Harvey Carnegie runs a portfolio of DTC brands that did $75M in revenue last year and is on pace to hit a $250M run rate in 2026. Across the portfolio, Harvey runs primarily on native ad platforms, with Meta making up roughly 25% of revenue, and a creative engine that ships around 100 new ad tests every week across four creative strategist pods. Harvey has been a CA Pro member (Stefan's mastermind for D2C operators doing $1M+/mo) for two years and was one of the earliest operators to fold StefanBrain into his daily workflow.

Scaling 100 ad tests a week without burning the in-house team on flop ideas

On a typical week, Harvey's pods ship ~100 new creatives across native and Meta. Every brief that gets handed to a video editor or designer burns hours of pod time. If the underlying idea turns out to be a flop, that's lost capacity that should have gone into iterating on a winner.

Harvey lived that bottleneck firsthand. As the founder, Harvey writes copy himself, manages four creative pods, and pushes new ideas straight into the ad account. His team has shrunk from 150 to roughly 105 over the past year as AI took over chunks of customer support and creative production, but creative strategy still has to be human-led. And the creative strategist seat is the hardest in his org to fill:

So Harvey did what most sharp founders try first. He tried to build it himself in Claude. He set up Claude projects with marketing briefs, training docs, and product context loaded into the system prompt, then ran his scripts and ideas through them. The problem was always the same. He had to manually feed Claude the same context every time he opened a new chat. And once the context window filled up with PDFs, the conversations would bloat and start contradicting themselves.

Claude isn't trained on the marketing reps Harvey actually needed feedback from. In his words, it's good, but it's not that good once you strip away the manual context. He needed something that already understood direct response, native ad flows, hook structure, and what kills conversions on a landing page.

A creative validation engine and a same-day video lab, both running on StefanBrain

Harvey started using StefanBrain soon after it launched exclusively for CA Pro members. He was already a member, so the early experience was less about onboarding and more about plugging into a system that was already aligned with how he thinks about creative. Two workflows took hold inside the first few weeks, and both are still running today.

Harvey's default loop is to write copy himself, then drop the script and the landing page it's pointing at into StefanBrain and ask why it works (or why it's about to fall flat). The tool breaks down which messaging beats are landing, which ones are off-pulse with the lander, and where the disconnect between ad and page is leaking conversions. He does the same thing with raw post-purchase survey data. He drops it in, asks for the read, and gets a straight callout on the gap between what customers actually say and what the page is leading with.

The reason this works is the training under the hood. StefanBrain is built on seven-plus years of Stefan personally reviewing operators' ads, email flows, product pages, landing pages, and full marketing flows, plus the deeper teaching on hooks, big ideas, and what to test next. As Harvey put it:

The second workflow is what compressed Harvey's testing cycle from weeks to hours. When he has a new angle he wants to throw at the ad account, he uses StefanBrain's Video Gen to produce a rough four-minute creative directly from his own script, then pushes it live without polish. The output is clean enough to drop straight into the ad account.

If the rough creative pulls spend, Harvey hands the brief over to his in-house AI and video editor team to produce the polished version. If it flops, no team time was burned. He uses the Static Ad Generator the same way for production-ready statics. The result is that he can prototype, test, kill, or scale a new angle in the same day, instead of waiting on a week-long creative cycle to find out the idea was a dud.

Harvey scaled his portfolio to a $250M run rate without growing the team

In the past 12 months, Harvey has scaled his portfolio of DTC brands from $75M to a $250M run rate target while pushing around 100 creative tests per week through native and Meta. StefanBrain became the validation layer underneath that engine. Every new angle gets pressure-tested against Stefan's marketing brain before a single video editor touches it, and Harvey can ship rough creative the same day to see if an idea pulls spend in the ad account.

Looking ahead, Harvey is testing new product lines on top of the existing portfolio, with native ads as the primary scaling channel.