Accelerating development without losing control

The state of software development in 2026 is at a critical juncture. Although AI-assisted development tools promise unprecedented speed, without proper foundations, this velocity can become recklessness rather than progress. This hands-on workshop will demonstrate how to embrace modern development practices, including AI agents, while maintaining control mechanisms that ensure quality, architectural integrity, and long-term sustainability.

Participants will learn how to leverage automation and modern tools not as a replacement for discipline, but as a means of enabling higher-velocity development that is still grounded in rigorous quality assurance. The workshop focuses on the principle that speed without control leads to technical debt and system fragility, while control without speed fails to meet modern market demands. The way forward is to establish comprehensive automated safety nets that enable teams to work faster because they can trust their quality controls.

Through practical exercises involving PHP and industry-standard tools, participants will implement a comprehensive quality infrastructure that facilitates rapid development. The workshop progresses from traditional quality practices to their evolution in AI-assisted workflows, demonstrating test-driven development patterns in which humans write specifications as tests and AI agents generate implementations.

Participants will leave with skills that they can apply immediately to accelerate their development processes while strengthening, not weakening, quality controls. The workshop shows that the practices that enable safe AI collaboration are the same as those that enable high-velocity human development. This convergence is the true path to sustainable speed, achieving both velocity and quality through automation and discipline.

Key takeaways

Participants will learn that in AI-assisted development, speed and quality are not opposing forces but mutually reinforcing ones: by building comprehensive automated safety nets and treating tests as executable specifications that govern both human and AI work, teams achieve sustainable velocity grounded in rigorous quality controls rather than reckless technical debt.

Prerequisites

• A laptop with specifications adequate for running an AI agent alongside your development tools
• The ability to run PHP 8.5 code, either directly on your machine or in a container
• A modern PHP IDE or editor configured for PHP development (PhpStorm recommended)
• Git installed and configured locally
• An active GitHub account with working credentials (we'll do collaborative exercises)
• Active access to at least one AI coding agent — Claude or Junie. Other agents are fine too, as long as they support file-system access and manipulation, test execution, and multi-step autonomous tasks. (I'll be using Claude Code.
• A stable Internet connection for the whole workshop — we'll be reaching AI agent APIs, installing dependencies via Composer, cloning from GitHub, and running collaborative, real-time exercises

Clone the project and make the quality gate pass. Please run the following before the workshop:

     git clone https://github.com/thePHPcc/EventFlow.git
     cd EventFlow
     ./tools/composer install
     ./tools/composer ci

"composer ci" must finish green. If it does not, please reach out to me before the workshop so we can sort it out in advance rather than on the day.

You will need:

• PHP 8.5.7 or later
• The Xdebug or PCOV extension (needed to collect code coverage)

Everything else — PHPUnit, PHPStan, Deptrac, Infection, PHP-CS-Fixer and
Composer — is already committed to the repository as PHAR files in tools/,
so there is nothing else to download.

Can't install PHP 8.5.7 locally? No problem; the repository ships with a Containerfile so you can run everything in a container with Podman or Docker. The image is pinned to the exact PHP version we use and bundles the Xdebug extension for coverage.

Build the image once:

     podman build -t eventflow-ci -f Containerfile .

Then run the complete quality gate against your working copy:

     podman run --rm -v "$PWD":/app:Z eventflow-ci

A green run means you're ready. Replace "podman" with "docker" if you prefer; on systems without SELinux you can drop the ":Z" suffix from the volume mount.