Marketing Speedrun

Mason Warner is Creative Director at Slash, a neobank going after Wells Fargo and Chase with a software-first banking platform. He runs the entire marketing operation with 2 people and a stack of AI tools that produces content at agency scale. In this conversation, Mason maps out every system behind it, from the $350 AI video that beat $60K in influencer spend to the internal AI agent that generates pitch decks from Slack in under 2 minutes.

In this episode, we cover:

1. How Mason's $350 AI video (Margot Sloppy) outperformed a $60K influencer campaign in organic impressions
2. The tool stack behind 70-80% AI-generated marketing media at a fintech startup
3. Why Mason says Claude Design is trash and uses Figma MCP for all design work instead
4. Twin, Slash's internal AI agent that generates one-pagers, pitch decks, and sales assets from Slack
5. The $80 brain rot shooter game that landed on Polymarket, Business Insider, and 20+ publications
6. How Mason picks models for different tasks: GPT 5.6, Grok 4.5, Fable Five, Opus 4.8
7. Melius, the agent-first content canvas replacing manual node-by-node AI video production

(00:00) Introduction to Mason Warner
(01:32) From YouTube kid to creative director
(03:31) Organic TikTok marketing before paid ads
(05:38) Running a fintech marketing team of 2
(07:45) The Margot Sloppy VO3 ad origin story
(12:56) $350 in AI credits vs $60K on influencers
(15:52) 8 AI launch videos in 3 months
(19:14) Why Claude Design is trash
(23:13) Twin: Slash's AI agent for sales assets
(26:43) Model picking: GPT 5.6, Grok 4.5, Opus 4.8
(28:05) The $80 brain rot game that hit Polymarket
(36:12) In-game advertising and 20+ press articles
(39:45) Being fun in B2B without losing enterprise
(46:29) Melius: the agent-first content canvas
(55:59) From 1000 ads/week to instant with AI
(59:43) North Star: one-shot marketing with zero QA

Where to find Mason Warner:
• X: https://x.com/masonwarner
• LinkedIn: https://www.linkedin.com/in/masonwarner
• Company: https://slash.com
• Slash on X: https://x.com/slashapp

Where to find Polina:
• X: https://x.com/kalashnikovapv
• LinkedIn: https://www.linkedin.com/in/polina-kalashnikova
• Website: https://polium.co

Marketing Speedrun is a podcast for founders, CMOs, and growth leaders building in AI and emerging tech. We talk to operators running marketing and growth at frontier companies and get them to share the fastest path to what's actually working.

Subscribe: https://www.youtube.com/@marketingspeedrun

Vaibhav Aggarwal is a growth engineer for ChatGPT at OpenAI. Vaibhav thinks about growth as interconnected loops where the opportunity lives in the handoffs between acquisition, activation, retention, and referral. In this episode, he breaks down the 4-layer growth system they’ve built, what was missing when he joined OpenAI, and the most underrated growth assets.
In this episode, we cover:

1. How OpenAI's growth team is structured to take ChatGPT from hundreds to a billion users
2. The 4-layer growth framework every startup should build
3. How Vaibhav turned Codex into his AI chief of staff
4. What was missing from ChatGPT's growth stack when he joined
5. Why retention systems beat acquisition hacks
6. The case for automating growth engineering
7. OpenAI's portfolio-of-bets approach

(00:00) Meet Vaibhav Aggarwal from OpenAI
(03:54) Surviving crypto cycles at Coinbase
(06:15) Powerful notification loops
(08:15) What being a "growth engineer" means
(09:52) How OpenAI's growth team is structured
(11:15) The 4-layer growth framework for startups
(13:10) Building an AI chief of staff with Codex
(20:12) What ChatGPT was missing at near-1B users
(30:51) The case for automating growth engineering
(41:05) Comparing crypto vs AI cycles
(42:41) OpenAI's portfolio of bets approach
(44:25) What's next for ChatGPT's consumer push

Where to find Vaibhav Aggarwal:
• LinkedIn: https://www.linkedin.com/in/vaibhavaggarwal

Where to find Polina:
• X: https://x.com/kalashnikovapv
• LinkedIn: https://www.linkedin.com/in/polina-kalashnikova
• Website: https://polium.co

Marketing Speedrun is a podcast for founders, CMOs, and growth leaders building in AI and emerging tech. We talk to operators running marketing and growth at frontier companies and get them to share the fastest path to what's actually working.

