We used to stress out about whether a *warning* could possibly be interpreted in a way that would cause a new bug, and now intellisense is recommending I write a dozen new bugs every time I use VS. No wonder the old guard are vexed.

I’ve often wondered what a language that treated instrumentation as a first class concept might look like.

Because people leave their accounts signed in on publicly accessible computers and then complain when their money or identity is stolen.

AKA We cannot have nice things.

When you use an LLM to help you write a regex, now you have three problems.

If I’m knowledgeable on a subject I don’t *need* an LLM.

If I’m not knowledgeable on a subject I don’t *trust* an LLM.

Yes, absolutely this is what you should do.

If you don’t do this, how do you know that the data you are inserting matches what it will look like in production?

There are many different bad guys. The lock is keeping some of them out.

It could be 3 months or 3 years or 30 years out. The vast majority of those affected will not notice or take action until sometime between 3 months before and 3 months *after* implementation.

How often are you context switching into the specifics of how the ProblemDetails object gets created when you are reading this method? It’s important to you right now because you are new to the code. Will that still be true the 10th time you read this method, or the 100th?