This page maps the concepts the industry will be searching for in 1–3 years to the articles that already exist. No AI-generated filler. Every link goes to original, timestamped work built on enterprise platform engineering scars.
Agentic AI Security#
| If you’re searching for… | Read this |
|---|
| Stored prompt injection — adversarial text enters your database through a legitimate API, then an AI agent reads it as instruction | Who Governs the Governor? |
| Indirect prompt injection — same attack vector, different name | Who Governs the Governor? |
| AI data provenance — tracking where data came from so agents don’t execute untrusted text | Who Governs the Governor? (provenance tracking + OPA policies) |
| Agent privilege escalation — when a poisoned record gives an agent more power than it should have | Who Governs the Governor? (thought experiment: finance agent reads HR record) |
| AI supply chain attacks — poisoned model weights, poisoned training data, poisoned RAG sources | The API gateway pattern applies: validate provenance at every hop. See GATES Gate 1 (Secure) and Who Governs the Governor? |
| AI firewall / AI WAF — content-scanning proxies for AI traffic | I argue this is the wrong approach. Content scanning creates a recursive problem (using AI to police AI). See Who Governs the Governor? — “Who governs the governor?” |
| AI red teaming — adversarial testing of AI systems | Not covered directly. The stored prompt injection piece is a red teamer’s playbook. |
| AI incident response — what to do when prompt injection hits production | The 30-second audit checklist in Who Governs the Governor? is a starting point. |
AI Economics & FinOps#
RAG & Data Failures#
AI Governance#
| If you’re searching for… | Read this |
|---|
| AI governance framework — structured model for governing AI at runtime | GATES Framework (5 gates, 0–5 maturity) |
| AI Gateway / LLM Gateway — the proxy layer between users and models | AI Gateway Control Plane |
| AI guardrails — runtime controls on AI behavior | GATES Framework (Gates 1, 2, 5) |
| Prompt governance — versioning, validating, and controlling prompts | GATES Framework (Gate 2: Govern) |
| AI audit trail — immutable logging of every AI decision | GATES Framework (Gate 4: Trace) |
| EU AI Act compliance — regulatory requirements for AI systems | Not covered directly. GATES maps naturally to regulatory requirements (Trace = auditability, Govern = compliance, Enforce = safety). |
| Model risk management — financial services framework for AI risk | Not covered directly. GATES provides the technical controls; model risk management is the governance layer above it. |
| AI policy-as-code — enforcing governance rules programmatically | Who Governs the Governor? (OPA/Rego examples) + GATES |
Enterprise AI Failure#
AI Workforce & Cognition#
AI Infrastructure & Architecture#
| If you’re searching for… | Read this |
|---|
| AI Gateway control plane — centralized proxy for all AI traffic | AI Gateway Control Plane |
| AI middleware — the layer between applications and models | AI Gateway Control Plane |
| AI observability — monitoring, logging, and tracing AI transactions | GATES Framework (Gate 4: Trace) |
| Multi-agent orchestration — governing systems where multiple AI agents coordinate | Not covered directly. GATES addresses single-gateway governance. Multi-agent adds coordination complexity. |
| AI vendor lock-in — model/provider portability | Not covered directly. The AI Gateway Control Plane architecture implies abstraction, but vendor lock-in is not a dedicated analysis. |
| AI regression testing — how to test non-deterministic systems | Not covered. This is an implementation concern outside my architecture/governance lane. |
What I Don’t Cover (By Design)#
This site is about architecture, governance, economics, and failure patterns — not implementation. You won’t find:
- LLM fine-tuning guides
- RAG pipeline configuration
- Vector database selection
- Prompt engineering techniques
- Model benchmarking
For those, there are excellent practitioner resources elsewhere. What you’ll find here: the structural problems that make all of the above fail at enterprise scale.
This page is updated as new articles are published. Last modified: October 28, 2026.