Old Laws, New Colleagues

Brooks's Law, Conway's Law, Goodhart's Law: for decades, laws like these have distilled what generations of developers learnt the hard way. They are not laws of nature but empirical observations about complexity and about people who build software together. Now AI agents write code, and the question is on the table: do these laws still hold?

This talk gives a differentiated answer. Some laws remain untouched because they are about software, not about how it is developed. Others hit harder than ever: if debugging is twice as hard as writing code, and writing suddenly costs almost nothing, then understanding becomes the bottleneck. And when an agent is given a metric as a target, it optimises that metric more literally than Goodhart could ever have dreamed.

It gets particularly interesting with the laws about teams. What does Brooks's Law mean when the new team members need no onboarding but generate review load? What does a system mirror, following Conway's Law, when prompts and context files are part of the communication structure? And what is the bus factor of code that an agent wrote and no human understands?

You will leave this talk with a map: which laws still give you orientation today, which ones you need to reread, and why the oldest observations of our field explain best what matters when working with AI agents.

This presentation currently exists only as an idea and has not yet been accepted by a conference.

About me

I created and maintain PHPUnit, the de facto standard testing framework within the PHP ecosystem. It is used by millions of developers and is embedded in the build pipelines of start-ups, Fortune 500 companies, and public sector organisations alike. I serve on the board of the PHP Foundation and am a co-founder of thePHP.cc, where I advise organisations on testing strategy, software quality, and security.

With over 25 years' experience in developing the testing and analysis tools that the PHP ecosystem relies on, I have witnessed the laws of software engineering come into effect through every new wave of tooling. This talk is informed by my experience of delegating non-trivial development tasks to AI agents, as well as by my daily work maintaining a project that millions of developers trust. This background influences my writing and speaking on software quality, keeping humans in the loop, and the practices that make AI-assisted development trustworthy.

Upcoming events

More events
Back to overview