The Thinking Archive

Notes from the search for general intelligence

Bengaluru, India
Est. 2026

From the archive

The brain is the only existence proof we have that general intelligence is possible. I am trying to understand how it works and build something inspired by what I find. These are the notes from that search.

Amritansh — AI Researcher
Essays & Research Notes
01
Metascience · Systems Thinking · Neuroscience
Why Science Gets Stuck
Modern research may be optimizing the known while neglecting the possible
I was doing nothing. Not reading a paper. Not running an experiment. And yet something clicked that months of structured work hadn't produced. The central bottleneck in science may not be intelligence, effort, or funding. It may be search.
April 2026
Astera Essay Competition
15 min read
02
Continual Learning · Phase Transitions · Neuroscience
Catastrophic Forgetting Is Not an Optimization Problem
Every standard fix is treating the wrong thing — the geometry, not the gradient
Every approach to catastrophic forgetting — elastic weight consolidation, replay buffers, adapter methods — treats it as an optimization failure. The brain treats it as a phase transition. These are not the same problem. They do not have the same solution.
May 2026
Continual Learning Series
20 min read
03
AGI · World Models · Causal Reasoning
The Oracle Problem: Why World Models Require Continuous Learning
A world model that cannot update without forgetting is not a world model. It is a snapshot. The difference between a snapshot and a model is the difference between a photograph of a river and an understanding of water.
Coming soon
18 min read
04
Experiment Log · CH-HNN · SAGE
23 Experiments: What Failure Taught Me About Hippocampal-Cortical Consolidation
A structured log of five months of continual learning experiments — including the 250 hours that failed. What each failure ruled out. What the Pareto improvements actually means. Published because the failures matter as much as the results.
Coming soon
25 min read

The Thinking Archive is a record of one researcher's search for the principles underlying general intelligence.

I am Amritansh, an independent AI researcher based in Bengaluru, India, working at the intersection of computational neuroscience, mathematics, physics, and machine learning.

What I have is a deep conviction that the brain — the only existence proof of general intelligence we possess — contains mathematical principles that current AI has not yet formalised. My research attempts to extract those principles and build with them.

This archive exists because research produces more than papers. It produces failed experiments, half-formed ideas, unexpected connections, daydreams that turn out to matter. Everything here is published openly, for anyone who finds it useful.

The archive does not forget. That is the point.

"The brain is not a computer that happens to be made of meat. It is a dynamical system that has learned, over five hundred million years, how to stay on the edge of a phase transition — just organised enough to act, just disordered enough to adapt."

Current AI optimises loss functions.
Biological intelligence navigates bifurcations.
These are not the same problem.
They do not have the same solution.

My research is an attempt to close that gap — formalising the brain's plasticity-stability mechanism as a mathematical framework for machines that learn without forgetting.