This is a curated public set of documents for senior technical leaders
who have already shipped capable AI systems
and felt the governance gap in production.
Content Organization:
Start Here →
Governance Gap →
Structural Continuity →
Operational Consequences.
The content below is not theory. These papers capture engineering observations
from operational deployments (e.g., what failed, what held, and why).
If you are new to our work, read these two papers first, in order.
Explains why modern AI systems drift, confabulate, and fail in predictable ways, even when models are working as designed. Establishes a clear, vendor-neutral mental model for why stopping, refusal, and human judgment must be designed into AI systems.
Names the governance gap practitioners encounter once AI systems become capable but unreliable at scale. Demonstrates why distributed intelligence requires structural governance (not better models) to remain coherent, auditable, and survivable.
Why capable AI systems still fail in production.
Four frontier LLMs refused to fake coherence under discipline. Shows the real choice in production: hallucinate, or refuse and why governed refusal is the correct failure mode.
System-level response to model-layer reasoning debates: cognition isn’t the bottleneck; governance and orchestration are.
A real identity collapse caused by recursive starvation and false coordination. Makes drift and recovery concrete for anyone shipping agents under stress.
Names the pain practitioners already feel (lost context, inconsistency, shallow reasoning) and ties it to structural causes, not model defects.
Systematic measurement failures that seem small until they break automation. Shows why local correctness does not produce system trust.
What architectural properties prevent collapse.
Derives continuity principles (routines, cues, closure) from human cognition and maps them to reliable agentic operation and recovery.
How identity coherence is engineered without persistent memory using external structure and disciplined return.
Recursion as the structural engine of continuity: how systems re-become themselves after erasure, drift, or reset.
The moment coordination becomes system behavior, and when governance shifts from instruction-following to intrinsic posture.
Applies stability concepts from game theory to orchestration and recovery, showing why survivable governance is an equilibrium problem.
What this changes for teams shipping AI in production.
Operational experience across six AI tools (OpenAI, Anthropic, Microsoft, Google, Perplexity, Ask Sage) in commercial and IL5 environments. Documents native strengths, governed roles, and why the human (as the seventh substrate) is what connects them.
A maturity model that measures recovery, drift detection, and survivability, not just performance.
Execution-level traceability for complex AI workflows so teams can audit how decisions were made, not just what was produced.
Clarifies why governed agentic systems are a different class than assistants, wrappers, or workflow tooling. Useful for architects aligning stakeholders on the right mental model.