You can, and you probably already do. Here is what changes when an experienced engineer is in the loop.
You start from a blank prompt, and the model starts from nothing it knew last week.
It answers confidently about a system it has only partly read, and you have no easy way to tell which parts it actually checked.
It will happily agree with a bad plan, because agreeing is what it is good at.
Nobody senior ever looks at what came out, so a mistake survives until it becomes expensive.
I already know your stack, your infrastructure and the decisions you made before, so nothing starts from zero.
I write the prompt knowing what a good answer looks like and what the model tends to skip.
I read the output the way a reviewer reads a colleague’s work: what is evidence, what is assertion, what is missing.
Prompts are one of my tools, not the thing you are buying. What you are buying is the judgment about which question to ask.
Mikhail Piskunov. System architecture, high-load systems and systems analysis, applied now to products built with AI tooling.
Designing systems that have to keep working when the load, the team and the requirements all change underneath them.
Fintech, SaaS, e-commerce, logistics and transport platforms, where being slow and being broken are the same thing.
Systems analysis, marketing technology and automation, and the integration work that is always harder than the estimate.
Architecture and technical guidance for engineering teams that already exist and already have opinions.
Pick the level of involvement that fits, describe your product, and we begin there.
Personal technical expertise for products built with AI coding agents. Three levels of involvement: advising the founder, leading the development, or building it hands-on.
