Latent Space: The AI Engineer Podcast

Episode: AI Video Is Eating The World — Olivia and Justine Moore, a16z

Date: July 9, 2025
Guests: Justine & Olivia Moore (Andreessen Horowitz/a16z)
Host: Alessio (Decibel), with wix (Small AI)

Episode Overview

This episode dives deep into the explosive rise of AI-generated video, its impact on pop culture, creative economies, and the tech stack enabling this next generation of content. Olivia and Justine Moore—identical twins, a16z partners, and active participants in generative media—join Latent Space to discuss viral Internet memes (“brain rot”), monetization opportunity, content trend-tracking, developer platforms, and the future of creative work. The conversation is rich with examples and practical insights both for creators and those building AI tools.

Key Discussion Points & Insights

1. The Viral Explosion of AI Video Content (00:42–02:51)

• Rapid Adoption:
  Justine & Olivia share their journey from early Stable Diffusion experiments (~late 2022) to today’s mass adoption, noting a swift shift as even their creative friends now seek AI video advice.

  "It used to be our friends would look at AI image and video generation and be like… 'I'm not worried about that'. And now, we'll literally have people coming over to our house on the weekend for tutorials." — Justine (01:37)

• AI Video’s Social Platform Takeover:
  Olivia notes, “90% of your feed is AI generated video,” referencing TikTok, Reels, and Shorts. Formerly fringe meme genres—like the “Italian Brain Rot”—now represent legitimate mainstream IP.

  “Even two months ago, it was just a few people making AI video. Now there's hundreds of thousands making and publishing, which is just awesome.” — Olivia (02:24)

2. Case Studies: Internet Memes and Community Remixing (03:04–04:44)

• Decentralized Meme Evolution:
  “Italian Brain Rot” started as images, then moved to videos. Characters get remixed, universe grows, and fans—kids and adults—relate to them as real TV stars.

  “The IP has evolved to people selling toy sets and plushies… A kid knows them by heart and treats them like Nickelodeon characters.” — Olivia (03:57–04:30)

• Speed of Character Recognition:
  AI video enables high-frequency, fan-driven content creation—unlike traditional media publishing cycles—which accelerates attachment.

  "Dozens, if not… hundreds, of videos every day. You get attached fast." — Olivia (04:30)

3. Trend Tracking, Model Evolution, and Platforms (05:00–10:47)

• Where Trends Originate:
  Early trends began on Reddit, then surfaced on Twitter/X, eventually exploding on TikTok/Instagram. Today, origin flips: viral content now arises from everyday users on consumer video platforms, not just technical circles.

  "Most viral content now is originating on the consumer platforms—TikTok and Instagram." — Justine (05:25)

• Experimentation with Models and Content:
  Olivia describes using Google’s VO3, Minimax, and other models to create viral ASMR and character videos, highlighting the trend-based, rapid experimentation and how certain genres (e.g., "eating lava" or "Tide Pods") consistently perform.

  "The trend that always works, like, in any format is lava. People love eating lava, squishing lava… People are just obsessed with it." — Olivia (07:39)

• Model Friction and Limitations:
  VO3 lacks image-to-video with audio, impacting character consistency. As a result, creators lean on known characters (Stormtrooper, Stitch, Jesus) for reliable outputs.

  "You can't do image to video with audio… That makes it super hard for character consistency." — Justine (09:57)

4. Creative Dynamics: Familiar vs. Novel IP (12:22–15:50)

• Power of Familiarity:
  Using existing IP (e.g., Star Wars) draws in established fanbases, but the twist comes from putting characters in unexpected contexts.

  "There are benefits to starting with established IP—you're tapping into a known association… But weird, entirely new stuff can work too." — Justine (12:49)

• Birth of New AI-Native Characters:
  Original creations like “Kim the Gorilla” thrive due to the newfound ease of building compelling content—even without technical, filmmaking, or acting credentials.

  "Who knows if this person would’ve been able to make this content before? Now, with VO3, anyone can come up with a fun idea and keep it going." — Olivia (15:27)

• Remixing as a Creative Engine:
  AI creators often remix each other’s ideas, iterating on concepts (e.g., Yeti to Stormtrooper to Gorilla vlogs). The meme universe constantly evolves.

