skilled in agentic coding
skilled in agentic coding
Posted Jun 8, 2026 13:04 UTC (Mon) by aimannajjar (subscriber, #184277)In reply to: skilled in agentic coding by kleptog
Parent article: BPF in the agentic era
That reminds of something. Did you know that there is a 70k-star "Claude Skill" project that allows you to communicate with Claude like "cavemen" with the aim of minimizing token usage: https://github.com/juliusbrussee/caveman ? Yes, it seems that trending skill today is how to minimize tokens by downgrading your English to caveman-level.
All that to say, there is no proof that prompt style has real effect on the quality of the output. This is why "prompt engineering" has quickly become an obsolete term, along with RAG and many other hyped terminology (I'm sure "agentic coding skills" will follow suit). The fact is, no matter how precise you are, you will have to repeatedly engage the LLM to get to the desired output, and in the process you will accumulate so much technical debt, as well as cognitive debt.
> We created programming languages *because* natural languages are imprecise
That is the crux of the problem, LLMs are inherently imprecise, due to them being stochastic and non-deterministic.
We are really doing a disservice to young, aspiring and smart engineers by giving them the illusion that they're investing their time in a valuable skill such as writing good prompts, when they should be their hands dirty on real coding exercises.