Quick tip: 30-second sanity check before big AI changes

Hey Architects,

As projects get bigger, AI can make things confusing if it doesn’t understand your app. For example, you might end up with two similar backend files that do the same job.

To save yourself some time, before a major change or fix, ask this:

What’s the blast radius?
Is this the lowest-risk option?

This usually gives the LLM a chance to audit the code and present a lower-risk option.

Try it on your next risky change and tell me if it saved you a debug session.

—

BONUS:

Here's a prompt I've used for the past 6 months. It's worked well, but I use it less now that I ask those two questions above. Try both and see which you like better.

--- The Prompt:

"Sanity check for our review:

- Preserves all existing functionality

- Minimal change, no scope creep

- Reduces brittleness, doesn't add it

- Simplest viable approach

- No new source of truth unless explicitly requested

Confirm or provide an improved plan — do not implement."

1 comment

Quick tip: 30-second sanity check before big AI changes