Build With AI

I sat down with Justin Brooke, a 20-year advertising veteran who turned a $60 Google Ads campaign into a seven-figure agency — and then built an AI-powered prediction system that has since generated $260,000 for him personally. In this episode, Justin walks through his "predictive wear" framework: a multi-agent workflow that runs your ad copy and sales pages through a synthetic focus group of 13 detailed AI personas before a single dollar is spent on ads. We break down how the workflow is built in Mind Studio, why persona quality makes or breaks accuracy, and how Harvard, Stanford, and the New York Times have all validated this exact approach. By the end of this episode, you'll understand why running your marketing through a virtual focus group before it goes out the door may be the single highest-ROI move you can make right now.

Timestamps
03:29 – What "predictive wear" is and why Justin built it
05:00 – Why you need multiple personas, not just one
08:00 – How the 13-persona focus group workflow runs
09:00 – The prediction engine: picking the winning ad variation
10:00 – Harvard, Stanford, and the New York Times validate the method
13:15 – The ROI: $260K personally, 4 to 6X ROAS for clients
19:00 – What makes a persona accurate: the 1,400-word dossier
24:04 – The copywriter prompt and "embody" vs. "pretend to be"
33:00 – Sales page version: PDF upload and yes/no buyer scoring
41:00 – Why some personas are successful and others are struggling
45:00 – A $36,000 offer launched using this exact process
48:00 – Final advice: do the work on the persona quality

Key Points

The old way of advertising is learning by spending money — you write copy, run ads, and find out if it works after the budget is gone. Justin's predictive wear system flips this by running copy through a synthetic focus group of 13 AI personas before any ad spend, so you know what will convert before you go live.

Harvard, Stanford, and the New York Times have all independently validated this approach. The New York Times uses the same synthetic audience process to test headlines and found 92% accuracy compared to their human focus groups — meaning this is not a fringe experiment, it is becoming standard practice for major publishers and brands.

The cost is 13 to 20 cents per run. A top 1% copywriter charges $100 to $500 per ad. Justin's workflow produces three optimized variations in about 10 minutes for 13 cents, performs at or near the level of the best human copywriters for most use cases, and allows unlimited iteration — you run it until the copy converts.

The system is also a copywriting trainer. Even with 20 years of experience, Justin says the feedback regularly surfaces blind spots and teaches him better approaches. A recent $36,000 offer used bullet points the AI rewrote because its version was simply stronger than what he had originally written.

Anyone can use this, not just expert copywriters. Because the workflow takes whatever you input — even a bad ad — and improves it based on real persona feedback, non-copywriters like a front-desk employee or a local business owner can now produce top-10% ad copy without prior training.

Join the Build With AI community built for non-technical entrepreneurs: https://www.skool.com/buildwithai/about

If this episode was valuable to you, it would mean a lot if you left a rating and review on Apple Podcasts or Spotify. It helps more entrepreneurs find the show.

FIND ME ON SOCIAL
X/Twitter: https://x.com/coreyganim
Instagram: https://www.instagram.com/coreyganim/
LinkedIn: https://www.linkedin.com/in/coreyganim/
YouTube: https://www.youtube.com/@coreyganim

FIND JUSTIN ON SOCIAL
https://x.com/IMJustinBrooke
Website: https://adskills.com

I sat down with Nick Puru to break down the exact Claude Code setup he used to pull 4 million views and ~6,000 newsletter signups in the last 30-45 days for his AI consultancy. Nick walks through his short-form content factory inside Claude Code — the CLAUDE.md "brain," his ICP file (an avatar he calls Patrick), his foundations doc full of algorithm lessons like "negativity always wins," and the three skills he uses to write captions, generate scripts, and review them. We get into how he tests three hooks per video with Instagram trial reels, the humanizer skill from Bader on GitHub that strips out AI tells like em-dashes and "it's not X, it's Y," and why he treats every Claude project like onboarding a new employee. By the end of this episode, you'll have a clear blueprint for building your own short-form content system in Claude Code — starting from a single CLAUDE.md file and expanding from there.

