Skip to content

Research

Research & publications

Questions about intelligence, machines, and the future — with publications on Zenodo, frameworks, and a book.

03 — Featured Research

A preprint on the geometry of learning

Preprint · 2026 · Zenodo

On the Spectral Dynamics and Eigenspace-Trapped Limit Cycles of SAM

Prabesh Dahal · 2026 · Zenodo

Theoretical analysis of Sharpness-Aware Minimization (SAM), deriving closed-form limit-cycle dynamics and convergence properties. The work characterises how SAM iterates become trapped in eigenspace-aligned limit cycles under specific spectral conditions.

Machine LearningOptimizationDynamical Systems
DOI: 10.5281/zenodo.20582640
Abstract visualisation of spectral dynamics and eigenspace-aligned limit cycles

Fig. 1 — Eigenspace-aligned limit cycles (concept)

Published via Zenodo — open access

Also published

Book

2025

The Last Human Protocol

Dahal, P. (2025). The Last Human Protocol.

A comprehensive exploration of human consciousness and its intersection with artificial intelligence. Recognised by the Asia Book of Records — earning the Grand Master title.

AI EthicsPhilosophy

Framework

2025

Seven Laws of the Robotic Age

Dahal, P. (2025). Seven Laws of the Robotic Age.

Seven core governing principles for AI and robotics development — addressing accountability, autonomy boundaries, human–robot coexistence, ethical deployment, and societal impact.

AI GovernanceRobotics Ethics

Theory

2025

Five Theories of Future Intelligence

Dahal, P. (2025). Five Theories of Future Intelligence.

Five original theories on how intelligence — human and artificial — will evolve. Covers emergent cognition, distributed intelligence, symbiotic learning, conscious machines, and post-human agency.

Future IntelligenceAI Philosophy