Agents and Automation

Past the hype: what an agent actually is, how the loop works, why most agent demos collapse in production, and how to build automation that survives contact with real data. Covers no-code tools (n8n, Zapier, Make) and code equally, and is honest about when a plain script is the better answer.

1. 1Three machines, one word6 minAn agent is a program where the model, not the author, decides what happens next.
2. 2A tool call is a request, not an action7 minA tool call is a request your code fulfils, and the result string is the model's only feedback.
3. 3Plan, act, observe, repeat7 minThe loop is trivial; deciding what goes back in as the observation is the entire craft.
4. 4Why the demo worked and production did not8 minPer-step accuracy compounds, so shortening the chain usually beats upgrading the model.
5. 5Budgets belong in the harness, not the prompt7 minAnything you need guaranteed goes in the harness; a prompt can only ask.
6. 6Where the person goes8 minGate on what cannot be undone, never on how confident the model sounds.
7. 7MCP, and why a standard for tools matters8 minA tool standard removes integration work; it does nothing about whether the model chooses tools well.
8. 8Build the boring version first9 minMost automation that survives is a deterministic pipeline with one model call and a real validator.