Browse Comments — Raw (as collected)

Clear inputs and tight review loops make AI coding actually useful.

Leonard Rodman, M.Sc. PMP LSSBB CSM CSPO Workato The strongest outcomes usually come from continuous refinement rather than one perfect intervention, because progress compounds through feedback loops, small adjustments, and disciplined attention to detail over time.

The vending machine framing is the right diagnosis. The failure mode isn't the first prompt being wrong, it's not catching the moment the model went sideways three steps later. Reviewing as you go is the expensive skill, not the prompt. Building snapfast.ai exactly this way. Fast builder next to me, but I'm still the architect, the reviewer, and the one on the hook when production breaks.

The think–build–check loop is a simple but powerful way to describe what good development already looks like, just faster with better tools.

The frustration many people feel with AI coding often comes from skipping the same steps that make traditional software reliable in the first place.

What stands out is the emphasis on ownership, AI can accelerate execution, but the responsibility for architecture and quality still sits with the human.

Is that guy barefoot? Yes, that’s all I got out of that.

The “vending machine” mindset is exactly where expectations break because iteration is still doing most of the heavy lifting

Leonard Rodman, M.Sc. PMP LSSBB CSM CSPO Workato Really well expressed. I like how you’ve shown that coding with Claude Code is about process and craftsmanship, not shortcuts or one‑line prompts.

Appreciate this perspective on coding with Claude Code. Learning to guide the work step by step is key for success in AI development.

The biggest gains come from those who are willing to slow down, ask tough questions and treat each iteration as a chance to actually improve the result Leonard.

Building with Claude Code is like leading a skilled crew through every phase not just tossing blueprints over the wall Leonard!

the best AI workflows I’ve seen aren’t about asking for a full finished thing in one go. they’re about giving clear context, checking the output, spotting gaps, then tightening the next step Leonard Rodman, M.Sc. PMP LSSBB CSM CSPO Workato

Craft, don't just command.

love this take, adoption curves don’t matter if your team isn’t confident enough to actually move with them

This is the right framing. AI coding works best when you treat it like a build process, not a vending machine.

AI coding still needs taste... judgment... and real problem solving Leonard

The vending machine analogy is useful because it highlights a common misunderstanding. Many frustrations with AI come from expecting completion instead of collaboration through stages.

The value is in the iteration, the testing, the catching edge cases.

AI as fast builder, human as architect. The output speed is the easy part. The structuring and review is where the skill lives.