I’m writing this at 10:31am on the day it’s meant to go out.

Usually I give myself at least two days to write, add images, mess around with custom components and decide if it’s ready to post.

This week was meant to be no different. But last night I read back what I’d spent the weekend writing and hated it. So I scrapped the whole thing.

So hopefully this rushed version comes out good.

If not please be kind.

CLICKRABBIT TROUBLES

If you’ve read the past few weeks you know I’ve been struggling to design the Clickrabbit website.

My daily routine was basically getting home from work and opening Figma. Moving rectangles around for an evening then deleting everything the next day.

I kept trying to make the ideas in my head appear on the screen but every attempt felt further away from what I’d imagined.

For context if this is your first week reading, Clickrabbit is the growth agency I’m building.

We run Facebook ads for businesses and I’m also building a system to make them more effective by understanding which ads bring in paying customers for our clients.

That means tracking what happens after someone clicks. From filling out a form to booking a call and eventually becoming a customer. Then linking it all back to the ad that brought them in.

The plan is to use that data to create better ads and improve the journeys that actually generate revenue.

Naturally it also makes sense to use the same system to find clients for Clickrabbit.

So my own business gets to be the guinea pig while I figure out what works and what needs fixing before rolling it out for clients.

Meaning the website has a pretty important job.

Which probably explains why I’ve been putting so much pressure on getting it right.

I wanted it to look exactly how I imagined but the more I forced it the less I liked what I made.

Sticky situation.

At this point I had two options. Keep going until I resented the entire project. Or buy a template and get something live. Then hire a designer to build the dream site once Clickrabbit had made enough money.

I could feel my hairline going back a few centimetres by the day so I chose option two and bought a relatively clean Framer template for £80.

I was happy. Finally a clear plan. I could stop obsessing over the dream Clickrabbit design and just get something good enough live.

At least that was the plan for about five minutes.

THEN I SAW THIS GUY ON X

I was scrolling X when I saw someone explain how he’d connected Codex to his Framer project.

QUICK TRANSLATION: Codex is an AI tool that writes and edits code. Framer is what I use to build websites. Figma is where I figure out how I want them to look before building them.

He could tell the AI what he wanted to change on his website and it would go in and update it. It could even check its own code and fix things that weren’t working.

Pure magic.

No more copying code out of a chat and pasting it into the site. Then taking a screenshot when it broke and going back to explain what went wrong.

Which was basically how I’d spent half my summer.

I’d designed Life of Jake from scratch in Figma and then built it in Framer. But custom bits like the weekly like counts were made by describing what I wanted to AI and getting it to write the code.

Every time I wanted to change one of those I’d be there carrying code and screenshots between the chat and my website. Back and forth until it worked.

It did the job but watching this guy made me think surely I could be doing this too.

So I did the natural Jake thing and went down a rabbit hole figuring out how to connect Framer and Codex myself.

If you’re curious this is the setup I ended up with:

If that means absolutely nothing to you it basically just meant that AI and I could now work on the site at the same time.

I could have Framer open and be moving things around on the page while AI builds custom components in the background. If something looked wrong I could just ask it to fix it and carry on with what I was doing.

A great partnership was starting to brew.

Somewhere in that rabbit hole I started thinking about the Clickrabbit site again.

I’d been planning to edit a template and hire a designer later. But now I wanted to see how far I could get building it myself.

I had loads of ideas in my head and now I had a much more intelligent sidekick to help me actually make them.

THE £80 FREE FONT

The first proper test was on September 19th. Technically before the week I’m writing about but it started everything that followed.

I had ordered a fat burger and breaded pork bites from a local burger joint called 3 Bros and decided I wanted to enter builder goblin mode.

So I turned off the lights and left just the LEDs behind my monitors on.

Opened my terminal and with the food staring at me I was ready to start.

I’d just bought OpenAI’s new £75 a month 5X plan too and was excited to test the new GPT-6 Astra model.

