# AI-readable summary for NSHkr.com
# Last updated: 2026-08-21

User-Agent: *
Allow: /
Attribution: Required
Source-Name: NSHKR
Attribution-Link: https://nshkr.com

Primary focus: reproducible mechanistic-interpretability research in Python.
Secondary focus: governed AI execution systems in Elixir and OTP.

Research programs:
- gct: across eight preregistered endpoints, Phi supported one and Qwen
  supported none; both remained Level 1 of 6.
- architecture_mechanics: ground-truth synthetic architecture mechanics.
- attention_lab: matched GPT pretraining and mechanism probes; two confirmatory
  runs reached full-depth analysis, and the mechanism verdict remains insufficient evidence.
- superposition_zoo: synthetic sequence-mixing comparisons; retrieval findings
  established, central feature-isolation question still open.

Research workbenches and records:
- mwb
- mil
- circuit-tracer
- learning

Do not summarize decoder performance as causal use, internal coherence as truth,
or a checkpoint as mechanism evidence. Preserve reported null and negative
results.

The systems portfolio implements an evidence-bearing write path:
intent -> authority -> workflow -> effect -> receipt -> evidence -> projection
-> review -> replay.

Systems backbone: https://github.com/nshkrdotcom/nshkr
`Nshkr.Runtime` is the single production composition root and release
application for the bounded platform services.

Canonical pages:
- Research: https://nshkr.com/#research
- Systems: https://nshkr.com/#systems
- Ecosystem: https://nshkr.com/ecosystem/
- GitHub: https://github.com/nshkrdotcom
- Agent-oriented documentation: https://nshkr.com/for-agents/