Hacktoberfest looks different this year.
Pull requests no longer count toward rewards. The focus is on building real projects with open-source and open-weight AI.
This guide is for beginners and experienced builders. I’ll explain the basic ideas first, then walk through a complete small project you can actually finish and submit in one focused session.
What Hacktoberfest 2026 Actually Is
Hacktoberfest is a month-long celebration of open source run by MLH and DEV (with DigitalOcean as presenting partner).
The 2026 theme is simple: AI belongs to everyone.
Instead of asking you to open four pull requests, the main in-person events (called Hack Days) ask you to:
Build something useful that uses open-source or open-weight AI.
Put the code in a public GitHub repository.
Give the project an open-source license (MIT, Apache-2.0, etc.).
Submit it through the official event system (OrganizerHQ Challenges).
Many Hack Days also have a partner track. If your event is a Google Gemma Event, there is a separate prize for the Best Use of Gemma 4. For that track the project must actually call Gemma 4 through the Gemini API.
Basic Concepts
Open-weight model
The trained model weights are available for people to download or run under the provider’s terms. Gemma is an open-weight model from Google.
Open-source project
Your source code is public and carries a license that lets others use, change, and share it. Common choices: MIT, Apache-2.0.
Gemini API
The official way to call Google’s models (including Gemma 4) from your code. You send a request and get a response.
Public GitHub repository
Anyone can see the code. This is required for the main Open-Source AI prize.
What “done” looks like for a Hack Day
A working demo + public repo + clear README that explains what you built and how the AI is used. You do not need a perfect production app.
How to Hack Like a Pro (Practical Tips)
Start with a tiny, clear problem
One sentence is enough.
“I want a tool that explains any image in plain English.”
Make the AI the core, not decoration
The model should do real work. If you remove the AI and the project still works the same way, it is not a strong entry.
Use an open-source license from the first commit
Add a LICENSE file early. MIT is simple and widely understood.
Write the README for a stranger
What does it do? How does the AI help? How can someone run it?
Keep the demo short and honest
Show the input, show the AI working, show the output. Two minutes is enough.
Ship something that works today
A small working tool beats a large unfinished idea.
Check the exact challenge rules for your event
For Gemma: the code must call Gemma 4. For the Open-Source AI track: public repo + open-source license + AI is important to how it works.
Building and Submitting a Real Project
The Idea (An example of an actual submission)
The idea is to upload any image (a sign, a whiteboard, a plant, a receipt, a UI screenshot, a handwritten note).
The tool uses Gemma 4 (via the Gemini API) to return:
A short plain-English explanation of what is in the image
Three useful next steps or questions someone might care about
Why this idea works for HacktoberFest:
Global (anyone can use it)
Easy to understand in 15 seconds
Multimodal (Gemma 4 handles text + images)
Can be built in a focused session
Clearly shows the AI doing real work
Step-by-step: Building It
- Create the repository
mkdir snapexplain cd snapexplain git init
Add a simple LICENSE file (MIT is fine).
Get a Gemini API key
Go to Google AI Studio, create an API key, and keep it out of the public repo(be sure your repo is public) (use an environment variable).Minimal working version (Python example)
`main.py
import os
from google import genai
from google.genai import types
client = genai.Client(api_key=os.environ["GEMINI_API_KEY"])
def explain_image(image_path: str) -> str:
with open(image_path, "rb") as f:
image_bytes = f.read()
response = client.models.generate_content(
model="gemma-3-27b-it", # or the current Gemma 4 model name available via the API
contents=[
types.Part.from_bytes(data=image_bytes, mime_type="image/jpeg"),
"Explain this image in plain English for a general audience. "
"Then give three short, practical next steps or questions someone might have."
],
)
return response.text
if name == "main":
print(explain_image("example.jpg"))
`
(Adjust the exact model name to the current Gemma 4 identifier shown in the Gemini API docs for your event.)
Make a simple interface (optional but strong)
A tiny Streamlit or plain HTML + JS front-end that lets someone upload an image and see the explanation is enough for a good demo.-
Write a clear README
`SnapExplain
Upload any image and get a plain-English explanation powered by Gemma 4.
How it works
- The image is sent to Gemma 4 through the Gemini API
- The model returns a short description + three practical next steps
Requirements
- Python 3.10+
- Gemini API key
License
MIT`
- Push to GitHub and make the repo public
- Submit through the event’s OrganizerHQ Challenges page
Select every challenge your project qualifies for (Best Use of Gemma 4 and/or Best Open-Source AI Project).
What Judges Usually Look For:
- Does the AI actually do meaningful work?
- Is the repository public and licensed?
- Can a stranger understand the project from the README and a short demo?
- For the Gemma track: is Gemma 4 clearly being called?
You do not need fancy design or a large team. Clarity and a working core matter more.
Final Advice
Treat the day like a focused build session, not a competition you have to win.
Pick one small idea.
Make the AI the center of it.
Ship a working version.
Write a honest README.
*Submit it.
*
Whether you are writing your first lines of code or you have shipped production systems, the same approach works: keep the problem small, keep the AI important, and keep the project understandable.
That is how you hack like a pro at Hacktoberfest 2026.
Yours,
devhabeeblateef,
Habeeb Lateef.
Top comments (0)