This is the dawn of a new age.
I have been observing software development for more than fifty years, ever since I wrote my first computer program. In that entire time, I have never witnessed a development that has changed the profession deeper, faster, or more pervasively than now. AI-assisted coding has escaped from the lab, and is impacting the work of every software developer.
In the communities I track I hear stories of software engineers regret taking a single week of vacation because of the innovations that they now have to catch up to.
I have lived through the revolutions of the Internet Dot-Com boom and cloud computing, but I feel this one is bigger and faster.
Over the past weeks (May/June 2025) we have seen software development agents mature. AI models are now prompted to write prompts for other AI models. These are AI agents on the next level. We can now orchestrate long running flows of code generation and improvement without our direct oversight.
Sure, many people point out that AI typically does not behave more thoughtfully than an “overenthusiastic intern, who’d happily push buggy code to production.” Failure rates are high, but that is typical of any new technology. At the same time, there are already people pointing out that this coding workflow allows them to write software they wouldn’t otherwise be capable of. That is effectively an “infinite” speed-up.
This too is typical of a disruptive innovation: it gets used in situations for which the previous technology was too expensive.
As for quality: we know how to produce good quality software that is also secure. It involves using test driven design, continuous testing, adherence to patterns and standards, and feedback loops. All these can be codified into AI agents and agent workflows. Here is a simple example. Recently, I had Jules, the background coding agent from Google, update a Dockerfile so that it no longer required root permissions. It worked like a charm. Imagine having AI agents continuously reviewing code bases for these types of hygiene patterns.
Few people currently have a complete understanding of where these developments are headed. Automatic coding agents have the potential to create or improve massive amounts of software. That software does not automatically have good quality, but, as sketched above, there are approaches to increase that quality.
But there is also a dark side. I shiver to think about the impact that this will have in the hands of malicious actors, from script kiddies to nation-states. The world of software engineering surely has changed.