Suggested conversations
AI Agents Are Leverage, Not Magic
A practical map of what agents do well, where they fail, and why judgment stays attached to the operator.
What Actually Breaks When You Put Agents to Work
Silent failures, stale memory, retry loops, wandering files, weak monitoring, and the guardrails learned afterward.
The Engineer’s Guide to Human-in-the-Loop AI
Requirements, permissions, test evidence, approval points, deterministic controls, and accountability.
Building the Thing You Couldn’t Build Before
How agents shorten the distance between an idea and a real market test—without making the idea good.
From Chatbot to Coworker
The operational stack around tool access, durable work, memory, monitoring, and supervision.
Guardrails Before Autonomy
Execution ceilings, spend caps, outcome watchdogs, secret handling, and kill switches before scale.
Short bio
Alastair Fraser is a Staff-level circuit-design engineer, engineering leader, and author of The Agent Advantage. He combines experience in space, defense, RF, microwave, avionics, and satellite communications with hands-on work building and supervising AI-agent systems for software, online stores, content, research, and operations.
Long bio
Alastair Fraser is an electrical engineer and Staff-level technical leader whose career spans satellite communications, RF and microwave systems, circuit and board design, avionics architecture, requirements, testing, troubleshooting, and engineering leadership. Outside his primary engineering work, he uses AI agents to build software products, operate online stores, research markets, create content, and test practical business systems. His book, The Agent Advantage, documents the useful work, the failures, and the guardrails required to keep human judgment in control.
Interview questions
- What turns a chatbot into an agent?
- Which failures surprised you only after the systems ran unattended?
- How can a non-programmer supervise software-building agents?
- What should always remain under human control?
- Why do you recommend manual work before automation?
- What can traditional engineering teach agentic systems?
- How do you decide when a demo deserves trust?