  "Many are just literally remixing other's work, and the universe evolves from that." — Justine (16:07)

5. Monetization Strategies and Challenges (18:35–23:12)

• Emerging Revenue Models:
  Monetization avenues include ad revenue from social platforms, sponsored content, IP-based merchandising (toys, shirts), prompter consulting, and the prospect of AI-native shows getting picked up by networks (Netflix, Hulu).

  "Number one way: get paid by a social platform for driving eyeballs… What I'm waiting for is: what is the first AI native IP that gets packaged and bought by Netflix or Hulu?" — Justine (19:06)

• High Cost of Generation:
  Currently, producing quality AI video is still expensive—VO3’s credits and complex content require multiple expensive generations for a single publishable result.

  "Generation is still expensive enough that you have to be smart about how you'll make money from it." — Olivia (20:27)

• Creators’ Motivations:
  For many, the dopamine of going viral and growing an account is a draw in itself—not just cash.

  "The dopamine hit of getting tens of thousands of likes… is motivating a bunch of people." — Justine (22:21)

6. AI Video’s Business Stack: Models, Interfaces & Platforms (24:16–28:25)

• Distilling the Value Chain:
  The ecosystem divides into:

  1. Core model providers (VO3, Minimax, Cling, etc.)
  2. Application/interface/enabler layer (CREA, ComfyUI, Replicate)
  3. End-user creators.

  "We call it the interface layer or application layer—companies making it easy to use others' models, then the core model layer itself." — Justine (24:16)

• Platform Usability & Friction:
  Accessing and using VO3 (via Google’s “Flow”) is cumbersome and expensive, pushing creators towards more streamlined third-party interfaces.

  "Honestly, a ton of creators are going to the model enablement layer, because it's easier than navigating the Google product infrastructure." — Justine (25:56)

• Open-Source Workflow Tools:
  Platforms like ComfyUI serve users needing precision and complex workflows (e.g., video style transformation, consistency, anime conversion), but remain difficult for everyday creators.

  "ComfyUI is really great for people who want a ton of control… but for the average person, there's just no way." — Justine (28:43)

7. Practical Advice for New Creator Channels (30:53–41:43)

• Blueprint for AI-Driven Educational Content:
  Justine suggests short, AI-generated “Brainrot” style education clips—using celebrity deepfakes or custom characters to summarize podcasts or topics, paired with engaging visuals and B-roll.

  "What I would do… take YouTube transcripts, put them in Gemini 2.5, ask for an entertaining Brainrot style clip… animate the script and overlay diagrams or auto-generated B-roll." — Justine (33:09)

• Clipping, Virality, and Automation Tools:
  Tools like Overlap can auto-clip podcasts/youtube, predict virality, add captions, reformat for platforms, and automate posting. The technical barrier for “going viral” is falling.

  "Overlap… automatically reviews, clips the best pieces, and then publishes those clips for you… this is the dream." — Justine (35:06, 37:09)

• Platform Customization:
  Consider tailoring clips for different platforms/audiences: technical, educational, or "brain rot"/memetic. Test engagement strategies—content context, style, and form matter.

  "Maybe you have an agent… generate different clips for different audiences… try educational for YouTube, give it a prompt for 'brain rot' for TikTok." — Justine (38:11)

• Repurposed vs. Native Short-Form:
  While native shorts are often stronger, examples exist of "clipping" (repurposing) succeeding—especially with the current algorithmic moment favoring video.

  "It’s a clipping era. For some reason, video is doing super well in the algorithm… especially if it’s a character people know." — Justine (38:50)

• Brand Integrity vs. Clickbait:
  Tension exists between virality tactics and brand trust. You may need sub-brands for riskier clickbait meme content.

  "You have to resist clickbait somewhat… but if you want your content to spread, you have to adapt to what people like whether they realize it or not." — wix (40:30)

8. Philosophical Trends: Prompt Theory & Generative Culture (42:00–46:41)

• Prompt Theory:
  Explored as both a creative genre and existential meme: AI characters questioning their own prompt-based existence, and whether humans are also “prompted.”

  "Prompt theory is, what if these [AI] characters realize they're AI generated? Or, what if humans are also just prompted?" — Justine (42:00)

  Viral “clapback” trends emerge, where AI characters roast one another about being prompted.

