mirror of
https://github.com/hanzo-docs/docs.git
synced 2026-10-29 00:27:34 +00:00
60 lines
2.5 KiB
Plaintext
60 lines
2.5 KiB
Plaintext
---
|
|
title: Why We Open-Source Everything
|
|
description: The case for radical openness in AI — why releasing weights is the right thing, strategically and ethically.
|
|
author: Zen LM Team
|
|
date: "2026-01-20"
|
|
tags: [open-source, philosophy, community]
|
|
---
|
|
|
|
Every time we release model weights, someone asks us: "Why would you give this away?"
|
|
|
|
The answer is simple: **the open frontier is how we win**.
|
|
|
|
## The Proprietary Trap
|
|
|
|
Keeping model weights closed creates short-term competitive moats that collapse. History shows this repeatedly: closed systems eventually face open alternatives that are "good enough" — and then the network effects of the open ecosystem compound until the proprietary system can't compete.
|
|
|
|
We've seen it with operating systems, databases, web servers, and now AI infrastructure. The question isn't whether open models will match proprietary ones — it's when.
|
|
|
|
By building openly from day one, we're on the right side of that curve.
|
|
|
|
## What We Keep Private
|
|
|
|
We're not naive about intellectual property. There's a meaningful distinction between:
|
|
|
|
- **Model weights** — released openly under the Zen Open License
|
|
- **Training data** — proprietary curation and filtering pipelines that we've invested in heavily
|
|
- **Infrastructure** — our serving stack, routing optimizations, and operational knowledge
|
|
- **Research** — papers we publish (soon™), insights we've earned through iteration
|
|
|
|
The weights are the democratizing artifact. The infrastructure and know-how are where sustainable business value lives.
|
|
|
|
## The Safety Argument
|
|
|
|
Some argue that open-sourcing powerful models is dangerous. We take safety seriously — which is why we think openness is the *safer* path.
|
|
|
|
Closed models create information asymmetries where only the developer knows what the model can and can't do. Open models allow:
|
|
|
|
- Independent red-teaming and vulnerability research
|
|
- Academic study of emergent behaviors
|
|
- Community development of alignment techniques
|
|
- Regulatory oversight that's grounded in facts, not marketing
|
|
|
|
Monsters hide in the dark. The best disinfectant is sunlight.
|
|
|
|
## The @zenlm Org
|
|
|
|
All Zen open-source code lives at the `@zenlm` npm org and the `zenlm` GitHub organization. This includes:
|
|
|
|
- `@zenlm/models` — canonical model catalog with specs and pricing
|
|
- `@zenlm/ui` — React components for model libraries and cards
|
|
- Inference server code, fine-tuning recipes, and more coming
|
|
|
|
Install and build with us:
|
|
|
|
```bash
|
|
npm install @zenlm/models @zenlm/ui
|
|
```
|
|
|
|
The open frontier needs builders. Come help us build it.
|