Louis Beaumont

louis beaumont · founder of screenpipe

Hi. I'm Louis. Let me tell you why I'm building this.

I first touched a computer at 8. My grandfather had dozens of them — he was a hobbyist. I got obsessed with competitive games. Learned English trading items in Diablo 2 online. I was 8.

At 13, I was diagnosed with leukemia. The doctors said I might have 6 months left. My first reaction? I was glad I could skip school and spend more time playing online games.

During treatment — total body irradiation, chemotherapy — I became a top player at a chess-like game called Dofus. Built a 4-figure bot business for Diablo 2. I was 14, getting radiotherapy, writing code to make money.

I went through the sight of other kids dying, the loneliness of being a misfit, and a stem cell transplant from my brother.

I didn't survive cancer to play small.

After that, I pursued a Master's in France. Worked on satellite communications for intelligence use cases — I learned a lot about building things that actually work in high-stakes environments.

I've read 150 books a year for the past 6 years. Made 50K+ contributions on GitHub across 250 repositories. Woke up at 4:45am every day to go to the gym and then work until night. Built a few software businesses — some worked, some didn't. I learned a lot about making money, working with people, and building products that matter.

Along the way, I turned down OpenAI (twice, before ChatGPT), Cursor (when they were 5 people), and Mercor. Not because they weren't great — but because I had something I needed to build.

the thesis

Here's what I believe: prompting AI is great, but it will be 10% of how we leverage AI soon. The real unlock is letting AI watch what you do and work for you.

Your screen is the highest bandwidth signal of human intent — 10 million bits per second of what you actually care about. Yet nobody's truly leveraging it.

Human bandwidth is too limiting to manage all those agents. We need a way to scale our attention. I personally run 16 parallel agents in a split terminal doing the most important work — but I'd rather have AI just watch me and figure it out.

what already works

Screenpipe records your screen and audio 24/7 locally, and lets AI search or run scheduled agents on that data. Some concrete examples:

Sales

AI updates your CRM when you meet someone — on a call, in a LinkedIn chat, Discord, anywhere.

Engineering

Update GitHub issues and Linear based on screen and audio activity. No manual status updates.

Marketing

Find distribution opportunities in real time based on what you're working on.

what I believe

your data should never leave your machine

Privacy isn't a feature. It's the foundation. Everything runs locally.

do a few things very well

Less features, higher quality. Be fanatical about user experience.

open source isn't a strategy — it's accountability

If we say your data stays local, you can verify it yourself.

charge more, don't compete on price

Build something worth paying for. Then charge what it's worth.

talk to users every single day

Ship daily product demos. Fix bugs. Repeat.

I want to create abundance, bring humans to explore the universe, and live forever.

Screenpipe is how I'm starting — by freeing people from meaningless work so they can focus on what actually matters.

where we are

3.2K
weekly users
16.8K
github stars
90+
contributors
574
new setups/week
7x
engaged WAU growth
100%
organic

If any of this resonates — come build with us, use the product, or just say hi.