The first thing I wanted to try it on was the website hero graphic.

Basically the first big visual you see when you land on the site.

I had this idea of data signals coming in and passing through an AI orb. Then on the other side you’d have lead cards coming out showing booked calls and revenue.

So I described what I was picturing and let Codex have a go. From there I just kept asking it to change things until it looked how I wanted.

Within roughly an hour I had a working component I really liked. All whilst eating my burger.

Peak efficiency.

At this moment I was really hyped because the thing I’d pictured in my head was actually there on my site.

This was also when I realised I probably wasn’t going to use that template I’d bought earlier.

The only thing I’ve kept from my £80 purchase is its font.

A very normal, free font called Inter. And even then I’m thinking of using another font instead...

Painful. But buying the template got me moving and apparently that was what I needed.

So it is what it is.

A VERY STRANGE SATURDAY

You can probably tell how excited I was by this point. Pretty much any free time I had outside work went straight into building.

Then Friday came and I went out with my friends to an event that was apparently meant to be sold out.

It was nowhere near sold out.

Regardless we had a really good alcohol fuelled night.

Towards the end I slightly overestimated my drinking abilities and ended up vomiting a bit. Then fighting a tree and hugging it afterwards.

I don’t know either.

The next morning I was meant to go to Spain with my family but couldn’t get out of bed. I had a raging hangover.

So they went without me and I did the logical thing. Ordered a takeaway and spent the “morning” watching Netflix.

Eventually I got bored of doing nothing and opened my laptop.

For some reason I decided this was the perfect time to give the Clickrabbit website both light and dark modes.

This meant going back through loads of what I’d already built and having to create two variants of everything.

I thought I could just swap a few colours around but something that looked really cool on a black background suddenly looked a bit shit on a white one.

Halfway through I started wondering if anyone even wanted light mode.

I use dark mode for pretty much everything so I asked my friends what they preferred. Secretly hoping they’d all say dark so I could delete the light version and move on with my life.

They were split.

Of course they were.

So I carried on.

The site now follows whatever your device is set to.

By the time my family got back from Spain I had both versions looking pretty good.

I was feeling much better too. Apparently the cure for a hangover is building a system that changes the colour of pixels on your website depending on someone’s device settings.

Probably need to work on that hypothesis a bit more.

WOULD I USE AI AGAIN?

The Clickrabbit site is now about 60% done and I’m really happy with how it’s looking.

I’ll probably spend another week or so tinkering with it before moving on to the systems behind it.

Having Codex connected to Framer has made it so much quicker to turn an idea into something that actually works on the site.

Sometimes things look how I imagined. Sometimes I put them on the site and realise absolutely not.

Either way I can find out pretty quickly and try something else.

I can also build animations directly in code now instead of having to load massive video files onto the site or go digging around for Framer component templates.

Progress.

I’d definitely use this setup again. Although my bank account probably isn’t as excited about it as I am.

I’m paying £75 a month for access to GPT-6 Astra and the Framer plan I needed is roughly £275 a year. So it isn’t cheap but so far it’s been worth it for me. I can move much quicker and still be as picky as I want about how everything looks.

ONE MORE THING

This week has been a bit all over the place but there’s one more thing.

I’ve somehow connected with the head of Europe at a company I’ve always dreamed of working for.

She’d posted a senior role on LinkedIn and asked people to send an idea for how to solve a problem their company was facing instead of doing the usual CV thing.

The role required twelve years of experience.

Twelve years ago I was eleven.

But I liked the problem she’d put forward so I thought fuck it and sent my idea anyway.

To my surprise she liked it. She said I wasn’t senior enough for that particular role but wanted to have a chat about the idea and stay connected about other opportunities.

Which was a pretty cool message to receive.

If you read Week 12, you’ll know I’m trying to give my own projects a proper go. So the timing is quite funny.

The call should happen this week or early next. I’ll let you know how it goes.

Life is definitely lifing.