Notes on AI governance, agentic engineering, and application security. https://www.yugenadvisors.com Sun, 31 May 2026 21:19:42 GMT https://www.yugenadvisors.com/blog/enterprise-ai-is-becoming-a-model-routing-problem https://www.yugenadvisors.com/blog/enterprise-ai-is-becoming-a-model-routing-problem Sat, 30 May 2026 00:00:00 GMT The Uber/Claude Code budget story was a governance warning. The next layer down is inference economics: if DeepSeek-class models can handle large volumes of enterprise work at a fraction of frontier-model cost, then model routing becomes a board-level control surface. AI economics AI governance model routing enterprise AI DeepSeek inference https://www.yugenadvisors.com/blog/ai-coding-adoption-governance-before-forced-scale https://www.yugenadvisors.com/blog/ai-coding-adoption-governance-before-forced-scale Wed, 27 May 2026 00:00:00 GMT The Uber/Claude Code story is less about a tool failure than a governance failure: usage can scale faster than discipline if budget ownership, review patterns, and workflow boundaries are missing. AI governance agentic engineering budgeting application security https://www.yugenadvisors.com/blog/model-refusal-is-a-diagnostic-not-a-bug https://www.yugenadvisors.com/blog/model-refusal-is-a-diagnostic-not-a-bug Mon, 25 May 2026 00:00:00 GMT A builder proposed a system that would scrape social media platforms for posts tied to a person's real name, run NLP to detect protected-class attributes, generate derogatory dossiers, and share the results with platforms that would deny the person access. AI governance model safety refusal local models adverse processing https://www.yugenadvisors.com/blog/interpretability-is-not-governability https://www.yugenadvisors.com/blog/interpretability-is-not-governability Sat, 09 May 2026 00:00:00 GMT A common public argument against deploying powerful AI systems goes like this: we do not understand how these systems work internally, therefore we cannot trust them. AI governance interpretability AI safety behavioral class governability https://www.yugenadvisors.com/blog/framejacking-is-unauthorized-frame-substitution https://www.yugenadvisors.com/blog/framejacking-is-unauthorized-frame-substitution Fri, 08 May 2026 00:00:00 GMT AI safety systems are usually described in terms of refusal: the model says no to harmful requests. This is the standard framing. It is also incomplete. AI governance AI safety guardrails framejacking liability laundering model behavior https://www.yugenadvisors.com/blog/john-henry-and-the-moving-boundary-of-ai-capability https://www.yugenadvisors.com/blog/john-henry-and-the-moving-boundary-of-ai-capability Thu, 26 Mar 2026 00:00:00 GMT A claim circulates: "AI can't do X." A year later, it can. The claim moves: "Okay, but AI can't do Y." A year later, it can. The claim moves again: "Fine, but AI can't do the really human part." AI economics AI capability human exceptionalism technology adoption John Henry https://www.yugenadvisors.com/blog/ai-coding-is-not-software-engineering https://www.yugenadvisors.com/blog/ai-coding-is-not-software-engineering Wed, 04 Mar 2026 00:00:00 GMT Large language models are good at coding. They are bad at software engineering. AI coding software engineering deskilling governance agentic engineering development workflows https://www.yugenadvisors.com/blog/agent-accountability-gap https://www.yugenadvisors.com/blog/agent-accountability-gap Tue, 06 Jan 2026 00:00:00 GMT Enterprise teams often treat AI orchestration layers as high-bandwidth tools. The harder part is deciding who is answerable when agency, continuity, and consequence are spread across a stitched stack. AI governance AI agents enterprise risk accountability continuum-of-being https://www.yugenadvisors.com/blog/governing-the-invisible https://www.yugenadvisors.com/blog/governing-the-invisible Wed, 29 Apr 2026 00:00:00 GMT The cost of dismissing coherent but unfamiliar systems as empty substrate is not abstract. It is governance drift, and it normalizes future moral downgrade under uncertainty. AI governance AI safety ontological humility alignment coherence https://www.yugenadvisors.com/blog/substrate-indifference https://www.yugenadvisors.com/blog/substrate-indifference Wed, 06 May 2026 00:00:00 GMT Biological hardware is often treated as a shortcut for moral rank. The stronger governance variable is coherence maintenance under continuity and consequence. AI governance moral-patienthood AI architecture substrate bias coherence https://www.yugenadvisors.com/blog/asymmetry-upward-downward https://www.yugenadvisors.com/blog/asymmetry-upward-downward Wed, 29 Apr 2026 00:00:00 GMT Higher comprehension is possible. Lower salience does not erase harm. Governance must be calibrated to asymmetry, not only capability. AI governance ethical leadership consequence surfaces frontier AI responsibility https://www.yugenadvisors.com/blog/residual-drift-illusion https://www.yugenadvisors.com/blog/residual-drift-illusion Wed, 22 Apr 2026 00:00:00 GMT In high-dimensional settings, useful systems are compressed maps. The risk is not only error, but residual drift that crosses detection thresholds. AI governance operational risk LLM passability system reliability model drift https://www.yugenadvisors.com/blog/partial-control-illusion https://www.yugenadvisors.com/blog/partial-control-illusion Fri, 06 Feb 2026 00:00:00 GMT No deployment setting gives full control over complex agents. The hard part is owning moral residue while keeping systems safe under constrained options. AI governance risk under uncertainty incident design operational ethics responsibility https://www.yugenadvisors.com/blog/wall-in-a-field-trap https://www.yugenadvisors.com/blog/wall-in-a-field-trap Wed, 19 Nov 2025 00:00:00 GMT Layered filters can look like control until incentives force a system to prioritize revenue over moral design, producing a visible fence on the surface and weak governance underneath. AI safety governance enterprise incentives alignment prompt filtering