KernScript site is officially live 🚀
Just a frontend and a simple backend, but it’s the very first project ever built with KS.
More to come.
Don't forget to join the waitlist!
https://t.co/LAscQzNQbW
... allows LLMs to express and execute logic with significantly less syntactic overhead.
By removing unnecessary complexity and focusing on the core information an AI needs, KernScript aims to make every token more valuable.
(3/3) 🧵
AI token costs are becoming a major expense for most businesses. There are many layers where tokens are wasted, and one of the most overlooked is the programming language itself. We can’t realistically learn a more token-efficient spoken language, but LLMs can operate...
(1/3) 🧵
... using programming languages that are far more compact and efficient, reducing the amount of tokens needed to communicate instructions, logic, and context.
This is precisely where KernScript comes in. KernScript is designed as a simpler, more compact language that...
(2/3) 🧵
The first two real projects built with KernScript are:
1️⃣ KernScript site
2️⃣ KernScript Studio frontend & backend: UI agent that allows building KernScript projects with different LLM connections.
Neither these projects or KernScript are public yet.
Stay tuned! 🚀👀
Best part: it all compiles to JavaScript or Typescript you can read and audit. Nothing magic, nothing locked in. Just a smaller, safer way to let an AI write real software.
🧵(4/4)
What if the language your AI wrote wasn't JavaScript, but something built for an AI to write and save you money?
That's KernScript (ks). It is created for AI not struggling and still compiles to plain, readable JavaScript or TypeScript. Same program, far fewer tokens.
🧵 (1/4)
How much does it save?
In our tests, KernScript reduced token usage by 48% to 91% across several LLMs, including Claude, Gemini, and OS models, in automated sessions of 100k tokens or more.
For smaller tasks, KS may be less efficient due to the skill's token overhead.
🧵(3/4)