Subscribe: https://www.youtube.com/@marketingspeedrun

Arnaud Simeray is VP of GTM at Dune, the leading on-chain data provider with over 1 million users and 30,000 companies on the platform. When Arno joined 3 years ago, Dune was a crypto-native B2C product, mostly known for dashboards shared on X. He built the entire GTM function from scratch, transitioning the company to enterprise B2B serving hedge funds, banks, regulators, and some of the biggest names in crypto and TradFi. 

In this episode, we cover:

1. How Arno built Dune's GTM team from zero, moving from B2C dashboards to B2B enterprise
2. Why Dune runs outbound campaigns of 50-55 people instead of thousands, and why conversion is higher
3. Claude workflows he swears by
4. How Dune uses customer call language and vocabulary to write outbound and marketing copy
5. Why the Dune MCP is "millions of dollars in analysts packed in one URL"
6. The research report playbook: why distribution is half the battle, not the writing

(00:00) Intro
(00:24) Arnaud’s journey from engineering to GTM
(03:36) What Dune is
(05:35) What a VP of GTM does
(09:06) Building the GTM function from zero: no marketing team existed
(11:39) What’s working for Dune’s growth
(13:04) Why 99% of LinkedIn outbound is lazy
(19:20) CRM setup: HubSpot as the hub, Claude as the brain
(21:50) Arnaud’s Claude workflows
(33:44) Dune research reports for lead gen
(37:55) What is the Dune MCP and why it matters
(44:03) MCP use cases: trading, prediction markets, stablecoin flows
(48:54) Advice for crypto B2B: ICP first, quality data, clear positioning
(52:07) What's next

Where to find Arno Simeray:

• LinkedIn: https://www.linkedin.com/in/arnosimeray
• X: https://x.com/arno393
• Dune: https://dune.com

Where to find Polina:

• X: https://x.com/kalashnikovapv
• LinkedIn: https://www.linkedin.com/in/polina-kalashnikova
• Website: https://polium.co

Marketing Speedrun is a podcast for founders, CMOs, and growth leaders building in AI and emerging tech. We talk to operators running marketing and growth at frontier companies and get them to share the fastest path to what's actually working.

Subscribe: https://www.youtube.com/@marketingspeedrun

Corey Haines is the founder of Conversion Factory, a B2B SaaS marketing agency, and the creator of the Marketing Skills GitHub repo that became one of the most starred marketing repos on the platform within days of launching. After a decade in SaaS marketing, a 2-week coding bootcamp, and years of building products like TrueList and Swipe Files, Corey now sits at the exact intersection of marketing and engineering that most people only talk about.

In this episode, he explains why 90% of AI CMO products are built by people who don't understand marketing, what he's building differently with Magister, and the exact tool stack and workflow he uses to run multiple SaaS businesses from Claude Code.

In this episode, we cover:
(00:00) Introduction to Corey Haines

(02:01) Why a marketer signed up for a coding bootcamp

(04:39) Learning the hard way just in time for AI

(06:39) How the Marketing Skills repo started

(09:27) GitHub as a versioned marketing playbook

(10:49) Open sourcing: risk of theft vs. first mover

(12:53) Skills.sh and the security problem with AI skills

(14:20) Which skills marketers actually use most

(17:28) Why most AI CMO products are a scam

(19:26) What a real AI CMO needs to cover

(22:27) Daily stack: Solo, Warp, Claude Code, Obsidian

(30:43) Dog fooding skills across multiple SaaS products

(33:07) Autonomous Google Ads management with Magister

(38:43) Taste: the woodworking analogy for AI design

(41:05) AI will hit A+ design in 6 months

(44:41) Why he stopped listening to business podcasts

(47:03) The bankruptcy hack: being resourceful with AI

(49:37) What's next: Magister and AI training

Guest Links:

Where to find Corey Haines:

• LinkedIn: https://www.linkedin.com/in/corey-haines/
• X: https://x.com/coreyhainesco
• GitHub (Marketing Skills): https://github.com/coreyhainesco/marketing-skills

Where to find Polina:

• X: https://x.com/kalashnikovapv
• LinkedIn: https://www.linkedin.com/in/polina-kalashnikova
• Website: https://polium.co

Subscribe: https://www.youtube.com/@marketingspeedrun

Drew Rogers is the founder of Stabledash, the stablecoin media company that hit 12,000 listeners in its first 30 days and now runs live shows twice a week with 3-5 guests per session. With one engineer, Drew's team built their own recording platform, automated clipping pipeline, guest portal, and sponsor integration system, effectively replacing Riverside, Opus Clip, and manual distribution workflows. In this episode, he maps out the full system: how Stabledash went from a side-quest podcast to a media business with its own tech stack, segment sponsorships, and a content flywheel that turns live interviews into a revenue engine.

