI pair programming tool that:
- Edits code in your local repository
- Understands git context and file relationships
- Works with GPT-4, Claude, and other LLMs
```bash
aider --model gpt-4 src/main.pyLive Demo
The power of this approach is that code examples maintain syntax highlighting, and you can even embed interactive components. For a hackathon setting where participants might be following along on their laptops, they can clone the repo, run npm install && npm run dev, and have the entire presentation running locally in seconds.
The second component is the Asciinema recording (demo.cast). Asciinema captures terminal sessions as lightweight text-based recordings rather than video files. A typical recording session might look like this workflow:
# Start recording
asciinema rec demo.cast
# Demonstrate Aider workflow
aider --model gpt-4
> /add src/api.py
> Add error handling to the fetch_user function
> /commit
# Stop recording
exitThe resulting .cast file is just JSON with timestamps and terminal output. It’s typically under 100KB compared to multi-megabyte video files, and crucially, viewers can copy-paste text directly from the recording. When a hackathon participant sees aider --model gpt-4 --no-auto-commits, they can highlight and copy that exact command.
The presentation flow likely follows a standard teaching pattern: introduce the problem (manual coding is slow), show the solution (Aider’s AI assistance), demonstrate core workflows (adding files, making changes, committing), and handle edge cases (what happens when the AI makes mistakes). By combining slides with terminal recordings, you cover both the conceptual understanding and the muscle-memory commands.
What makes this architecture effective for hackathons specifically is the offline-first nature. Once a participant clones the repo and installs dependencies, they don’t need internet connectivity to review the material. They’re not dependent on YouTube staying up or conference WiFi handling video streaming for 50 people simultaneously. The Slidev dev server runs locally, and Asciinema playback is purely client-side.
The technical limitation that becomes an advantage: this isn’t trying to be a comprehensive course or documentation site. It’s scoped to exactly one presentation’s worth of content. That constraint forces clarity. There’s no sprawling wiki, no maintenance burden of keeping multiple pages in sync, no version compatibility matrix. Just: here’s what Aider does, here’s how to use it, go build something.
Gotcha
The biggest limitation is right in the description: this is presentation material, not a tool you’ll integrate into your workflow. If you’re looking for actual Aider functionality, scripts that automate Aider commands, or plugins that extend Aider’s capabilities, you won’t find them here. The repository delivers exactly one thing—a presentation—and nothing more.
The Asciinema recording presents its own challenges. Terminal recordings are fantastic for showing command-line workflows, but they’re terrible for explaining why certain commands matter. You see the output, but you don’t get the internal monologue of the presenter explaining why they chose --no-auto-commits or what to do when the AI suggests a change you disagree with. The slides presumably fill this gap, but the tight coupling means the recording only makes full sense when viewed alongside the presentation. This also means the content ages quickly—if Aider updates its CLI flags or changes its default behavior, the recording becomes misleading rather than helpful. With only 2 stars and no recent updates visible, there’s a real question of whether this material reflects current Aider behavior or represents a snapshot from a specific hackathon date.
Verdict
Use if: You’re organizing a hackathon or workshop and need a proven format for teaching AI coding tools in time-constrained settings. The Slidev + Asciinema combination is genuinely smart for technical presentations, and you can fork this repo to create similar workshops for other tools. Also use this if you’re specifically interested in how Aider was used at Nebula Fog and want to see a worked example rather than reading documentation. Skip if: You want to learn Aider itself—go straight to the official paul-gauthier/aider repository instead. Skip if you need production-ready code, reusable libraries, or anything beyond presentation materials. This is a teaching artifact from a specific event, not a maintained educational resource. For current Aider tutorials, you’re better served by the official docs or community-created video courses that receive regular updates.
