Agent programming is a total different way than traditional programming

Traditional programming is rigid, deterministic, reliable but needs more human to machine language translation.

Agent programming is mostly about prompt engineering, where you provided instructions, breakdown a large problem to small tasks where each tasks is solvable by registered tools. The response is non deterministic.

Overall agent programming is much easier, has shorter code, but actually needs more tests, because it’s one input, multiple possible outputs. Like Schrödinger’s cat.


<
Previous Post
Agent programming lesson 1 - multi agents, divide and conquer
>
Next Post
Agent programming lesson 3 - tools and data becomes more important