Skip to content

Simulate and visualize Continuous Thought Machines (CTMs) — biologically-inspired neurons with memory and phase-based synchronization. A Streamlit app for exploring neural dynamics, built for learning, research, and a bit of fun.

Notifications You must be signed in to change notification settings

calvinx16/ctm-concept

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🧠 Continuous Thought Machines Visualizer

A hands-on project inspired by the Continuous Thought Machines (CTM) paper by Darlow et al.

This interactive simulation demonstrates the behaviour of biologically-inspired neurons that:

  • Process inputs over time
  • Store recent memory
  • Fire only when synchronized to a global rhythm

It’s a visual, intuitive way to explore concepts like memory, temporal weighting, and neural phase alignment — all from a browser.


Streamlit App


🚀 What This App Does

  • 🧠 Simulates CTM-style neurons with short-term memory and weighted history
  • 🔄 Synchronizes neurons via configurable phase cycles
  • ⚙️ Lets you control number of neurons, memory size, input stream, and step-by-step simulation
  • 🎨 Visualizes outputs over time using a dark-themed chart — great for demos or concept teaching

📌 Key Features

  • ✅ Step-by-step neuron activation with animation-style controls
  • ✅ Dynamic charts reflecting real-time neuron behavior
  • ✅ Clean Streamlit interface for showcasing machine learning concepts
  • ✅ Great starting point for further explorations into neural dynamics, adaptive compute, or temporal AI models

💡 Why This Project?

Inspired by the 2024 CTM paper, this project bridges the gap between neuroscience concepts and visual, code-based experimentation.

It’s built for:

  • Curious ML engineers
  • Students of neural computation
  • Anyone who wants to see what "thinking over time" looks like

📦 Getting Started Locally

git clone https://github.com/calvinx16/ctm-concept.git
cd ctm-concept
pip install -r requirements.txt
streamlit run app/main_app.py

🧠 Tech Stack

Python · Streamlit · NumPy · Matplotlib

Fully modular — core neuron logic lives in utils/neuron.py

No GPU needed, runs entirely in the browser or locally


📜 Credits

Built with neurons, caffeine, and curiosity
by Calvin Paperwala

This project was inspired by the paper:
Darlow, L., Regan, C., Risi, S., Seely, J., & Jones, L. (2025).
Continuous Thought Machines
arXiv:2505.05522

Special thanks to ChatGPT for co-developing, debugging, and hyping every neuron of this project.


About

Simulate and visualize Continuous Thought Machines (CTMs) — biologically-inspired neurons with memory and phase-based synchronization. A Streamlit app for exploring neural dynamics, built for learning, research, and a bit of fun.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages