About the lab

One operator directing a roster of codified AI implementers. Mastery-tier execution at boutique-agency prices. The lab is the output of decade-plus practice hitting a forcing function: cross-disciplinary marketing engineering needs operators who carry code, copy, sales psychology, CRO, UX, and agentic systems engineering at depth, and that profile is rare at any price.

Decades, named

The substrate underneath

Brandon McPeak built his first website in 2008 with HTML and CSS to land a job at Naked City Magazine, a Wichita culture-and-arts publication. The site worked. The job followed. Other businesses started asking him to build theirs.

College ran on a double major: journalism plus cultural anthropology. Heavily involved in the student newspaper; copy-editing internship at the Wichita Eagle. The interviewing skills became deep-dive client intake. The marketing degree came next, after the technical-and-architecting strength surfaced first.

A move to Austin pursued tech-world opportunities. By 2019, freelance work was full-time. The roles in between built the substrate: SEO specialist promoted to senior SEO at a cutting-edge unit; designed an in-house messaging-and-branding service alongside the SEO offering; marketing manager; web experience developer at Serenova running ads alongside dev work in B2B SaaS contact-center software.

Baker Johnson, then VP Marketing at Serenova, named the pattern publicly: “marketing unicorn… the prototype for a new breed of Marketing DevOps professionals.”

Same operator. Different stack now.

Differentiator

Deep intake.
Per-client copywriting bots.

Every engagement opens with a deep-dive interview. Web scrape on top. Public voice-of-customer data layered in (reviews, comments, transcripts, support tickets). The combined substrate gets turned into a structured per-client knowledge set, then into an expert copywriting bot trained on that client's actual voice.

Most agencies do not do this depth. Clients consistently report being impressed by the intake itself, before any copy ships. The work that follows is on-brand because the knowledge base is built from the client's own substrate, not adapted from a generic template.

The codified practice

The codified practice

The lab is a roster of codified processes that operate as implementers under one operator's pilot oversight. Each one a real system, named and reachable.

The Odyssey design skill

Produces layouts, patterns, and UX scaffolds.

The ad-ops skill

Runs Google Ads optimization. Codified from career-spanning managed spend approaching $1M.

The custom recording system

Captures intake interviews and client meetings.

GTM and GA implementations

Baked into every build at elite-tier depth.

Per-client copywriting bots

Train on each client’s voice substrate.

TITAN Research Engine

Public at titanresearchengine.com. Multi-agent CLI research with adversarial filtering, triangulation, and a red-team synthesis pass.

The agency analytics + reporting backend

Vercel, Inngest, Tremor, Anthropic API. Inngest functions retrieve data; the Anthropic API drafts weekly client reports; Tremor renders the dashboards.

HarnessLab

Governs how new lab implementers get built and validated.

Working with the lab

Working with the lab

Selective by design. The lab takes a small number of engagements per cycle, premium-priced, intentionally limited so the depth of the practice survives delivery. Most engagements start with a deep-dive conversation, scope into a three-phase value-ladder, and ship in roughly half the timeline of a traditional small agency on equivalent depth.

Phase 1 gets you off the old stack and onto the integrated foundation. Phase 2 surfaces from real-user data 30 to 60 days post-launch. Phase 3 handles enhancement and system depth as the foundation compounds.

Brandon is also currently exploring senior full-time and fractional roles where this same practice scales to company-wide application. The lab continues either way.

Client engagements

Five operators we've shipped for

Five engagements across the past several years: tattoo studios, restaurants, founder-led service brands, early-stage B2B work. Each one a different shape; each one shipped on the same integrated foundation.

  • Mike Davenport Tattoo Artist
  • Mike Henn Owner, Safeview Auto Glass
  • Zach Rubin CEO, PBC Guru
  • Ron Ronhaar Owner, The Grooves Band
  • Craig Dacy Financial Coach

Common questions

Before you start the conversation

How is this different from a freelancer with a ChatGPT subscription?

A freelancer with a ChatGPT subscription has access to the same general-purpose model as anyone else. The lab is a roster of codified processes, named and reachable: the Odyssey design skill, the ad-ops skill, the custom recording system, per-client copywriting bots trained on your own voice substrate, the agency analytics and reporting backend on Vercel, Inngest, Tremor, and the Anthropic API. Each one is a real system Brandon built across years of practice, not a chat session. The result is closer to having a senior operator running a small, codified team than to a freelancer renting a chatbot.

What types of engagements is the lab a fit for?

Founders and small teams who want mastery-tier marketing engineering done as one integrated practice rather than stitched across three vendors. Site rebuilds on Astro or Next.js with the analytics, ads, SEO, and lead-delivery email all on the same foundation. Ongoing Google Ads management on top of that for the engagements that warrant it. Verticals served today include tattoo studios, restaurants, and founder-led service brands. Budget shape: SMB to mid-market, premium-priced for the depth.

Why such a small caseload?

The depth of the practice does not survive volume. Each engagement opens with a deep-dive intake that produces a per-client knowledge set and an expert copywriting bot trained on that client’s actual voice. Each one is a foundation the rest of the work compounds on. Brandon runs a small number per cycle on purpose. Selective by design, premium-priced.

What technologies does the lab build on?

Astro and Next.js for the frontend, with Sanity or Strapi where the content scope warrants a CMS. Vercel hosts production. Inngest handles scheduled jobs and background workflows. The agency’s analytics and reporting backend runs on Vercel, Inngest, Tremor, and the Anthropic API. GTM and GA implementations get baked into every build at elite-tier depth. Internally, the codified-process layer runs on Claude Code with the Odyssey design skill, the ad-ops skill, the custom recording system, and a library of per-client copywriting bots underneath. Schema-first, component-driven, mobile-app-grade UX as a default.

Does the lab take on ongoing Google Ads management?

Yes, primarily Google Ads. Roughly $1M in managed ad spend across years on a reliable optimization system targeting high-value, low-demand search. The ad-ops skill runs the day-to-day dial-turning; Brandon directs the strategy. Most site-rebuild engagements transition into ongoing ad management when the foundation goes live, because the integrated stack makes ad performance compound off the conversion baseline. Meta is sparing. The one-operator high-touch model does not stretch across both platforms by default.

What about ongoing optimization after launch?

Surfaces 30 to 60 days post-launch from real-user data running through the analytics layer that ships with the foundation. Conversion-pattern audits against actual traffic, review-count social proof additions, A/B test setup, additional ad landing pages where the search demand maps, deeper analytics-tracking-reporting infrastructure. This is Phase 2 of the engagement arc on the Home page. The optimization rhythm is what dials to turn, when, and against what ROI signal. Built from years of A/B testing patterns.

Is Brandon exploring full-time or fractional roles?

Yes. The same codified practice that runs the lab also scales to company-wide application inside a senior in-house or fractional role. Brandon is open to the right opportunity. Full FT-search signal lives on the BrandonMcPeak.com umbrella; this answer is the single-paragraph version on the lab surface. The lab continues either way.

Start a conversation

Selective caseload. Replies within 2-3 business days. If the fit is wrong, you'll get a fast no.