From To-Do Lists to Machines

Log 003 – From To-Do Lists to Machines

Things are moving now. For real.

After the first two logs — why I’m doing this and actually shipping something — the project has entered a new phase. Less exploration. More direction. More machine.

AdSense in motion (finally)

Google has now confirmed that they’ve started reviewing VadKostarBadet.se for AdSense.
It’s still a waiting game, but just having the process in motion feels like a major mental shift.

Once ads are live, the next steps are straightforward:

  • Double-check that the calculations are actually correct
  • Validate the numbers with real traffic
  • Address technical debt that was intentionally postponed

This site has always been a proof of concept. It’s now approaching its first real test.

First gated freebie live

At the same time, we’ve shipped:

  • Email integration
  • The first downloadable PDF
  • A clear “free → email → data” flow

This matters. Not for revenue yet, but for learning.
We’re not building content. We’re building relief.

The free products exist to show what people actually download — not what we think they want.

Stricter formats – less freedom, more scale

A major change has happened in how digital packages are created.

We’ve introduced a stricter format and a clear process for every new product. Less ad-hoc. More structure. More machine.

Cursor is currently checking the entire process into the project so that:

  • every product starts from the same base
  • quality stays consistent
  • creation gets faster with every iteration

It’s not romantic.
But it’s exactly what’s required if this is going to become more than a one-off project.

Competitor analysis + customer interviews

In parallel, I’ve started running:

  • Competitor analysis via deep research
  • Customer interviews with people who’ve bought similar products online

The goal isn’t to copy. The goal is to understand why people buy, what works — and what doesn’t.

That work has led to a very clear short-term goal:

Have one strong Easter product and one strong birthday product ready before Easter.

Not ten products. Two that actually deliver value.

From to-do files to prompt-driven work (and doubt)

The way I work has also changed.

Before, a lot was driven by:

  • to-do files
  • manual checklists
  • sequential execution

Now, more is driven by:

  • prompts in Cursor
  • clearer instructions to the machine
  • fewer micro-decisions from me

Honestly, it has sometimes felt like a step backwards. I’ve questioned it more than once.
But that led to something important: we updated the ChatGPT project routines to give the machine a more active, responsible role.

Earlier, the AI wanted to teach me how to build. That made sense then.
Now I want to let go. Delegate. Let agents work.

The current tool setup

Right now, the setup is clear:

  • ChatGPT 5.2 → brain, project lead, direction
  • Cursor → developer (no fixed model)
  • Me → account setup, GUI clicking, quality control

I’m also continuing to study Claude Code — the tool that started this whole project, but one I’m not actively using right now.

What I miss most is long-running agents that can work over time. Less one-off prompts. More continuity.
But we’re getting there.

This is where the project stands right now.