I built Local MI Lab because SELF-GROUND had become too heavy for the amount of mechanistic interpretability practice I actually had. The negation SAE work taught me a lot, but it also wrapped every …
Articles
Technical articles and public-interest analysis from GTCode, including AI systems research, information-control audits, trauma research synthesis, and computational investigation methods.
I did not get the result I wanted from my first serious mechanistic interpretability attempt. Probably the best thing that could have happened. I started with an attractive idea: take sparse …
[!NOTE] Scope and status: This article describes the design rationale, theoretical framework, and experimental protocols of the Bio-Digital Interface research program. The BARLI-QM and Q-CHIF phases …
Most AI systems are built for answer production. Ask a question, retrieve some documents, generate a response. If the retrieved material looks authoritative, the answer often sounds authoritative too. …
Harness engineering is the executable control layer around AI coding agents: repository knowledge, capability boundaries, context bundles, architectural invariants, runtime evidence, normalizing …
What if code review could work like editing a film rather than proofreading a manuscript? This article introduces the Cinema Debugger paradigm: a fundamental reimagining of how developers navigate, …
The field of mechanistic interpretability has matured rapidly over the past two years, transitioning from an academic curiosity to a critical component of AI safety research. As large language models …
What if AI systems could communicate the way neurons do—not through the bottleneck of language, but through direct geometric exchange of meaning? This article introduces the Global Cognitive Brain …
How do AI systems make sense of contradictory information? When multiple sources disagree, when interpretations conflict, when evidence points in different directions—what happens in the space …
Update: This article reflects the earlier CNS/SNO framing of epistemic fragmentation. The current CNS research line is CNS 7.1 / GCTS, which extends the work with record-access states, possible-world …