Sam Schillace explores the emergence of AI-driven software development teams achieving exponentially growing productivity through innovative approaches to coding and system design. These compounding teams represent a technological transformation comparable to major innovations such as personal computers, web browsers, or smartphones.
The distinctive characteristics of these teams are striking. Contrary to popular belief, they do not use basic AI coding tools like GitHub Copilot. Instead, they build complete, sophisticated frameworks around AI models, creating proactive systems endowed with independent capabilities. Their approach relies heavily on low-level programming tools such as git, markdown, and Kubernetes.
The working methodology adopted by these teams is characterized by a recursive approach: "build a tool to build a tool." This philosophy of extensive automation allows AI systems to create and implement their own tools autonomously. Problems are broken down into modular, solvable pieces, facilitating massive parallel development. Schillace mentions teams running 5 to 10 parallel processes simultaneously, with daily spending reaching hundreds of dollars in API calls.
A particularly telling element is that some of these teams have not directly touched code in months. Instead, they orchestrate AI systems that generate, test, and deploy code autonomously. In this new paradigm, coordination becomes the primary challenge rather than writing code itself.
The organizational structure of these teams differs radically from traditional development teams. They prefer to stay small, composed mainly of senior designers capable of designing architectures and systems rather than coding line by line. Strict modular boundaries are essential to allow different components to function independently while integrating coherently.
The technical workflow relies on constant code execution and continuous acceptance testing. This approach ensures that automatically generated systems meet the functional and quality specifications defined by human designers. The Amplifier framework is mentioned as a concrete example of these new architectures.
Schillace suggests that this transformation will likely extend beyond software development to affect knowledge work in general. The ability to create systems that improve and develop autonomously represents a fundamental shift in how creative and technical work is accomplished.
The author concludes with conviction that "the new, better way to do things is now very clear," signaling that we are at an inflection point where these approaches will move from experimentation to widespread adoption. Organizations that fail to adapt to this new reality risk being outpaced by those that fully embrace the potential of AI-driven compounding teams.