Anti-patterns while working with LLMs

The post lists five “don’ts” for working with AI chatbots known as large language models (LLMs): don’t repeat info, don’t force it to do what it’s bad at, don’t stuff it with too much context, expect weaker results on niche topics, and don’t vibe‑code blindly. But the comments went nuclear. User sharkjacobs called it messy and suspected it’s just promo for click3. Others piled on about the “fish climbing a tree” analogy, turning it into a meme while debating whether the advice is even actionable.

On the constructive side, veterans like Scotrix boiled it down to street rules: be specific, keep requests small, and push strict tasks to real code. pedropaulovc shared war stories with Claude Code and the obscure SolidWorks toolkit: hallucinated methods and parameters made it “like pulling teeth.” Then willvarfar dropped a surprise: AI is better at understanding big projects and finding bug roots than writing new code. Meanwhile, isodev went full skeptic, saying LLMs simply aren’t capable of logic or true creativity. The biggest fight? Whether we should treat AI as a helpful analyst that writes code to call tools — or accept that the magic isn’t coming. The takeaway is pragmatic: less hype, more guardrails.