In improvisational comedy, and even in brainstorming, there is a principle of “Yes, and…” that encourages people to build on each other’s ideas. “There are no bad ideas” is a common mantra in these settings.

In software development, however, this principle can be dangerous. Saying “Yes, and…” to every new feature request can lead to scope creep, or can lead to projects becoming bloated with features that were not part of the original plan.

It’s even worse if you never really had a plan in the first place.

Without a product vision, and a clear roadmap, it’s easy to take well meaning ideas, suggestions, or (in the case of OSS) Pull Requests and to merge every one of them into your product.

There’s one thing that can sometimes prevent this happening, and that’s the effort of implementing those ideas. Even merging someone’s code into your own can be time consuming.

However…with AI, LLMs, and coding agents the cost of implementing and merging every single well-intentioned suggestion has dropped significantly. It’s easier than ever to say “Yes, and…” and bolt another feature on.

Unfortunately, this often leads to products that are hard to understand, hard to explain, and harder for people to adopt. An effective one-use tool becomes a bloated and unwieldy mess that appears to have everything you want, but does none of them particularly well.

In an age of Agentic Software Engineering it is more important than ever to have a clear vision, a direction, and a strong opinion on where you want to go and what your software is aiming to achieve.

If you publish your vision and roadmap you can refer to them when you politely, but firmly, say “no”. In fact, you could even use AI to help you to say “no”.