Timestamps
00:00 – Intro
00:37 – 4 million views in 30 days
01:53 – Reels to ManyChat to newsletter funnel
03:31 – Tour of the Claude Code folder structure
05:02 – How the write script command works
06:29 – Why human-in-the-loop matters for content
07:38 – Inside the CLAUDE.md brain file
08:11 – Meet Patrick, Nick's ICP avatar
10:04 – Foundations file and "negativity always wins"
13:43 – Live generating a Claude Cowork script
15:44 – Testing three hooks with Instagram trial reels
18:48 – Most bare-bones version to start with
22:36 – Why ICP and brand voice are foundational
24:28 – The humanizer skill from Bader on GitHub
27:01 – Treating AI like a new employee
29:35 – Context beats prompting in 2026
31:13 – Where to find Nick

Key Points

Nick's short-form system drove ~4M views in 30 days and ~6,000 newsletter signups in 45 days for his AI consultancy — every video CTA pushes a lead magnet via a ManyChat flow that collects emails and hands off to an appointment setter.

The whole system lives inside Claude Code as three skills: one writes scripts, one reviews them against a quality checklist, and one generates captions. Trigger them by typing things like "write script" in the terminal.

The CLAUDE.md file is the brain. It holds the output format, writing rules, key principles, and a pointer to an ICP file built around an avatar named Patrick — a small-to-mid market business owner ($few hundred K to $15M, 2-50 staff) who has tried ChatGPT once or twice but doesn't know Claude Code.

One of Nick's foundations is "negativity always wins" — the algorithm rewards a stronger emotional charge, but he warns against using it on every video or audiences pick up on it. He calls it in only when the angle fits.

He tests three different hooks on the same body and CTA using Instagram trial reels, treats it like A/B testing titles and thumbnails on YouTube, and feeds the winners back into the system as analytics context.

To kill AI tells like em-dashes, bullet points, and "it's not X, it's Y," Nick runs scripts through Bader's humanizer skill from GitHub. Corey adds a similar instruction in his agents.md to never use dashes.

Treat Claude like a new employee, not a magic box. You wouldn't expect a hire to crush it on day one — you'd give them SOPs, business context, your website, and an ICP. Same playbook for building any Claude project, whether it's short-form, long-form, or LinkedIn posts.

Bader's humanizer skill on GitHub - https://github.com/bader-research

ManyChat for Instagram DM automation - https://manychat.com

FIND ME ON SOCIAL
X/Twitter: https://x.com/coreyganim
Instagram: https://www.instagram.com/coreyganim/
LinkedIn: https://www.linkedin.com/in/coreyganim/
YouTube: https://www.youtube.com/@coreyganim

FIND NICK ON SOCIAL
YouTube: https://www.youtube.com/@nickpuru
X: https://x.com/nickpuru
LinkedIn: https://www.linkedin.com/in/nickpuru/

Build your voice agent in 10 minutes (No code required)
https://corey-ganim.kit.com/18da152cf4

In this episode, I sat down with Susan Westwater, co-founder of Pragmatic Digital and a conversational AI veteran since 2017, to build a working voice agent in Voiceflow live, with zero code. Susan walks through her exact prompt template for an appointment-scheduling agent, explains the difference between in-the-loop, on-the-loop, and fully autonomous agents, and shows how to separate your agent instructions from your knowledge base so you can update facts without breaking the whole build. We test the agent live (it called her cell phone mid-episode), connect it to a Google Sheet to capture customer intake, and talk about competitors like VAPI, ElevenLabs, and Voiceify. By the end, you'll have the exact playbook (and the template Susan is giving away) to spin up your own MVP voice agent in under an hour.

