Initially I got up thinking: „FUCK“.

I am behind on delivering client work.

Then I very briefly helped at Laura’s house, but left after 30min since I was not necessarily needed.

I went to Spar, got a sandwhich + redbull and rode Astrid to Appian Way (Allee).

Said to myself: „Must run up the Iron SOPs – think“.

Jenissa delivers great designs.

But I have to go in an create texts.

Could we not streamline that?

First thought actually: let’s create a masterprompt that specifies all context for an LLM to generate relevant content.

Whipped out Claude on the iphone in the browser and let it craft that. Too lazy to even read the long masterprompt it spit out – told it to create a UI for this using the OpenAI API.

UI actually looks pretty decent actually. I made one with lovable then as well.

https://preview–webtext-blender.lovable.app/

and some local versions in standalone HTML files.

Refined it with Claude 4-5 times.

However – the output was suboptimal – too broad context.

It is actually better to prompt each section – do it manually so to speak.

I’ve now spent 5+ hours doiung that.

Glorifying and bloating something up into a UI or app – that can be handled by simply „doing it“.

The eternal programmers flaw: „Why spend 10min doing something when I can spend 10 hours automating it?“

Fuck off. Do the work needed.

Momentum builds more momentum.

Also came up with an „Idea“:

But I’m not sure how much merit it holds.

Right now, he has more to show for than me.

And just saying „Trust me, we’ll do that…“ is something I’d like to avoid going forward.

Now fuck off – and get to designing and building.

Actual RAW progress.