• AI-Native Brands:
  Example: “Breadclimp”—an AI-generated possum becoming a meme, with subsequent merch (e.g., sweatshirts), illustrating how content creation leads rapidly to commerce.

  "We actually made Breadclimp sweatshirts… someone said they wanted Breadclimp sweatshirts, so we made and distributed them." — Justine (45:55)

• Platform-Driven Content Differences:
  Viral trends move in waves between platforms, exploiting arbitrage windows—what works for TikTok may later go viral on Threads, Facebook, then X.

  "There’s almost a content arbitrage thing… viral on one platform, one or two days to post it on the next." — Olivia (46:41)

9. The Professional Side: Agencies, Commercial Work, and Workflow (47:39–48:11)

• Professional AI Video Production:
  Shoutouts to creators like Pjace, who blend decades of media production experience with AI video for commercial work (e.g., brands, musicians, ad campaigns).

  “If you want to see how a pro uses AI for commercial work, check out Pjace on Twitter.” — Justine (47:39)

10. The Future: Frictionless Content-to-Commerce Automations (48:23–End)

• From Clip to Merch:
  The path from AI video to physical product is smoother than ever. “We need the Overlap for merch”—automatically generate and sell merchandise based on viral clips.

  "It feels much easier to make content and get an audience than to actually get good merch. Imagine something that just generates merch based on your latest videos." — Alessio (48:23)

Notable Quotes & Memorable Moments

• On the Satisfying Nature of Hands-On AI Creation:

  "I did that with the Minimax model and new ElevenLabs sound effects… it took a long time to line them up, but it was very satisfying when I was done." — Olivia (09:23)

• On AI Influencers Supplanting Traditional Beauty:

  "Before, to be an Instagram or YouTube influencer, most of them are hot people. Now anyone can be a popular influencer—and not have to be hot. Anyone can create a character who meets the beauty standard and then it's their brain behind it.” — Justine (17:31–18:07)

• On Brand Integrity in the Algorithm Age:

  "We have to resist clickbait … But if you want your content to spread, you have to adapt to the way people want to consume info, whether or not they know it." — wix (40:30)

• On AI’s Existential Comedy:

  "One of the big trends for VO3 on TikTok right now is AI clapbacks… The young person says, 'well your hair is just prompted,' and the old person fires back that they're prompted too." — Justine (42:42)

• On the Business Stack Friction:

  "You have to make sure you're logged into the correct Google account… just to get to VO3. Honestly, most creators just use an easier interface." — Justine (25:56)

Suggested Timestamps for Important Segments

• AI Video Meme Breakout: 02:24–04:44
• How Content Trends Originate & Spread: 05:25–10:47
• Creativity & the Remix Ecosystem: 15:27–16:52
• Monetization Paths & Economic Realities: 18:35–22:21
• Platform Layer & Business Stack: 24:16–28:25
• Automation Tools for Viral Content: 35:06–37:09
• Brand vs. Virality Tension: 39:19–40:51
• Prompt Theory & Generative Memes: 42:00–44:40
• Breadclimp/Bespoke Merch Example: 45:13–45:55

Additional Resources / Mentions

• Models & Tools:
  • VO3 (Google), Minimax, Cling, CREA, ComfyUI, Replicate
  • ElevenLabs (audio), Yapper (deepfake), Overlap (clipper automation)
• Trends:
  • “Italian Brain Rot,” “Animals Diving,” “Kim the Gorilla,” “Breadclimp”
• Accounts to Watch:
  • Pjace (professional AI video), AI Warper (advanced open-source workflows)
• Platform Insights:
  • TikTok/Instagram (mainstream, memes); Threads/Facebook (older users); Reddit (anonymity/content origin)

Summary Takeaway

AI video isn’t just a new creative tool—it’s an accelerator for Internet culture, entrepreneurship, and the creator economy. As tools evolve and friction drops, meme-makers become IP-holders, viewers become merch buyers, and anyone—from professionals to “brain rot” memers—can shape pop culture at viral speed. The infrastructure still presents usability and cost hurdles, but the landscape is already upending who and what goes viral, how content is made and sold, and the very definition of digital creativity.

For full show notes and references, see: https://latent.space