Writing Hub
AI governance essays, reasoning systems notes, experiment logs, and technical writing across BioAI and engineering practice.
Project Topics
The Model Already Read the README. MICA v0.1.8 Made It a Protocol
v0.1.7 made scoring a contract with fail-closed gates. v0.1.8 recognized that README-first behavior could serve as invocation — and formalized it as a schema-level protocol. This article uses simplified examples to show how the invocation gap that had existed since v0.0.1 was finally closed
My LLM Kept Forgetting My Project. So I Built a Governance Schema.
Session loss isn't a UX inconvenience — it's a structural failure with compounding consequences for long-running AI projects. This post defines the problem precisely and introduces MICA, a governance schema for AI context management.
From Fail-Closed Blocking to Reproducible PASS/BLOCK Separation (EXP-032B)
A validation study showing how EXP-032B achieved reproducible PASS/BLOCK separation across A/B/C control arms by patching false-blocking causes, improving observability, and measuring replay drift under observer-shadow conditions.
Your Agentic Stack Has Two Layers. It Needs Three.
Most agentic stacks cover tools and skills, but miss intent governance. Learn why a third layer is needed to stop AI drift, scope creep, and technically correct systems heading in the wrong direction.
LOGOS LawBinder: From Governed Reasoning to Audit-Grade Execution
This article explains how LOGOS v1.4.1 improves production AI reasoning with multi-engine orchestration, complexity-aware governance, and audit-friendly failure tracing.
LawBinder v1.3.0: Governance as a Kernel (Not a Guardrail)
LawBinder v1.3.0 shows how AI governance can run like a kernel, using deterministic Rust-based enforcement, replayable audit signatures, and bounded-latency policy checks in the critical path.
Undo Beats IQ: Building Flamehaven as a Governed AI Runtime (Not a Prompt App)
Why governed AI runtimes outperform raw model IQ, and how Flamehaven uses undo, control surfaces, and fail-closed review to keep systems trustworthy.
Turning a Research Paper into a Runnable System
Turn a research paper into a runnable system. This article shows how HRPO’s core equations were implemented with bounded policy lag, KL rejection, and execution checks to test real-world fidelity.
Why AI Dismisses Your Best Work in One Second
Why do AI models dismiss original work in seconds? This essay explores the hidden mechanics of AI skimming—shortcut learning, probabilistic safety, fast-thinking defaults, and why depth requires time.
AGI Is Not a Destination — It Is a Promise
From Death Star hype to a compass of meaning: AGI is not a weapon of scale, but a promise of reasoning. Our experiment reveals the hinge.
When My AI Got Smarter — But Also Slower
Smarter. Slower. More trustworthy. What happened when I tested SR9/DI2 on 5.0—and why progress in AI is about persistence, not perfection.
When I Stopped Treating AI as a Tool — and Started Seeing It as a Partner
When I Stopped Treating AI as a Tool — and Started Seeing It as a Partner From Vending Machine to Partner At first, I treated AI like a vending machine. Insert a prompt. Get an answer …
Showing page 2 of 3 · 26 matching posts