So there’s this programming language called Jacquard that I can’t stop thinking about. Not because it’s another language you need to learn — we’ve got enough of those. But because it’s the first language I’ve seen that’s designed from the ground up with the assumption that AI is writing most of the code.
And that changes everything.
The Problem Nobody’s Talking About
Here’s the thing. Right now, when ChatGPT or Claude spits out Python code, you review it. You look for bugs, logic errors, the usual stuff. But there’s something you can’t really check easily — what side effects does this code have? Is it phoning home somewhere? Does it make network calls you didn’t expect?
In Python, any function can do anything. A function called calculate_discount could literally be sending your database to a server somewhere. You wouldn’t know until you traced every single import.
Jacquard says: nah, that’s silly. Let’s make side effects trackable.
Effect Signatures — Like Types, But for Honesty
Jacquard uses something called effect signatures. Think TypeScript’s type system, but for operations. A function’s signature tells you exactly what it touches — the file system, the network, the database, randomness. Everything’s explicit.
If a function called formatDate somehow makes an HTTP request, the compiler catches it before it ever runs. It’s dishonest code that can’t hide.
This is huge when you’re reviewing AI-generated code at scale. Imagine a team that accepts 60% AI-written code — Airbnb says that’s already their reality. You need mechanical guarantees, not just human eyeballs.
Identity That Actually Means Something
Jacquard also does something clever with canonical identity. Every entity in the system has a globally unique, permanent identifier. Not a string that could change when someone refactors. Not an address tied to a server. A real identity that stays the same across deploys, across databases, across time.
In traditional languages, you build this yourself — and usually mess it up. Jacquard makes it the default.
Probabilistic Programming, Built In
And here’s the wild part — probabilistic programming is right there in the language. You can write code that goes “if there’s a 70% chance of rain, carry an umbrella” and it’s not some hacky random library call. It’s a first-class language construct.
Because Jacquard’s designers realized: software in the real world deals with uncertainty all the time. Weather predictions, fraud detection, recommendation algorithms. Why pretend the world is deterministic when it isn’t?
The Honest Truth — There’s a Catch
Jacquard has a chicken-and-egg problem. Large language models are trained on millions of lines of Python, JavaScript, Rust. They’ve barely seen any Jacquard code. So today, asking Claude to write Jacquard is like asking it to write in a language it read one Wikipedia article about. It’ll be creative. In the bad way.
But that’s fixable. As more people write Jacquard, the models will learn. And the bet Jacquard’s creators are making is that languages designed for the human-AI pair — where verifiability is baked in, not bolted on — will win in the long run.
So… Should You Learn It?
Not today, unless you’re curious. But watch it. The languages we use right now were built for a world where humans wrote every line. That world is ending. What comes next should be designed for the world we’re walking into — where AI and humans write code together, and trust needs to be mechanical, not blind.
Jacquard might be the first real glimpse of what that looks like.