-What we need instead are areas where programmers can set up environments where they can safely learn from their mistakes. We need areas where developers can feel OK about trying new things. But we also need an area where developers can test those changes and ensure that they don't have other rippling effects on other code.
+We need environments where programmers can safely learn from their mistakes. We need spaces where programmers can feel good and confident about trying new things. We need places where developers can try out their ideas and not have those changes ripple out to other unrelated systems. This is the best way that developers can learn and be brave in their learning process.
+
+These environments must model the target systems, and they must be as close as is practical to those target systems. That doesn't mean you need to make exact copies of expensive production environments, but you do need to create models of production environments that test most of the pieces with which your code will come in contact. Having models that mirror production systems means that when you move your code to production you'll introduce fewer changes with unintended consequences. Your changes will have already existed in a production-like environment. You can take comfort in knowing that the changes you enact in these models will be the same changes that will appear on the target system.
+
+Ideally you'll need to have an environment like this on a machine that you control. This means that you're not competing with other programmers in your organization who are also being brave with their changes. You'll also want to ensure that your environment is kept up-to-date with their changes (and any production changes) so your development model matches what's on the target system and what will be on the target system. A good model is one that is kept current with what it is modeling. It's the same as a map of a city: it's best when it matches the area its modeling and is kept current with changes that occur in that city. A good map of the city might tell you about the recent construction happening on your route. A useless map doesn't even show your route because it wasn't built when the map was created. If our model of production is constantly falling behind what's in production we will spend more time rectifying the changes that we're making with the changes between our model and production.
+
+This also means having an environment that you can rebuild quickly and replicate as needed. Having a model that becomes its own separate reality becomes one more system to maintain. This model should be something that you can delete and rebuild at will in order to remove any previous experiments. It's best to think of it as an ephemeral copy of your target environment that has limited use and can be tossed when no longer necessary. It should be quick to replicate this environment so there's little friction in creating new environments to play in. That can mean scripting the building process for these environments. How you decide to do this is up to you but keep in mind that you want something that's as simple as you can make it and requires as little thought as you can manage to replicate it.