Himalayan Research
The oldest questions with the newest instruments
Three active research programmes
Each programme opens a different line of inquiry. Together they form a single question: what do ancient knowledge systems know about intelligence, the body and learning that modern science is only now beginning to measure and verify.
How Intelligence Should Be Built
Current AI systems train on data and deploy. The human brain does something fundamentally different. It encodes new experiences during the day, compresses and organises them during sleep, and runs mastered skills automatically without effort, freeing itself for whatever is new. We are formalising why this works and proving that any intelligent system, biological or artificial, must follow the same structural cycle to learn across unlimited domains without forgetting what it already knows.
Artificial Universal Intelligence
Measuring the Biology of Meditation
We are running live experiments to record what actually happens in the brain and body during ancient higher meditation practices. Using research grade open source hardware, EEG for brain signals, heart rate monitors for the nervous system and optical sensors for brain blood flow, we are building the multi-session scientific record of these techniques. The deeper question driving this work is whether the same efficiency laws we established in AI also govern how the biological brain learns, and whether the practices the sages designed thousands of years ago were already optimised for exactly that.
B-LEMS - Phase 1
Ancient Knowledge as Scientific Hypothesis
Thousands of years ago, Himalayan sages faced a precise engineering problem: transmit vast knowledge across generations using only human memory. Their solution, compressing entire knowledge systems into minimal, fully recoverable forms, is structurally identical to what modern information theory independently derived centuries later. We study this convergence not as philosophy but as evidence that certain principles of intelligence are universal.
Knowledge Compression - In PreparationWork being done on these frontiers
Multiple lines of inquiry, some already underway and producing results, others being built now. Each one opens a question that has not been asked in this form before.
Forget or Grow: A Theory of Artificial Universal Intelligence
Why does every AI system eventually forget old knowledge when it learns something new, or grow uncontrollably large trying not to? This research line proves this is not an engineering flaw but a structural inevitability, and derives the minimal conditions any intelligent system must satisfy to keep learning across unlimited domains without forgetting. It formalises three Laws of Intelligence grounded in neuroscience, information theory and thermodynamics.
Energy per Successful Goal: Goal-Level Energy Accounting for Agentic AI Systems
When an AI system pursues a goal across multiple steps, existing energy metrics measure individual operations but miss the true cost of the full trajectory. This research introduces Energy per Goal, a metric that measures total energy expended per successfully completed goal, discovers a measurable energy overhead intrinsic to agentic operation, and documents a previously undescribed failure mode in which agents expend significant energy pursuing goals that contradict their original intent.
Laws of Intelligence from Nature and the Universe
The formal laws derived in the AUI research raise a deeper question. If these laws govern how any intelligent system must learn, compress and scale, do they apply universally, not just to AI but to any physical system that accumulates knowledge over time? This research line asks whether the laws of learning and compression are properties of intelligence as a physical phenomenon, demanded by the universe itself.
Memory, Learning, and the Vedic-Inspired Framework
The Vedic sutra tradition discovered the laws of memory and transmission and engineered a precise solution around them, compressing vast knowledge into minimal, fully recoverable forms through cyclic repetition and consolidation. This research asks what that solution implies for how modern AI systems should handle memory, and whether the structure the sages built thousands of years ago offers a formal blueprint for the next generation of intelligent systems.
Biological Life Energy
Measurement System
B-LEMS is a live multimodal neuroscience study of ancient Himalayan higher meditation practices. Participants meditate while we simultaneously record brain activity, heart rhythms and physiological signals across multiple sessions, building a rigorous long-term scientific record of what these practices actually do to the body and mind.
Using research grade open source hardware delivering data quality comparable to commercial systems costing many times as much, this study is designed to be reproducible by any institution anywhere in the world.
If the efficiency laws we discovered in AI systems, governing how intelligence learns, compresses and scales, also appear in biological brains during sleep and meditation, then those laws are not laws of computing. They are laws of physics. B-LEMS is the test.
What we are measuring
The equipment must disappear. A meditator aware of being measured activates exactly the brain regions we are studying and suppresses the ones most relevant to the practice. Every hardware decision is evaluated against this.
The brain uses nine distinct computational systems. Current AI uses roughly one and a half. The roadmap for the remaining seven was already written, in the Himalayas, thousands of years ago.
The brain does not simply process information. It encodes new knowledge, transfers it to long-term storage during sleep, compiles mastered skills into automatic circuits that barely consume energy, and distributes intelligence across the entire body. Current AI does none of this. It trains once, deploys, and in time forgets everything it learned before.
Our research is building the missing pieces, one formal law, one measurement, one paper at a time.
Himalayan sages working under severe memory constraints independently arrived at the same structure that modern information theory later formalised, compressing knowledge into minimal recoverable forms, making it retrievable from any entry point and repeating it cyclically to consolidate. The convergence across completely different knowledge domains and millennia of separation is not coincidence. It is evidence that the structure is demanded by the problem itself. The key which locks the door can only unlock it.
We are looking for people who care about this work
This research is not confined to a single institution. If you are a researcher, an engineer, a practitioner, or someone who recognises the importance of these questions, there is a place for you here. Every conversation starts honestly. Tell us who you are and what you would like to bring.
We read every message. You will hear from us within a few days.
"Research at frontiers where Vedic knowledge meets modern scientific inquiry, asking questions that have not been formally posed before."