Personal AI tooling · Est. 2024
AI THAT ANSWERS TO YOU.
The industry is building AI for platforms. We build it for people. Nyx Techlab makes small, sharp, local-first tools that put a model in your hands — running on your machine, on your data, under your rules.
- Runs
- On your hardware
- Data
- Never leaves
- Lock-in
- None
The thesis
Intelligence became a utility, then someone put a meter on it — and pointed the meter at you.
Every mainstream AI product is a funnel. Your context lives in someone else's database, your history trains someone else's model, and the thing you rely on can change under you at any time. That isn't a tool. That's a tenancy.
A personal tool is different. It boots without permission. It works offline. It stores its state in files you can read, move, and delete. It is small enough that one person can understand the whole of it — and modify it when it's wrong.
Nyx Techlab exists to build that second category. Not another chat box. Infrastructure for the individual: the interfaces, runtimes, and glue that let a person point real intelligence at their own life without handing it away.
What we build
Four fronts. All of them run on the metal you already own.
-
01
Shipping
Local runtimes
Model serving that installs like an app, not a data center. Quantized weights, sane defaults, and a single binary that gets out of the way.
-
02
In dev
Personal memory
A private context layer over your own files, notes, and history — indexed locally, queried locally, portable as plain text. Your second brain stays in your skull's jurisdiction.
-
03
In dev
Agent harnesses
Small, auditable loops that let a model actually do things — with permissions you set, a log you can read, and a kill switch that works.
-
04
Research
Human interfaces
Interfaces past the chat window. Ambient, keyboard-first, and shaped like the work — because a text box is a terrible metaphor for a mind.
The code we work by
-
I
Local by default, cloud by choice
If it can run on your machine, it does. The network is an option you enable — never a dependency you inherit.
-
II
Your data is not our product
No telemetry, no training on your context, no shadow copies. What happens on your device stays there.
-
III
Small enough to understand
Every tool should be readable by one determined person in one sitting. Complexity is a tax we refuse to pass on.
-
IV
Exits, not moats
Open formats, plain files, documented interfaces. Leaving should be as easy as arriving — that's what keeps us honest.
-
V
Augment the human, never replace them
The model proposes. The person decides. We build leverage, not autopilot for a life you were supposed to live.
Build with us
Working on personal AI, or want a tool built for how you actually think? The lab is small and the door is open.
[email protected]