Claude Opus in Daily Work: A Practitioner's View

For a year now, Claude has been the first thing I open when I need to think out loud. Not because it's smarter than me at my own job. Because it doesn't get tired, doesn't carry the “we've always done it this way” bias, and isn't afraid to say an idea is weak.

I describe the task, ask for 10 approaches. Out of 10, I take 2–3 forward. Faster than running through options in my head — a written “here are all the ideas” unloads working memory and lets me choose with a cold head.

I describe the architecture or the logic before writing any code. I ask it to find weak spots. I get a list of questions I didn't ask myself. Saves hours of debugging later.

I explain what was built, ask for a README or a technical section. I tweak the tone — done. Not because I'm too lazy to write, but because turning loose thoughts into coherent prose is a separate skill, and Claude handles it well.

One important point: the quality of the answer is linear in the quality of the prompt. Bad prompt, bad result. Learning to describe a task clearly pays off more than hunting for the “best” model.