Dive Club Content OS
Turns podcast episodes into 80%-ready written content, in the host's voice. No UI. Claude Code is the interface.
Every episode = 4 hours of manual writing
Show notes, newsletter, social posts, hooks. All in Ridd’s voice. All by hand.
AI that sounds like a specific person
Not "good enough." Exact. Three major versions to crack voice matching.
60 seconds of a human talking to himself
The model stayed the same. A 60-second brain dump changed the voice.
Record → Run → Done
Episode + brain dump in, newsletter + social + notes out. Zero UI. Recursive learning.
Ridd hosts Dive Club, one of my favorite design podcasts and a big influence on how I think about my career. In 2025, I got to work with him on an agentic pipeline to produce the written content that goes out with each episode.
After every episode, Ridd needed show notes, newsletter copy, social posts, and hooks. All manual, all slow, all eating into time he could spend on his next episode or his startup.
A cleaner transcript still didn't sound like Ridd.
The bigger challenge was making AI-generated writing sound like a specific person's voice. Early drafts read like every other AI newsletter: technically correct, zero personality.
Each version got closer. V1 nailed the ideas but sounded generic. V2 had better rhythm but still read like an AI wrote it. V3 finally sounded like Ridd. The input changed.
“Why designers are severely undervalued when they’re actually the key differentiator in the AI era”
Close on idea, not on voice“Why design becomes the differentiator when everyone can build software”
Tighter rhythm, still generic“what happens when everyone is a 7/10 designer”
Ridd’s voice, Ridd’s angle60 seconds of a human talking to himself became the single most important input to the entire system
No UI. No dashboard Ridd has to learn. He records an episode, runs it through the Claude Code pipeline with his brain dump, and gets back a newsletter draft, social posts, show notes, and hooks. All in his voice.
What do you do when everyone becomes a 7/10 designer overnight? Carl Rivera thinks most companies are asking the wrong question. As Shopify's new CDO, he's betting that design, not AI models, will be the critical differentiator when software becomes infinitely generatable.
The 7/10 Designer Transformation: AI makes everyone a 7/10 at every job, including design.
In an AI era, what does it look like to reimagine e-commerce from first principles? That's the question Shopify's Chief Design Officer, Carl Rivera, has set out to answer, so this week's episode is all about Shopify's big bet on design and craft as the ultimate differentiator.
Why design is undervalued: for everyone else, design is the most valuable talent. When software is essentially free, how do you make something people will actually pay for?
An end-to-end pipeline with zero UI
Every episode flows through the same sequence: no manual steps, no copy-paste, no prompt engineering on Ridd's end. Here's what came out of it for Episode 17.
See the “+ a lot more” at the end of Ridd's bullet list? That's a signature Ridd phrase, italics and all. The AI picked it up from the training data and the published version kept it.
hot paths
what happens when everyone is a 7/10 designer
In an AI era, what does it look like to reimagine e-commerce from first principles? That's the question Shopify's CDO, Carl Rivera, has set out to answer so this week's episode is all about Shopify's big bet on design and craft as the ultimate differentiator.
We go deep into:
Episode and brain dump in. Newsletter, social posts, show notes, and hooks out.
Inputs
Pipeline
Deliverables
I spent weeks chasing a perfect transcript through local Whisper and cleanup scripts. The useful change came when I stopped polishing the source and asked Ridd to react while the conversation was still fresh.