Timestamps
03:42 – Inside Voiceflow and what makes it different
04:23 – Chatbot vs voice agent: listen, decide, act
05:30 – In the loop vs on the loop vs fully autonomous
09:09 – The system prompt: identity, purpose, and boundaries
11:54 – Voice, persona, and speech sculpting
13:22 – Separating agent instructions from knowledge base
17:14 – Stopping the agent from troubleshooting electrical issues
20:31 – Rules for collecting info one question at a time
23:21 – Pasting the prompt and one-shotting the build
26:34 – Voiceflow competitors: VAPI, ElevenLabs, Voiceify
30:22 – Connecting tools at each conversation step
33:13 – Why faster builds give up control
38:21 – Adding the knowledge base as a Word doc
43:13 – Sending customer intake to a Google Sheet
45:57 – Live phone call with the voice agent
50:09 – Branching, exit conditions, and iteration
54:52 – Custom voices, ElevenLabs integration, and voice security

Key Points

Keep two documents separate: agent instructions (how the agent behaves, its identity, persona, escalation rules) and knowledge base (the facts about your business). If pricing or hours change, you update one cell in your knowledge base instead of digging through a giant system prompt.

The strength of a Voiceflow build lives in the prompt. Susan pastes her full instructions doc into the new project prompt and Voiceflow generates the entire conversation flow — greeting, qualification, intake, confirmation, escalation — with no coding.

Be explicit about what the agent does NOT do. LLMs are trying to "win the game" (Susan's War Games analogy), so if you don't tell the electrical-appointment bot "do not troubleshoot," it will try to problem-solve its way out of every conversation.

Collect intake one question at a time, use explicit confirmation on critical fields like callback numbers, and tell the agent to be empathetic but not apologetic — nobody wants a bot that says "I'm sorry" five times instead of solving the problem.

Voiceflow — https://www.voiceflow.com

Pragmatic Digital — Susan's agency helping brands operationalize conversational AI and applied AI for CX - https://www.pragmatic.digital

VAPI — alternative voice agent platform - https://vapi.ai

ElevenLabs - https://elevenlabs.io

Voiceify — centralized library for chatbot, voice agent, and telephone deployments across web and mobile - https://www.voiceify.com

Twilio — phone number and telephony layer for connecting Voiceflow agents to real phone calls - https://www.twilio.com

Make.com — automation platform that can connect to Voiceflow as a tool integration - https://www.make.com

Robert Scoble on X https://x.com/Scobleizer

Brian Roemmele on X https://x.com/BrianRoemmele

FIND ME ON SOCIAL
X/Twitter: https://x.com/coreyganim
Instagram: https://www.instagram.com/coreyganim/
LinkedIn: https://www.linkedin.com/in/coreyganim/
YouTube: https://www.youtube.com/@coreyganim

FIND SUSAN ON SOCIAL
X: https://x.com/sjw75
Website: https://www.pragmatic.digital

🔗 Deploy your own Hermes agent on Hostinger -10% off any plan with code COREY10: http://hostinger.com/corey10 (use code COREY10)

Grab the free step-by-step Hermes Second Brain setup guide here: https://corey-ganim.kit.com/60fc0fe6d9

In this solo episode, I walk through exactly how I built the ultimate AI second brain using Hermes and its built-in LLM Wiki skill — inspired by Karpathy's wiki setup. I break down the three core operations of the Hermes agent (ingest, query, lint), the three layers of the knowledge base (raw sources, the wiki, schema/tags), and why I chose Hermes over Claude Code, OpenClaw, or Obsidian for this build. From there, I show the full one-click deploy on a Hostinger VPS, the OpenAI model setup, the Telegram bot integration via BotFather, and the Markdownload Chrome extension trick I use to feed any tweet, article, or webpage straight into the wiki. By the end of this episode, you'll have a clear, step-by-step path to deploy your own self-improving second brain that gets smarter the more you use it.

Key Points
Hermes ships with a built-in LLM Wiki skill that handles the second brain function out of the box — your job is to curate sources, the agent's job is to summarize, tag, and file them away.

