Notes on agents, software, and work.
Systems / August 16, 2026 / 10 min read
AI delegation: how much authority should an agent get?
AI delegation works when each action has a clear authority boundary, proof of completion, stop condition, and human approval point.
Systems / August 16, 2026 / 10 min read
AI agents vs agentic AI: where the acting unit ends and the system begins
AI agents perform tasks for users; agentic AI shifts the unit to long-horizon work, with orchestration and controls shaping the system.
Systems / August 16, 2026 / 10 min read
Agentic AI vs LLMs: where the model ends and the system begins
LLMs generate language; agentic systems add memory, tools, execution, controls, and recovery around the model to complete multi-step work.
Systems / August 16, 2026 / 11 min read
Step by step AI agent architecture: build from recovery backward
Build an AI agent architecture from recovery backward, with explicit tools, durable execution, approval gates, safe retries, and crash tests.
Systems / August 16, 2026 / 10 min read
Multi-agent orchestration architecture design starts with a coordination budget
A coordination budget reveals whether multi-agent orchestration earns its handoffs, shared state, authority boundaries, observability, and recovery costs.
Systems / August 16, 2026 / 10 min read
Agentic AI vs multi-agent systems: choose the architecture, not the label
Agentic behavior does not require multiple agents; use context, parallelism, specialization, boundaries, and coordination cost to choose.
Systems / August 16, 2026 / 10 min read
MCP vs agentic AI: compare the layer, authority, and control points
Agentic AI decides and acts; MCP connects models to tools. Compare their layers, authority risks, control points, and where A2A and APIs fit.
Systems / August 16, 2026 / 9 min read
How to build agentic workflows that can stop, resume, and recover
A recoverable agentic workflow starts with explicit state, side effects, verification, approvals, stop conditions, checkpoints, and recovery.
Systems / August 16, 2026 / 10 min read
AI agent architecture patterns: choose by workload and failure mode
Choose AI agent architecture patterns by workload, failure impact, coordination needs, human stop points, permissions, and recovery controls.