Almost Entirely Human

Tell it what task it's doing and what scope it's working in. The "you are an expert" prompt is mostly cargo cult now; a list of NEVER statements is brittle.

I double down on the idea that keeping up with AI-generated code looks a lot like gardening: seeds, soil, water, pruning, weeding, and sometimes pulling up a plant.

The number one reason folks struggle with AI tools is they don't give the model enough to work with. Context is the cheapest upgrade you can make to any AI workflow today.

Esri now has a whole family of AI assistants built into ArcGIS—Arcade, Pro, Survey123, Item Details, Story Maps, Hub, Documentation, and more. I suggest starting with a workflow you do every week so you can start to learn when to trust the AI tools.

I built an 8-bit video game-style AI red teaming game that lets you try to break various chatbots with different guardrails so you can see how it works.

I built a chatbot for my website using the five-layer guardrail stack from Episode 58. My AI agent wrote the code, then a separate AI agent attacked it. You can try to break it too.

I built an autonomous AI agent that runs 24/7 on a Mac Mini, manages my home office, triages my email, helps write content, and improves itself every night at 3 AM. Here's an introduction to how it actually works.

Poison pills are hidden instructions in data your AI consumes. They’re a fun defensive trick for content creators and a serious attack vector if you’re building AI tools that touch external data.

It takes more than a good system prompt to protect an AI from abuse, and it is worth making it as hard as possible for attackers.

Do you garden? I do, sometimes. I have a raspberry bed that I am proud of. Sometimes I grow a few tomatoes, lettuce, and the occasional pepper.