MOXXYLABS/Kern
Kern is a family of small expert models for code that run on your own machine. Every answer goes through the tools a developer already trusts: the compiler, the linter and the tests. Only what holds up reaches you.
Talk to us about KernBEFORE YOU SEE AN ANSWER
Kern works the way a developer does. It writes a change at the level of a function or block, compiles it, runs the linter and the tests, and reads the errors. When something fails, it repairs only the part that failed and checks again.
ARCHITECTURE
A compact open-weight general model, chosen by measurement. It stays loaded and runs offline.
A small adapter per language or stack. Switching experts swaps the module and keeps the trunk in memory.
The build, lint and test commands for each language, with diagnostics mapped to one format.
Picks the right module from the file and the project. When it is unsure, it asks you.
IT KNOWS ITS LIMITS
Small probes read the model’s internal state and estimate whether a request is in scope and whether Kern is likely to solve it. Kern then answers, asks one question or declines, and points you to a larger model when a task is too hard.
Kern is an expert in editing code: fixing build errors, changing functions, adding tests and fixing types. It does not chat about everything, and it does not replace large models on long, hard tasks.
Kern starts with JavaScript and TypeScript, and with React frontend code checked in a browser. The proof of concept runs on these today.
The same trunk with new modules and new validators. Each language brings its own toolchain as the judge.
Kern ships inside Moxxy, our open-source agent for developers. Compile, lint and test run as Moxxy tools, and consequential actions are shown before they run.
GOOD QUESTIONS
No. Kern is designed to run entirely on your device. Your code and data stay where they are, and it works without a network connection.
Kern is in development and runs as a proof of concept. It will ship inside Moxxy. If you want to try it early on your own codebase, write to us.
No. Kern is an expert in code edits in its languages. When a task is out of scope or too hard, it says so, and you can hand it to a larger model.
A small model runs on hardware you already have. With the compiler and the tests checking each answer, it does not need to memorize everything a giant model knows.
LET’S TALK ABOUT AI
Tell us about your domain, your data and how a good answer can be checked. We’ll tell you whether an expert model, a ready model or plain software fits best.
Let’s talkcontact@moxxylabs.com