SLA Engineering for LLM Agents: Specifying Failure, Recovery, and Cost
A concrete, testable approach to LLM agent SLAs: specify failure semantics, observability, replanning budgets, and economic duties across tools and vendors.
Read essayEssays
Writing
Long-form essays on AI orchestration, technical program management, AI in project management, and the operational reality of agentic AI in regulated environments.
Topics:
-
-
Design Patterns for Cost-Aware Agent Orchestration
Agentic AI can quietly melt budgets. Practical patterns for metering, staged fidelity, simulations, and circuit-breakers to make agents accountable.
Read essay -
Signed Skills: A Practical Standard for Safe, Open Agent Marketplaces
A case for a neutral standard for cryptographically signed agent skills and runtime attestation that keeps marketplaces open while enforcing safety and provenan
Read essay -
Auditability-first Agent Memory: Design Patterns for Trustworthy Recall
How to build agent memory you can trust: cryptographic provenance, scoped capability tokens, and SLOs for enterprise-ready, auditable recall.
Read essay -
Operationalizing Defenses Against Persistent-Control in Agentic AI
A layered model—provenance, attestation, confinement, automated red teaming—against persistent-control attacks in agentic AI, with deployable steps.
Read essay -
Push Agent Skill Distillation to Devices, Not Just the Cloud
Argues for hybrid agent architectures that distill signed skills to devices, cutting cloud costs, reducing data exposure, and escaping platform gatekeeping.
Read essay -
Survive Platform Gatekeeping: Architecting Enterprise AI Beyond App Stores
How to ship agentic AI under tightening app store rules: architect attested modules, supply-chain proofs, and hybrid delivery to keep control and security.
Read essay -
Cryptographic Action Tokens: A Governance Model for Agent Orchestration
Require cryptographic action tokens and reproducible execution transcripts for every agent effect to enforce least-privilege, auditability, and rollback.
Read essay -
Designing Agent-Aware Public APIs: Versioning, Limits, and Contracts
Public APIs need agent-aware design: versioning for agents, capability-bound tokens, semantic rate limits, and verifiable intent contracts to curb abuse and cos
Read essay -
Platform Gatekeeping Is Killing Agentic Open Source — Fixes
Platform gatekeeping is throttling agentic AI open source. A practical blueprint—manifests, provenance, sandboxes, legal and CI—to keep it portable.
Read essay -
From Explorer to Trailblazer: A Project Leader's GenAI Roadmap
How project managers can move beyond personal productivity hacks into systematic, transformative use of generative AI - without losing the discipline that makes project management work.
Read essay -
Multi-Agent AI in Government Isn't a Chatbot Problem
Why the next generation of government digital services needs orchestration, not assistants - and what it takes to deliver multi-agent AI in a regulated, multi-entity environment.
Read essay