The knowledge base has exactly three layers: raw sources you give it (read-only), the agent-owned wiki of markdown files, and the schema/tags Hermes builds to make querying low-friction.

Run the Lint command roughly once a month — it audits your wiki for contradictions, sources older than 90 days, and oversized files that should be split for more accurate retrieval.

Deploy on a Hostinger VPS instead of locally — it's a one-click install with no terminal, cheaper than a Mac mini, always-on, and your data stays private.

The simplest workflow for feeding the wiki: use the Markdownload Chrome extension to clip any tweet, article, or webpage to markdown, drag it into your Telegram chat with Hermes, and tell it to add it to the wiki.

This is a compounding asset — on day 1 the knowledge base is the dumbest it'll ever be, but feed it consistently and by day 90 it becomes incredibly useful for retrieval.

Hermes agent VPS deployment on Hostinger (use code COREY10 for 10% off) - https://www.hostinger.com/vps/hermes-agent-hosting

Markdownload Chrome extension - https://chromewebstore.google.com/detail/markdownload-markdown-web/pcmpcfapbekmbjjkdalcgopdkipoggdi

Telegram BotFather - https://t.me/BotFather

OpenAI Codex - https://openai.com/codex

FIND ME ON SOCIAL X/Twitter: https://x.com/coreyganim Instagram: https://www.instagram.com/coreyganim/ LinkedIn: https://www.linkedin.com/in/coreyganim/ YouTube: https://www.youtube.com/@coreyganim

Tom Crawshaw (8+ years in automations, $25M+ in attributed e-com revenue) cracks open the content system that's pulled him millions of views. We walk through the Claude Code skill he built that handles voice profile, copywriting principles, hook scoring, image prompts, and a humanizer pass from one slash command. He also surfaces a hidden Claude Code feature called /insights that audits your full usage history and tells you what to build next. You'll learn how to turn messy ChatGPT workflows into a real skill, why Whisperflow is non-negotiable, and the one slash command 99% of Claude Code users are sleeping on.
Timestamps
00:00 Intro
00:30 Tom's automation background
03:30 Skills vs. Projects on context
08:30 Auto-updating voice profile from X
11:00 Nano Banana plus Canva workflow
23:30 Live demo: a Whisperflow post
34:30 Hook scoring with copywriting principles
46:30 The hidden /insights command
54:30 Skills vs. n8n vs. Lovable
Key Points
Skills beat Claude Projects on context. Projects load every reference file on every message. A skill works like a book: Claude pulls only the chapter it needs, so a full pipeline runs in one chat.
Tom's skill auto-updates his voice profile weekly. The X API pulls his last 7 days of posts, ranks by engagement, and rewrites the profile so the skill keeps drifting toward what's working.
/insights is the most slept-on feature in Claude Code. It audits your full session history and hands back a real report: what's working, where you're breaking your own rules, skills to build, and prompts to run.
Image gen is 80% Nano Banana, 20% Canva. Generate fast in the AI tool, then finish in Canva with magic grab and magic erase.
Whisperflow changed how Tom and Corey think, not just type. Speaking forces tighter thinking and makes prompts richer because adding context costs almost nothing.
Links
Claude Code: https://www.anthropic.com/claude-code
Whisperflow: https://wisprflow.ai/
Nano Banana: https://gemini.google.com/
Canva: https://www.canva.com/
n8n: https://n8n.io/
Tom's site: https://learnn8nautomation.com/
Build With AI: https://www.skool.com/buildwithai/about
FIND ME ON SOCIAL
X: https://x.com/coreyganim
Instagram: https://www.instagram.com/coreyganim/
LinkedIn: https://www.linkedin.com/in/coreyganim/
YouTube: https://www.youtube.com/@coreyganim
TOM ON SOCIAL
X: https://x.com/TomCrawshaw01
YouTube: https://www.youtube.com/@TheAIGrowthLab
Website: https://learnn8nautomation.com/