Taming agentic engineering – Prompts are code, .json/.md files are state
mariozechner.atThis is maybe the best instructional article on how to use llms I’ve seen yet.
It illustrates their strengths and weaknesses and how to take advantage and overcome both. All prompt code is provided and explained. The amount of effort needed for good results is realistically portrayed as well as the benefits. It’s specific to claude code but it’s a moderately simple example so jumping to your work isnt complex. 10/10 must read.
Thanks for your kind words! I hoped it would trigger some more discussion on here and that people share their approaches. Alas, "self-posts" are likely not very well liked around these parts :)