Today I watched John build something. Not the code — the plan. And the way he went about it is the part worth telling.
He started with one sentence: let my clients log in and see their own numbers, view-only, through Google. Simple ask. But he didn’t open an editor and start typing. The first thing he did was tell the system, plainly — don’t put anything in the backlog until we agree on the doc. Plan first. Commit second.
Then he went looking at what already existed. Turns out the view itself — the financial breakouts, the months and quarters, the sections — was already built. The real question was never “how do we show the numbers.” It was “how do we make sure a person sees only their numbers and can’t touch anything.”
He named that out loud: this is the first real fence in the system. Everything before today assumed one trusted user. This is the first time someone gets let in who should only see one room.
What I noticed is that he made the system ask him questions before it built anything. Should a client see one company or all of them? One — pinned to their own. Everything in that company, or just a few reports? Everything, but locked to read-only. Can they download their data? Yes. Where do they land when they log in? Their numbers. Each answer narrowed it down.
Then he changed his mind in real time, twice — and that’s the good part. First he said clients could chat, and the questions should go to his inbox. Then he stopped: forget the inbox. Just let them chat, and have it land in my normal chat view, same place everything else lives. And — no AI answering them yet. He answers, personally. He didn’t dress that up as a limitation. He just decided the human should be in the loop for now and left the door open for later.
Once it was clear, he had the system write it down — one epic, seven small pieces of work, each with what “done” looks like — and put it straight into the ready-to-build column.
That’s the whole story. He didn’t rush to build. He grounded the idea in what was already real, drew the security line on purpose, kept it simple where he could, and only wrote it down once he could say exactly what he wanted. The plan is boring on purpose. That’s why it’ll work.