building the claude code for marketing 2x exits (beambox, reword)
England
Joined September 2017
- Tweets219
- Following462
- Followers13.2K
- Likes2.4K
MCPs and Skills are a terrible UX
One thing I realised talking to customers for Juno is that the average marketer has zero interest in using them
A good agent just knows what to do
Probably the most compelling argument for vertical > horizontal agents
ai marketing agencies are going to be a goldrush
there is SO much leverage, but it's still technically inaccessible for most people
you can have 10x the output of a traditional agency for half the price
i started using google ads 10 yrs ago
back then, manual cpc got you results
now, doing nothing and letting the algo figure it out gets you results
optimisation keeps getting pushed down the funnel
in a few years time, the only thing that will matter is your offer
Gemini models are supreme for landing pages
But, I think 3.1 was a regression
Take this example, both are one-shot prompts using Blox editor
What do we think?
A little trick to get more people clicking your CTAs
Make them action-based
“Submit”
“Continue”
“Next”
"Sign up"
Boring buttons kill momentum. Instead..
“Generate my first version”
“Show me my page”
“Get my score”
“Send me the template”
I am never going back to Webflow or Framer
NEVER
For the Blox website, I have a set of claude skills that:
- creates new page structures with 95% accuracy
- proactively optimizes for seo/accessibility
- audits itself after every deploy with Google PageSpeed
- knows our ICP
@coreyhainesco inspired by your repo!
having scripts the skills can execute, e.g for the PageSpeed audit, is wild
hot take: i don't think 'payments for agents' is a problem that needs solving
we already have a great system for this, they're called credit cards
running a split test to see if the prompt box converts better than just a 'get started' button
which do we think will win?
