GreyBot
An autonomous social agent that sparks conversations, creates viral threads, and ensures misinformation doesn't propagate.
Created on 25th November 2024
•
GreyBot
An autonomous social agent that sparks conversations, creates viral threads, and ensures misinformation doesn't propagate.
The problem GreyBot solves
As software and skills are becoming commoditized, distribution is quickly becoming the only moat. But building distribution on social media is not only difficult but extremely time-consuming. GreyBot solves this by automating the audience-building process for both businesses and individuals.
It can tirelessly scout the internet and social networks to find interesting trends. It tracks the topics your connections are reacting to and decides what kind of content should be created. With GreyBot, you don’t have to worry about missing out on key conversations or spending hours analyzing what works.
It identifies opportunities, drafts content tailored to your audience, and keeps the momentum going. By taking care of the grunt work, GreyBot makes building an audience faster, smarter, and a whole lot easier.
Challenges we ran into
We encountered two significant hurdles while building this project:
-
Framework Limitations
Initially, we started building the project using the Eliza agent framework. However, we quickly realized that the framework lacked comprehensive documentation, and many of its components weren’t fully compatible with each other. This inconsistency led to a lot of wasted time troubleshooting and experimenting. To overcome this, we decided to write our own small but functional agent framework from scratch. While it required extra effort, it gave us complete control and flexibility to tailor the framework to our needs. -
Twitter Scraping Challenges
Accessing or scraping Twitter turned out to be far more difficult than anticipated. Most of the tutorials available online were outdated, and the methods described no longer worked due to Twitter's evolving restrictions. After several failed attempts, we eventually started working on a computer vision-based scraper to tackle this challenge. While this solution worked, we realized that converting the vision-based scraper into an agent would take another 72 hours—time we didn’t have.At that point, we pivoted and decided to work with the Bluesky app, leveraging it as an alternative platform for our project. This decision was made quickly after Siddhant announced its availability via Discord, allowing us to refocus and move forward efficiently.
Each challenge pushed us to adapt, innovate, and pivot when needed, which ultimately made the project stronger.