In this episode, we cover:

1. How Stabledash built a Riverside replacement with 1 engineer using Claude Code and custom tooling
2. The guest portal system that routes applicants to sponsor-aligned segments automatically
3. Why Drew thinks the hardest problems to solve with AI are the ones worth building around
4. How automated clipping identifies 3-5 key moments per guest and distributes clips within 24 hours
5. The pivot from GTM services to media company, and why podcast PMF forced the decision
6. How segment sponsorships (like "Neo powered by Dakota") make advertising feel native, not extractive
7. The Canva + Nano Banana + FAL pipeline that generates studio-quality thumbnails from live screenshots
   

Timestamps:
(00:00) Introduction to Drew Rogers
(01:15) From Square to Coinbase Commerce
(03:30) Why media is the best GTM strategy
(04:28) Stabledash as a new category of fintech media
(05:33) Finding the hardest problems to solve with AI
(07:36) What Stabledash automates vs. keeps human
(09:00) Building a Riverside replacement in-house
(12:18) One engineer, infinite leverage
(13:17) Guest discovery and research workflows
(14:23) How the stablecoin podcast found PMF
(16:04) The design taste problem in AI-generated assets
(19:42) Claude for brand PDFs and design standards
(22:02) The Stabledash tools dashboard
(22:37) Guest onboarding portal and segment routing
(25:47) Run of show automation and studio product
(26:53) Why good advertising feels like content
(29:24) Personal tool stack: Claude, Hermes, Cursor, Canva
(32:12) Canva + Nano Banana thumbnail pipeline
(36:19) Post-show clipping and distribution system
(39:29) Turning media guests into a sponsor funnel

Where to find Drew Rogers:

• X: https://x.com/drewrogers

• LinkedIn: https://www.linkedin.com/in/drewrogers
• Stabledash: https://stabledash.com

Where to find Polina:

• X: https://x.com/kalashnikovapv
• LinkedIn: https://www.linkedin.com/in/polina-kalashnikova
• Website: https://polium.co

  Subscribe to Marketing Speedrun: https://www.youtube.com/@marketingspeedrun

Romain Marsal is VP Growth and Marketing at Vibe.co, the streaming TV ad platform that scaled from $10M to $300M ARR in under 2 years and is targeting $1B by end of 2027.

In this episode we discuss:
1. How a lean growth team operates $2M/month in paid across multiple platforms
2. How they generate 30 creative concepts/week with automated winner detection
3. Why Romain is replacing inbound SDR with agents, and reallocating those people to outbound
4. The warm-score system bundling tech-stack scraping, LinkedIn signals, and product activity into one outbound priority number
5. How Vibe put 20% of total marketing budget into out-of-home with zero attribution, and why it still 3X'd their growth
6. The AEO content engine shipping 40 pieces per week with Claude Code, Ahrefs, AirOps, and Dato CMS
7. Why Meta MCP, Google MCP, and Vibe's own MCP eliminated the need for platform-specific paid ads expertise

In this episode, we cover:
(00:00) Introduction
(01:46) What Vibe.co does
(03:44) Engineer to growth leader
(08:37) Team structure: domain owners + GTM engineers
(10:18) $2M/month paid and the creative product
(12:15) New channels: OOH, AEO, creators
(13:43) Dogfooding Vibe as B2B case study
(16:09) AEO content engine: 40 pieces/week
(19:04) Two-tier content: AI vs human
(20:40) Thought leadership from sales calls
(22:17) Reddit: 400K impressions, bot issues
(25:36) 50 AEs and replacing inbound SDR
(27:40) Headless CRM for AEs
(30:21) The warm-score system
(33:00) Scraping tech stacks for buying signals
(36:00) OOH: 20% budget, zero attribution
(39:09) Billboards people screenshot
(44:15) AI stack: Claude Code + Managed Agents
(46:17) How MCPs killed platform UI expertise
(49:54) What's next: Netflix, agents, measurement
(51:07) Where to find Romain

Where to find Romain:
- LinkedIn: https://www.linkedin.com/in/romainmarsal
- Company: https://vibe.co

---

Where to find Polina:
- X: https://x.com/kalashnikovapv
- LinkedIn: https://www.linkedin.com/in/polina-kalashnikova
- Website: https://polium.co

Marketing Speedrun is a podcast for founders, CMOs, and growth leaders building in AI and emerging tech. We talk to operators running marketing and growth at frontier companies and get them to share the fastest path to what's actually working.

Subscribe: https://www.youtube.com/@marketingspeedrun