Katalyst: The Zero-Entropy AGI Engine

Stillness floor

I. The Inspiration: Overcoming the Stochastic Barrier

In modern cloud and artificial intelligence systems, they suffer from severe "stochastic shaking"—inherent computational noise that manifests as AI hallucinations, endless processing loops, and massive GPU power waste. Standard frameworks attempt to resolve this by simply burning more electrical power, which leads to systemic decay. Our inspiration was to build Katalyst: The Zero-Entropy AGI Engine—an autonomous, self-correcting agentic pipeline designed to stabilize raw, chaotic data inputs before they ever reach the AI reasoning layers.

II. What It Does: The Self-Correcting Data Gate

Katalyst takes highly chaotic, unstructured inputs (such as unformatted text, logs, or system data) and instantly forces them to settle to the Stillness Floor (F_c=0). When a data packet is submitted, the engine automatically calculates its Torsion Gradient (\tau) representing coordinate instability : If this instability exceeds the infinite-pressure Buchdahl Limit (4/9 \approx 0.4444) , Katalyst executes a 90-Degree Vector Pivot. Using a custom Collatz (3x+1) loop , the engine rotates the expanding vector onto a stable, centripetal orbit , applying our Torsion Variable (\zeta_H = 0.001756) as geometric drag to slow and condense the signal : This forces all computational "shaking" to cancel out, freezing the transaction into a stable standing wave of zero entropy and generating a secure certificate ID anchored to our global constant, \Omega_G \approx 0.835102.

III. How We Built It: Shipped on an Old Phone and a Chromebook

The technical execution of this platform represents a monumental triumph of resourcefulness. This entire full-stack project was designed, coded, and deployed by a single developer working from an old cell phone and a basic Chromebook over public Wi-Fi. We bypassed the mobile web browser keyboard bugs that ignore "Enter" keys inside virtual terminals and defeated platform text-stripping filters by writing a custom, 100% bracket-free Streamlit interface. Using this mobile setup, we built:

The Interactive Frontend (app.py): An elegant dashboard featuring our custom Spacetime Siphon Visualizer mapped with matplotlib, letting users slide coordinate drift values and watch the geometric boundary warp and resolve in real-time.
The Backend Engine (backend/main.py): A lightweight FastAPI web server that acts as the core API tool our Google Cloud AI agent communicates with to execute vector stabilization on autopilot.
The Deployment: Streamlined and pushed live to the web via Railway under our custom railway.json configuration. #### IV. What We Learned We proved that advanced, abstract geometric physics is not just a theoretical concept—it can be successfully translated into clean, running Python code to eliminate AI hallucinations and computational waste. More importantly, we showed that you do not need a massive corporate budget or an elite engineering team to ship production-ready, low-entropy software. All it takes is the right math, a phone, and the drive to make it happen.

Built With

3.9+
architecture
cloud
cloud\omega-g
constant
custom
gemini
google
katalyst
ledger
pro+
python
run
sovereign
streamlit

Submitted to

Google Cloud Rapid Agent Hackathon

Created by

As the sole builder of Katalyst: The Zero-Entropy Sovereign Identity & Optimization Agent, my individual contributions spanned all layers of the technical design, mathematical integration, and cloud deployment:
1. Mathematical & Agentic Architecture
Engineered the KatalystAGICore: I designed and implemented the core KatalystAGICore class to serve as a self-correcting, low-entropy engine.
Implemented the 90-Degree Pivot: I formulated the mathematical logic for the 90-Degree Vector Pivot using the Collatz (3x+1) conjecture. This algorithm calculates the Torsion Gradient of incoming data. If the instability exceeds the Buchdahl Limit (4/9), it executes a right-angle pivot to force the chaotic inputs to stop shaking and resolve directly to the Stillness Floor (F_c = 0).
Lattice Alignment: I mapped this logic to our global Stability Constant (\Omega_G = 0.835102) and Torsion Variable (\zeta_H = 0.001756) to create a mathematically closed coordinate system.
2. Full-Stack Software Development
Built the Interactive Frontend (app.py): I developed the user-facing dashboard in GitHub Codespaces using the Streamlit library. I integrated a dynamic "Coordinate Drift" slider and used matplotlib to plot the centripetal flow of coordinates siphoning toward the central anchor.
Built the FastAPI Backend (backend/main.py): I built a lightweight FastAPI background engine to serve as the actual OpenAPI Tool that our Google Cloud AI agent communicates with. When the agent receives a raw query, it sends a secure request to this backend, which runs our katalyst_logic to stabilize the input.
3. Cloud Infrastructure & Configuration
Configured railway.json for Nixpacks: I authored the railway.json file to manage container building via the NIXPACKS builder and successfully configured our startup commands.
Resolved Port Binding & Private Networking: I configured our services to dynamically bind to the $PORT variable injected by Railway and enabled Private Networking to ensure that our public domain could locate and route traffic to our running Python container.
Dependency Management: I managed all required libraries inside requirements.txt to ensure a clean build, bypassing the mobile keyboard terminal bugs entirely.

Johnnie Hammons

Updates

Johnnie Hammons started this project — Jun 09, 2026 12:04 AM EDT

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.