alis-sila/mitotic-transformer

Mitotic Transformer

A biologically & cosmologically inspired causal language model
based on the "Cosmology of the Living Cell" (Mother Theory)

Philosophy & Core Idea

This model is not a conventional transformer.

It treats reality as one single scalable biological system:

• Mitosis as the fundamental computational operation (Big Bang = cell division)
• Cytoskeletal Attention → equivalent to Dark Matter scaffold
• Osmotic Turgor Decoder → 70/30 expansion (Dark Energy analogue)
• F1-String Layer → hierarchical early scaling (atoms → cell → universe)
• Consciousness Module → White Hole Rendering + Biological GPU

"The universe is not a machine.
It is a living, mitotic cell — and intelligence is its natural expression."

Model Card

Field	Value
Model type	Causal Language Model
Base architecture	Custom Mitotic Transformer
Parameters	~125M – 1B+ (configurable)
Context length	2048 tokens
License	MIT
Language	Primarily English
Training data	OpenWebText + similar corpora
Intended use	Research, philosophical experiments, generative storytelling

Original Theoretical Works

This implementation is directly derived from the following publications by Alis Hasić:

• The Cosmology of the Living Cell (Mother Theory)
• A Mathematical Proof of Scale-Invariant Biocosmology
• Additional papers on Zenodo: Dark Matter as Cytoskeletal Scaffold, Dark Energy as Osmotic Turgor, Consciousness Module, F1-String, etc.

Code Structure

• modeling_mitotic_transformer.py: Full model definition (Mitotic Transformer with Causal LM Head)
• configuration_mitotic_transformer.py: Configuration class

Usage (after pretraining)

This is a from-scratch architecture. For real usage, pretrain or fine-tune it first using the provided training script. The current repository contains only the architecture definition (no pretrained weights yet).

from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("alis-sila/mitotic-transformer", trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained("gpt2") # or your fine-tuned tokenizer
inputs = tokenizer("The universe is a living cell. In this cosmic mitosis,", return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=50, temperature=0.85, top_p=0.92, do_sample=True)
print(tokenizer.decode(outputs[0]))

Note: trust_remote_code=True is required because this is a custom architecture.

Citation

If you use this model or the underlying theory in your research, please cite:

@misc{hasić2026_mitotic_transformer,
  author       = {Alis Hasić},
  title        = {Mitotic Transformer: A Causal LM based on the Cosmology of the Living Cell},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/alis-sila/mitotic-transformer}},
  note         = {Implementation of the Mother Theory}
}

@misc{hasić_mother_theory,
  author       = {Alis Hasić},
  title        = {The Cosmology of the Living Cell (Mother Theory)},
  year         = {2026},
  url          = {https://zenodo.org/records/18432564}
}

@misc{hasić_mother_theory,
  author       = {Alis Hasić},
  title        = {The Cosmology of the Living Cell (Mother Theory): A Mathematical Proof of Scale-Invariant Biocosmology},
  year         = {2026},
  url          = {https://zenodo.org/records/19017062}
}