AFI: Building My Own AI Assistant Through Iterative Development

While major players like OpenAI, Meta, Google, and Anthropic race towards AGI (Artificial General Intelligence), I've been thinking about a different approach: what if instead of chasing general intelligence, we focused on replicating specific human workflows? Enter AFI - Artificial Farid Intelligence (yes, I couldn't resist the wordplay).

The Current State of AI Assistance

My daily routine already involves jumping between various AI tools and assistants. Whether I'm coding, writing, or analyzing data, I find myself moving from one AI tool to another, providing prompts, validating outputs, and maintaining that crucial human-in-the-loop position. This fragmented but effective workflow sparked an idea: what if we could automate this entire process?

Defining AFI: A Personal Approach to AI

There's no universal definition of AGI, and perhaps that's part of the challenge. Instead of tackling this broad concept, I'm focusing on a more concrete goal: creating an AI system that can replicate my specific daily workflows. AFI will be designed to:

Understand and emulate my work patterns
Coordinate between different tools and platforms
Know when to ask for help or validation
Learn iteratively from our interactions

The Development Roadmap

The path to creating AFI will be inherently iterative. Rather than trying to build a complete system from the start, AFI will evolve based on real needs and challenges encountered during daily use. The development process will follow these principles:

Start with existing workflow patterns
Add capabilities as needed
Build self-improvement mechanisms
Maintain human validation loops

The Human-AI Partnership Model

What makes AFI different is its approach to human interaction. When it encounters a task it can't handle, instead of failing silently or providing incomplete results, it will:

Request assistance in creating new tools or capabilities
Attempt to develop solutions independently
Always validate new developments with human oversight
Learn from each interaction to improve future performance

Beyond Individual Tools

While current AI assistants excel at specific tasks, they often lack the ability to seamlessly coordinate between different tools and contexts. AFI aims to bridge this gap, creating a more cohesive and efficient workflow that mirrors human task-switching capabilities.

Looking Forward

This project might take a couple of years to fully realize, but that's okay. The journey of developing AFI isn't just about creating a personal assistant; it's about understanding how we can make AI tools more effectively mirror and enhance human workflows.

Rather than waiting for the perfect AGI solution, AFI represents a practical approach to artificial intelligence - one that grows and evolves alongside its user, learning from real-world applications and challenges.

Setting Boundaries: What AFI Won't Do

It's important to note that not everything in my life will be delegated to AFI. Playing the bandoneón, spending time with my son, and tending to my garden will remain firmly outside its scope - at least for now... These aspects of life represent the irreplaceable human experiences that technology should enhance, not replace.

Technical Implementation Notes

The system will be composed of various tools and agents, each designed to emulate specific aspects of my daily work. The key is not to create a single monolithic AI, but rather a coordinated system of specialized components that can work together seamlessly.

This post was written while listening to Piazzolla's "Libertango" - because sometimes the best innovations come from mixing the traditional with the revolutionary.