nshkr-research

Hardware-Secured Temporal Blockchain

Chiral Narrative Synthesis - Dialectical reasoning framework for automated knowledge discovery

Evolutionary Autonomous Development System

Phoenix LiveView interface for CNS dialectical reasoning experiments

Mechanistic Workbench (mwb): Local-first mechanistic interpretability workbench for IPython research, agent-readable state, artifact validation, evidence graphs, claim-safe MechanismCards, provenance, run ledgers, SAE/TransformerLens workflows, and reproducible MI experiments

Leakage-resistant experiments testing contextual transport and latent-coordinate structure across Qwen3-4B and Phi-4-mini, with frozen controls and reproducible reports.

Local mechanistic-interpretability workbench with IPython capture, activation patching, SAE analysis, controls, and reproducible visual artifacts.

Deterministic epistemic investigation core, causal hypothesis exploration engine, and strategic-network benchmark evaluating structured causal reasoning against monolithic LLM baselines.

Synthetic ground-truth lab for testing how sequence-mixing architectures shape transport, feature geometry, overwrite behavior, and causal legibility.

Mechanistic-interpretability experiments, research records, and ML visualization tools, including SELF-GROUND and construct-mismatch studies.

Controlled GPT pretraining and mechanistic-interpretability harness for testing alternative attention architectures against matched baselines.

Anti Agents - Inspired by Sakana AI's String Seed of Thought paper

A research harness for studying how sequence-mixing architectures (softmax attention, linear attention, state-space models, delta-rule associative memory, and more) causally shape superposition and feature isolation in learned representations, using ground-truth synthetic benchmarks instead of natural-language probing.