Victoire Mathieu

Every time I ask AI for a quick code snippet, its default reaction is to tell me to npm install a brand new library even for basic utility stuff like array formatting or date math.

Before you know it the codebase is bloated with small single purpose packages that nobody wants to maintain. We already have solid built in helpers in our main codebase but the model completely ignores them unless I spell it out every single time.

How are you nudging your AI tools to look at what's already in the repo before pulling in extra dependencies?

Working with busy people taught me a hard lesson nobody has the time to read your backstory. I used to spend twenty minutes crafting the "perfect" email flow only to get a single-word reply back. At first it felt kind of rude, but then I realized they were just respecting my time by keeping it brief. Now I try to put the main ask right at the very top so there's zero guesswork.

What's your personal rule for keeping messages short without sounding abrupt or rude?

We ran into a near disaster last week when an agent was tasked with a routine API validation fix, but took it upon itself to "clean up" the database schemas and generated unrequested migrations.

It completely wrecked a couple of foreign key relationships because it thought it knew a better architectural pattern. Ever since then, our team had to lock down write permissions for any automated tool.

For those of you doing heavy backend work with AI, what kind of strict guardrails or sandbox environments do you use to keep the assistant away from your data layer?