Hacktoberfest Meetup Melbourne x Melbourne GitHub User Group

Event Description

This is your Hacktoberfest Meetup Melbourne edition! We're super excited to host this in Australia, and hope you can all come along.

Join us for a night of mingling with other developers, hear about GitHub, open-source, and open-weight, and get excited for all the amazing things happening around AI assisted coding. You'll hear talks and content from people who are building and using open-source AI models and open-weight AI models.

Food will be provided, and there will be SWAG (door prizes with GitHub merch, SWAG from our sponsors, and special Hacktoberfest goodies), so get excited!

Thank you to our generous venue sponsor: Arinco, and our food and drinks sponsor Twilio.

Talk details:

Talk 1:
Roll for Recap: Turning Session Chaos into Agent Powered Summaries by AJ Bajada

In Roll for Recap, AJ walks through how vibe coding can bring a brand new feature to life for tabletop play. This time the focus is a Session Notes and Recap Engine that takes messy player notes, structures them, and transforms them into clear and entertaining summaries using skills.md and open weight models. No prophecy from the DM, no ancient architecture scrolls, just curiosity, instinct, and a willingness to let the notes and the agent fall where they may.
AJ shows how raw session notes can become tactical recaps, narrative recaps, humorous recaps, and even radio show style summaries. Along the way you will see how skills.md shapes agent behaviour and how open weight models help the system stay consistent, expressive, and grounded in the rules and lore of the game.
It is a real example of how vibe coding, when done with intention instead of chaotic neutral energy, can build useful agent features and bring a tabletop world to life with surprising speed and style.

Talk 2:
Local AI: Twenty Gigabytes and suspicious amount of confidence by Dennis Vroegop

Somebody on your team has downloaded 20 gigabytes from Hugging Face and announced that you're now doing open-source AI. Nobody has read the licence. But it runs on a laptop, so apparently security is sorted too.
A remarkable amount of confidence to extract from a progress bar.
In this 45-minute session, we'll run models live on a laptop: no cloud, no API key, and no screenshot of something that worked last week. We'll cover what you've actually downloaded, why open weights aren't automatically open source, which models deserve your disk space, and what hardware you need. We'll also look honestly at speed, quality and cost.
Then we'll tackle what "local" doesn't solve. A wrong answer doesn't become right because you paid for the electricity. And keeping data on your machine is a useful guarantee, not your entire security strategy.
You'll leave knowing how to get useful results locally and when to choose local over cloud. With numbers and a working example. Which is a nice change from the unshakeable conviction of someone who has watched half a tutorial.