At a recent conference in Europe, Linux kernel maintainer Greg Kroah-Hartman shared an interesting observation. After over a year of AI-generated pull requests and security reports that often missed the mark, he noticed a sudden improvement. The reports, which once earned the nickname “slop” for their poor quality, had become surprisingly useful in the last month or so. He guessed this shift was due to better tools and a deeper understanding of how to use them effectively. This change highlights how AI is starting to impact the way developers find and fix issues in code.

Large Projects and the Challenge of Critical Bugs

Big open-source and corporate projects have large teams and resources to handle security issues and bugs. When new problems emerge, they can patch things quickly thanks to their extensive manpower. These teams include both company employees and volunteers from around the world, working together to address vulnerabilities. But smaller projects often don’t have that luxury. Often run by just one or two people in their spare time, these projects face a real crisis of developer productivity. They need quick fixes, but lack the skilled developers to deliver them fast enough.

This creates a dilemma: how can small teams keep up with the rising tide of critical vulnerabilities? They need a way to boost their capacity without adding more human developers, which is often impossible. The answer might lie in leveraging AI agents—powerful tools that can act as force multipliers, helping small teams do more with less. By orchestrating teams of AI agents, even tiny projects could respond to issues faster and more efficiently.

Harnessing AI Agents for Better Coding and Security

AI agent frameworks like OpenClaw have become popular because they can coordinate multiple AI models and tools. These general-purpose applications are capable of managing complex workflows, but they can be expensive and sometimes produce inaccurate results due to hallucinations or errors common with large language models. Still, a structured approach based on a solid methodology and rich data sources can help teams improve productivity. Using code and API structures as a foundation, AI agents can assist in various parts of the software development lifecycle, from planning to testing.

What’s needed is a way to combine these tools with techniques like spec-driven development—where clear specifications guide the process—and agent orchestration. The goal is to give developers their own team of AI agents that work alongside them, tackling bugs, writing code, and managing documentation. This approach could help small teams stay ahead of threats, reduce technical debt, and keep pace with larger organizations that have more resources. The rise of AI-powered agent teams may soon become a key part of modern software development.

The Squad Project: A Team of AI Agents in Action

One promising example of this idea is Squad, an open-source project developed by Brady Gaster, a Principal PM Architect at Microsoft. Squad creates an agent team around GitHub Copilot, orchestrating multiple AI agents to help with various development tasks. With a simple command-line interface, developers can set up an entire team that includes roles like a lead developer, front-end and back-end developers, and even a test engineer.

The goal of Squad is to mimic a real development team working together on a project. Once installed, it manages the different roles, allowing each agent to focus on specific tasks. For example, one agent might handle coding, another tests the code, and others could work on documentation or deployment. This setup aims to make small teams more efficient and capable of handling complex projects without needing a large number of human developers.

By automating routine tasks and coordinating efforts between multiple AI agents, Squad represents a step toward more intelligent, scalable software development. It’s an example of how leveraging agent harnesses can help developers keep up with the demands of modern coding, improve security, and reduce technical debt. As AI tools continue to evolve, projects like Squad could become essential for small teams aiming to compete with larger organizations in the software world.

Inspired by

• https://www.infoworld.com/article/4164601/harness-teams-of-agentic-coders-with-squad.html

Sources

• github.com
• github.io
• github.io
• github.io
• github.io