Hi, my name is

Rain Chong.

I research theoretical AGI.

AI researcher and developer exploring the theoretical foundations of general intelligence. I am working on my independent research full time after leaving my role as Product & AI Engineer at Fellou AI, formalizing a mathematical framework for cognition at the intersection of machine learning, cognitive psychology, and philosophy of mind.

Chat with My AI Clone (obsolete)× Close Chat

Hi, I am Rain’s AI clone. Ask me about his research, his work, or anything he has written.

About Me

Hello! My name is Rain, and my work focuses on understanding and constructing intelligence. What began as a curiosity at 16 about consciousness has grown into a formal investigation of intelligent systems, both natural and artificial.

I graduated from CUHK in 2025 with a major in AI and minors in Physics and Economics. I have worked across applied AI, research, and venture, including roles in automation system, quantitative trading, and early-stage company building. These experiences shaped a systems-level view of how intelligence can be engineered and deployed.

My current research examines the topological and functional structure of cognition, with the goal of modeling learning, memory, and awareness as transformations on evolving state spaces, in the pursuit of general intelligence. I also write blogs regularly on AI, consciousness, and the future of intelligence.

Here are a few technologies I’ve been working with recently:

  • Python
  • C/C++
  • PyTorch
  • Node.js
  • LangChain
  • NestJS

Where I’ve Worked

Independent Research in AGI @ Stealth

Mar 2026 - Present

  • Building a machine learning paradigm that models the structure of human cognition directly, instead of approximating functions by gradient descent. Target is a model homogeneous to human intelligence in organisation, evolution and topology. Unpublished.
  • Learning runs purely forward and locally, with no backpropagation, no backpropagation through time, and no target or reward function supplied from outside. Feedback is endogenous: the model learns what counts as good rather than being told.
  • Internal structure self assembles from the temporal structure of input, so unsupervised, reinforcement and supervised learning arise as stages of one process rather than as three algorithms chosen in advance. Structure and its evolution are interpretable by construction, not by probing.
  • Local updates make the model parallelisable, distributable, and implementable on analog and neuromorphic hardware.
  • Core tools: non-autonomous dynamical systems, functional analysis, generalised hypergraphs, coupled systems, differential geometry. Drawing on physics, philosophy and psychology.

The Question I’ve Been Working On Since I Was 16

“What is intelligence, actually?”

Everyone is building toward it. Nobody knows what they are building toward.

I started this work out of curiosity about psychology and philosophy, before I knew it had anything to do with AI. When I studied different schools of thought, I found that all of them held themselves to be correct, but they contradicted one another. Rather than adopt one, I wondered what sits underneath all of them.

I realized that different thoughts are just structure held inside a human mind as an instance of cognition, and I see intelligence as the process of approaching a more general structure that can contain different thoughts rather than collapsing into one. The core of this work is to determine what human intelligence actually is, and to build a mathematical foundation for it.

Read the research brief (coming soon)

I Write about AI, Philosophy, and More

View all blogs (5)

Some Things I’ve Built

Other Noteworthy Projects

View all projects (9)

What’s Next?

Get In Touch

Whether you have a question or just want to say hi, I’ll try my best to get back to you!