I have my own special perspective on how systems should be built—how they should function, scale, and interact with their users. It’s quite difficult to express this in words or formalize it, but I usually feel that something “fits” me precisely when I see it in practice. Sometimes companies and teams turn to me for advice because they trust this intuition.
Artificial intelligence doesn’t change anything here—it’s more of a powerful tool to enhance capabilities. I can delegate some tasks to machines, but the final result still needs to meet certain standards. Often, achieving this immediately isn’t possible. You need to try different approaches: clarify constraints, rephrase tasks, experiment and adjust until you find the solution that feels right.
Today, tools are truly powerful and multifunctional, but the main filter remains the sense of correctness—the very taste.
Your, @llm_under_hood
P.S. I’m currently struggling with ChatGPT Pro and Codex trying to design a simple API/MCP for efficient sharing of my personal context database between agents and tools in our system. I’ve been working on this design for three days already, trying to find one that meets all my requirements and remains simple—so far without success.
Created with n8n:
https://cutt.ly/n8n
Created with syllaby:
https://cutt.ly/syllaby