TL;DR: This post is a case study of a social listening tool our team built during the Code4Compassion hackathon. In about six hours, we created a functional n8n workflow that monitors Twitter for content about factory farming, uses an AI model to score it for "outrage," and sends prioritized alerts. We are sharing our process and the open-source workflow as an example of how modern tools can be used to rapidly prototype potentially impactful projects. The goal is to offer inspiration and a technical breakdown for others in the EA community.
Acknowledgements: A huge thank you to the organizers of the Code4Compassion hackathon for creating the space for this project to happen. This was a team effort, and I want to thank my incredible teammates Adeniyi Joseph, Diana Rivera, and Shreya Shukla along with the Code4Compassion team for their collaboration and creativity. Thanks to Gemini for critiquing this post and refining the tone for a forum audience.
Hello everyone,
I'm sharing a project our team developed during the recent Code4Compassion London 2025 hackathon. Our primary goal in sharing this is to showcase what a small, focused team can build in a very short time and to provide a transparent look at how it works.
The Concept: An Automated Filter for Public Outrage
The project was designed to address a common challenge for animal advocacy organizations: filtering the noise of social media to find high-signal moments of public outrage that are worth amplifying. The idea was to automate the work of a communications team member who might otherwise spend hours manually searching for these opportunities.
Our Approach: Rapid Prototyping with Low-Code and AI
Our main takeaway was the power of modern tools to accelerate development. Working across several time zones, we went from idea to functional prototype in about six hours. We achieved this by using the low-code platform n8n to orchestrate the entire workflow, which allowed us to focus on the core logic instead of the technical overhead.
How the Tool Works
The system functions as a pipeline that ingests raw social media data and outputs prioritized alerts. In essence, the workflow:
- Searches Twitter for relevant keywords.
- Passes each result to a specialized AI agent for analysis.
- Scores the content based on sentiment, topics, and a specific "outrage" metric.
- Filters for posts that meet a minimum outrage score (>= 0.5).
- Alerts the team via two types of emails.
The entire logic is contained in this single n8n diagram:
The analysis step is powered by a specialized AI agent. We gave it the following prompt to ensure it returned a structured object we could easily use:
System Prompt:
"You are an expert in social media analysis. Given a social media post, return valid JSON with these additional keys:
{ "sentimentScore": a number between -1.0 (very negative) and +1.0 (very positive), "topics": a list of keywords relevant to the post, "outrageScore": a number between 0.0 (no outrage) and 1.0 (maximum outrage) }
Output only raw JSON—no markdown, no explanation, no code fences."
If a post crosses the outrage threshold, the system sends a real-time alert email for immediate action:
A separate branch of the workflow also compiles an aggregated email of all high-scoring tweets from the period, giving a broader overview:
The Key Takeaway: Inspiration for Rapid, Impactful Projects
We are sharing this to offer a concrete example of what is now possible in a very short timeframe. The barrier to building and testing ideas has never been lower.
This project demonstrates that with a clear goal and the right combination of low-code platforms and powerful AI models, a small team can create a surprisingly functional and potentially useful tool in a single day.
For anyone interested in the technical details or looking for a template for a similar project, we have made the entire workflow and documentation available as an open-source resource.
- View the project on GitHub: https://github.com/shreyashukla01/social-listening-for-public-outrage/
As a bonus point on rapid execution, our final presentation was also built with an AI tool and took only about 10 minutes to create.
- See the presentation here: https://gamma.app/docs/Real-Time-Outrage-AI-Powered-Alerts-on-Animal-Abuse-2l3e6bv67eok5f9
We hope this case study serves as inspiration for others to tackle problems and build their own rapid prototypes. Thank you for reading.