I came across a post on Facebook summarising a recent talk by Kent Beck; it’s on YouTube too. He’s the creator of Extreme Programming (XP), a key champion of JUnit and Test-Driven Development (TDD), and one of the signatories of the Agile Manifesto. The summary of the transcript goes roughly like this:
AI output doesn’t mean correct → build tests, acceptance checks, and independent verification
Features consume the Future → check changeability after every round of feature delivery
Fast output doesn’t mean fast learning → keep time for experiments, retrospectives, and redesign
One-shot doesn’t suit every project → deliver long-lived systems incrementally
The spec isn’t always right → keep revising it as new evidence comes in
Output isn’t Outcome → track whether user behaviour actually changes
Outcome isn’t necessarily Mission → define the value the product ultimately wants to deliver
Leadership is more than handing out work → build a mission the whole team shares
This lines up with my own practice about 87 out of 100. But the one thing I want to call out is the Mission: how you define the value the product ultimately wants to deliver. In the course of putting it into practice, if you don’t keep challenging it, it’s very easy to be led along by AI and end up with thoroughly ordinary output.
To deal with that, and drawing on my experience working at UI, when I was practising and building heavily in March I defined my own development mode, BDD (Boss-Driven Development). So I basically run my development work on a TDD + BDD dual track. The point is not to let AI lead, but to have humans drive AI with bolder questions and goals.
So seeing the write-up of this talk reminded me of all that, and I talked it over with GPT. The result is in the image below: a bold idea…
BOSS-DD — that really is interesting!
Comments & Feedback