RK.
← Home

Field Notes

Working opinions from the current build.

Some of these will age badly. That's the point of writing them down with a date. Published here first, then on LinkedIn.

Essays

Each one started as a problem inside a build and turned out to travel. That's the test of whether it was an idea or a workaround.

Short notes

Less finished opinions. A paragraph each.

AI makes production cheap and review expensive

Everyone budgets for the speed-up. Nobody budgets for the domain expert who now has to read three times as much generated documentation. This is the cost that will quietly eat your velocity gains.

The PRD is the wrong artefact for AI builds

Requirements, then plan, then tasks beat a traditional PRD by a wide margin. The format matters more than the detail. A shorter, better-structured doc hallucinates less than a thorough, badly structured one.

Never open with a blank chat box

Structure first. Show the actions before you invite the question. Open-endedness is a capability, not an interface.

Enterprise-ready doesn't mean complex

Competing with twenty-year-old GRC platforms on flexibility is a losing game. Win by being the product that looks and works simpler while solving the harder case.

A settings request is usually an undecided design

When customers ask for a toggle, they're often telling you that you haven't made a call. Opinionated flows beat configurable ones in almost every small-business context.

Most UX problems are IA problems in disguise

Seven tabs, inconsistent click targets, stacking drawers. None of those are bugs. They're the accumulated cost of never deciding what the area is for.