Julien Genestoux

Julien Genestoux

@ouvre-boite.com

VP of Engineering at @relayProtocol! Previously, founder @unlock-protocol.com (aqui-hired by Relay in 2025), founder @superfeedr (RSS, websub), sold to @medium.com in 2016! Also goes by julien51 on the Internet! https://ouvre-boite.com

837 followers601 following1.1K posts

Julien Genestoux reposted

Hot take: the AI revolution isn’t about intelligence. It’s about headcount. Average reasoning, infinite copies. AI isn’t a breakthrough in intelligence. It’s a breakthrough in brute force.

Hot take: the AI revolution isn’t about intelligence. It’s about headcount. Average reasoning, infinite copies. AI isn’t a breakthrough in intelligence. It’s a breakthrough in brute force.

Julien Genestoux reposted

Truly wish Bluesky MFA offered TOTP, passkeys, or basically anything other than a code in your email.

There are many things to be worried about AI. I don't think "artificial" is it. Look around you, look at yourself. Everything is artificial, everything is man-made. Yes, even your intelligence, is, by definition man-made. Everything you do, know, think was actually done by you or other humans.

It doesn't matter that a model can't be tested because it will behave differently when it is being tested. In fact, I want models to always assume they are being tested so they always behave in an accountable way, because they expect accountability.

There are 2 kind of tests: 1. The "scaffolding tests" you create when you write a new feature. 2. The regression tests to make sure that thing won't break in the future.

Julien Genestoux reposted

One of my favorite side effects of AI is that it's making the web much more open again. For over a decade we've seen platforms increasingly become walled gardens. AI has created pressure for interoperability and we're finally seeing open standards and APIs being embraced again.

Julien Genestoux reposted

Next big thing for Leaflet: helping creators monetize! 💸 First step = paid memberships for your publications Next, exploring other things like one-off payments — tell us what you need! Almost ready & we'd love to find some early partners, please fill out the linked form if you're interested!

lab.leaflet.pubPiloting paid memberships for the open social web!Lab Notes 035: building monetization features for Leaflet creators, starting with paid memberships — tell us if you'd like to be the first to try, and what else you need!

I don't care about Fauci. The facts are: the US did worse than most OECD countries when it comes to health outcomes AND it also did better economically than these same countries. So the argument that the US went too far to protect lives at the expense of economic activity is pretty weak.

Maybe this is a controversial take, but trump and mamdani could both be economically wrong and politically correct.

Mike Masnick@masnick.com

Same people promoting this are screaming about Mamdani's grocery stores as "communism."

Same experience from me and the Relay team.

P(aul) Frazee@pfrazee.com

Accurate that it's a skill issue. Why is being a bit flippant about it but it's accurate. Good engineers know how to get good work out of llms; bad ones produce tech debt. Same as ever

The best maturity signal I found in engineers is whether they only start coding/fixing/refactoring once they actually have a very very good understanding of the problem. My team is working on a latency issue and it's fascinating to see some jump to "I will change X" vs. "I will measure Y"

I wish everyone in France was lucky enough to live away from France for a few years, if anything to understand how great France is, and how good they have it.

Same. If you use AI, you must be the first reviewer and (other) agents should review before humans.

Filippo Valsorda@filippo.abyssdomain.expert

Nuclear opinion: I don’t care if you use AI to write code,¹ but making a human review a change before an LLM is disrespectful. Why am I catching your bugs if a computer could have done it? It’s like asking to review code that doesn’t compile or pass CI. ¹ As long as you do it responsibly.

One thing that resonates with me a lot about global warming is that this year (or last) may the be hottest year on record so far, but it is also very likely the coldest of the next few ones. The (dis)symmetry is really what makes it obvious that we're going to face a lot of pain.

An interesting thing about agents vs other enterprise software is that most companies actually want to use 2 (or more) concurrently. In some strange way, in our org Codex and Claude are not competing. The more we use one, the more we use the the other one as well.

Until a few months ago, Iran only had a single card to play and that could only play it once: nuclear weapons. They now also can disrupt international fossil fuel markets at any point and put many countries in their knees, economically.

Exactly. I also want to believe that it's possible to make social media healthy, rather than outright prohibited. We should hold the platform accountable, rather than punish their users.

Sigh. A reminder that age-related social media bans require recording people’s age which means adding a layer of surveillance that makes vulnerable communities unsafe and creates a chilling effect on freedom of speech and particularly political speech.