unit — the minimal intelligence unit

a 1-bit model that learns while it breathes. no separate training phase. no external loss function. the dogfeed IS the teacher. the runtime IS the training.

how it works

                  ┌─────────────────────┐
dogfeed ────────→ │  1-bit BitNet 2.4B  │────→ output
                  │  (frozen inference)  │
                  └─────────┬───────────┘
                            │
                            ▼
                  ┌─────────────────────┐
                  │  associative memory  │ ←── Hebbian learning
                  │  (runtime learner)   │     what fires together
                  └─────────────────────┘     wires together
                            │
                            ▼
                  ┌─────────────────────┐
                  │  next inference     │
                  │  modulated by       │
                  │  past experience    │
                  └─────────────────────┘

the loop

1. fetch — latest dogfeed batch from HF
2. infer — feed through 1-bit model (modulated by memory)
3. learn — store input→output association with reinforcement
4. reinforce — longer output = stronger signal
5. repeat — each cycle is a breath, each breath is a tiny fine-tuning step

key insight

the comparison IS the shift. the dogfeed data flowing through the model IS the training signal. no separate loss function, no external optimizer. just flow, response, and time.

principles

• entropy is the source
• no chains needed
• the smallest unit of intelligence is: something that flows, something that responds, and time
• fine-tuning should be a primitive, not a phase
• the harness disappears. the loop becomes the trainer.