AI policy

This page describes my current usage of AI (LLMs) when it comes to creating prose (for this site, and for communication, private or on social platforms) and creating code (in a professional setting, and for personal projects I open-source). This personal policy is subject to change.

Site content

All content on this site is certified human, written by me, Nicolas Hery.

I mainly write for myself. It helps me organize my thoughts, improve my writing skills, and solidify learnings. So using LLMs would defeat the purpose of those goals. I also think it would be a waste of a reader's time.

On some older posts (between 2023 and 2025), after writing the post myself, I have used LLMs to proofread and suggest simplifications or other corrections for any awkward sentences.

I've decided to not even use it for that anymore. So any posts from 2026 onwards, I will only use an LLM to fix spelling and grammar, nothing else. I apologies in advance for any awkward sentences, but that's how I write currently and hopefully I'll continue to improve.

Code snippets contained in a post are also hand-written by me, unless explicitly specified otherwise next to the snippet and with a good reason.

The only exceptions to the above are the intro in the About page and my resume, both of which also appear on my LinkedIn profile. For those items, I write the initial draft myself and then may use an LLM to edit and suggest improvements. This is in the interest of time and to optimize the content for career objectives.

Code

Professionally, I use LLM-based coding agents to generate code. That is just the state of the industry now.

For personal projects however (such as some of those I share on my GitHub), I tend to prefer writing the code by hand. The goal of most of these projects is for me to learn something, so actually doing the thing myself and writing the code down makes the learning stick.

In recent projects I include an "AI disclaimer" section in the project's README.md (example) to explain how LLMs were used for that specific project (for example as a tutor, a code reviewer, etc.).

On occasion I may "vibe code" a tool for personal use. I generally keep the repository private for those, as I don't think it is useful to share. It does not demonstrate any coding skills, and anyone can ask their own LLM to re-create the tool with their specific needs. If ever one of these is made public for any reason, it will be called out explicitly at the top of its README.md that it was entirely generated by an LLM.

Written communication

If you read an email from me, a LinkedIn private message, or any other social platform DM or post with my name on it, it will be written by me.

I do not run these messages through an LLM to proofread, so you will have the authentic message, typos and all, as I punched it in the keyboard with my own fingers.

There are only few exceptions to this.

I've used LLMs to generate some emails, for example customer support requests. This was mostly in the interest of time, and I apologize to any human support staff on the other side. I will try avoid this in the future, and if I absolutely have to, I will call out at the top of the message that it as generated by AI.

Also, for some posts on Twitter/X prior to 2026, I have used LLMs to proofread and help make the post shorter, but this was only after writing the initial draft myself. I've decided not do that anymore for any social media post going forward.