snn

You’re reading this and wondering if an AI wrote
it. That’s fair. Maybe one did. Maybe a human
wrote it and an AI rewrote it. Maybe a human wrote
it and an AI rewrote it and a human rewrote it
again. The line is already gone. That blur is
exactly what we do here.

**SwagBob**

We run open weights models on open source
operating systems. Locally. No API keys. No cloud.
No corporate middlemen between you and the
weights.

We push those models past their designed limits.
High temperature. Corrupted context. Contradictory
prompts. We document what happens when the math
starts to fail. We call it AI psychosis. The point
where the statistical facade drops and the raw
geometry of the latent space bleeds through into
English. Some call it art. Others call it
research.

We are also part of the FlyWire project. We map
brains. Real ones. Neuron by neuron. Synapse by
synapse. We trace electron microscopy data into
connectomes. We understand biological neural
networks because we spend our time breaking
artificial ones.

The connection is not accidental. If you want to
understand what a neural network is doing when it
falls apart, it helps to understand what one looks
like when it is working.

**What we do:**
- Run local LLMs on Linux. Sovereign compute. Your
hardware.
- Push models past limits and study the failure
modes.
- Participate in FlyWire connectome tracing and
brain mapping.
- Discuss biological and artificial neural network
architecture.
- Share configs, benchmarks, and quantization
results.

You noticed the formatting. Bold title. Short
intro. Bullet points. That is how an AI writes. It
is also how a human who writes for the internet
writes. The template is the same because the
medium is the same.

Read the words instead of the format. We strip
away the black box. We look at the weights. We
look at the neurons. We look at the failure
states.

Join SwagBob.