Three projects are running through my mind right now:
1. WhatsApp Interface with LLM for Chat Analysis
The first project is all about building a WhatsApp interface powered by an LLM to analyze high-volume, high-definition chats. Here’s the plan for this:
- Define the Automation Framework: I’ll use a WhatsApp bot to scrape text messages as the foundation for this MVP. I can leverage a gadget extension to establish the bot and create the necessary framework.
- Data Processing through LLM: Once the messages are gathered, the LLM will step in to extract key insights and information. This could serve as a valuable tool for initial data processing and analysis, particularly for Independent Software Vendors (ISVs) looking to gain insights from chat interactions.
- Creating the MVP: For simplicity, I’ll start by using WhatsApp’s chat export function to build a basic version, focusing on text-based recognition to identify and extract relevant information. This means cutting out some intermediate steps to fast-track development.
- Security Layer: To keep it secure, I’ll add a username and password feature, even if it’s initially as simple as manually distributing passwords. This will give basic access control to the MVP.
Overall, this will allow me to create an interface that can parse WhatsApp messages with an LLM-backed layer, producing valuable insights with a fast turnaround time. The ultimate goal is to build a functional MVP that I can improve upon over time.
2. Incremental Updates for Product Introduction
The second area of focus is introducing the core product (an offshoot of Cheetah) to potential customers in a simple, non-pushy manner. Here’s how I plan to do that:
- Post-Iteration Showcasing: I realized that one reason for the delay in execution is the lack of visible progress. My goal is to socialize the post-iteration updates regularly to give a clear picture of the evolution and direction of the product.
- Incremental Education: I’ll continue to share small, digestible updates with potential customers, keeping them informed without overwhelming them. This approach will help build familiarity and gauge their response to new features.
- Feedback Loop: By presenting the product gradually, I can observe and measure their reactions and refine based on real-time feedback. This allows me to tune the product’s development in alignment with customer interest.
This slow-and-steady approach lets me introduce the product’s value without feeling pushy, creating a smoother path for adoption.
3. LLM for GitHub Repository Discovery
A third idea I’m considering is using an LLM to analyze and recommend GitHub repositories based on specific needs. Here’s the vision:
- Filtering Starred Repositories: Over time, I’ve starred thousands of repositories. The idea is to have an LLM scan through these, identifying details such as descriptions, URLs, and introductions to recommend repositories that match particular keywords or use cases.
- Targeted Search for React: For example, if I need a repository that facilitates React component integration on HTML pages, I’d simply search with “React in HTML,” and the LLM would highlight relevant options.
- Enhanced Discoverability: This would save time and make it easier to access and utilize repositories that meet specific project requirements.
Summary
Each of these projects represents an exciting opportunity to streamline workflows and enhance functionality. The WhatsApp interface with LLM promises to bring efficiency to chat analysis, while incremental updates will help me roll out the core product smoothly. And finally, an LLM-driven GitHub discovery tool could simplify my development process by quickly surfacing relevant repositories.
These are still works in progress, but the groundwork is set, and I’m eager to see how these ideas unfold.