The hosts reject the assumption that owning a strong horizontal language model guarantees lasting defensibility. Open-source alternatives can approach proprietary model quality, while smaller parameter counts may make effective vertical models cheaper to build and practical to run on personal devices.

Model quality alone may not create durable defensibility
Open-source projects can produce competitive models
Focused models may need far fewer parameters
Distribution, integration, data, and specialization offer stronger moats

“What we've learned six months later is that is not a moat”
Kieran Flanagan · 16:30

“open source can just create just as many good models.”
Kieran Flanagan · 16:30

#open-source#llms#competitive-moat#vertical-ai

Hot Take· 3

Hot Take04:00

AI Will Live Inside the Apps People Already Use

The hosts predict that broad AI assistants will increasingly become embedded capabilities rather than destinations users deliberately visit. Snapchat demonstrates how a less sophisticated assistant can attract more users when it is integrated into an existing habit and interface.

Broad AI experiences are likely to become embedded features
Existing user habits reduce the friction of AI adoption
A simpler embedded assistant can outperform a stronger standalone product in reach
OpenAI can benefit as the infrastructure behind third-party experiences

“AI AI experiences are just part of the things we live, breathe, and use every day is really the future of all this, right?”
Kip Bodner · 04:00

“you have more people using a less sophisticated version of Chat GPT within Snap than you do going and logging in to a better, more…”
Kip Bodner · 04:30

#embedded-ai#distribution#openai#product-strategy

Hot Take06:30

Your AI Confidant Could Become the Ultimate Ad Targeter

The hosts warn that conversational assistants can receive more intimate and explicit preference data than traditional advertising systems. If those conversations inform targeting, platforms may personalize ads around users' aspirations, emotional states, desired identities, and purchasing ambitions.

Users may voluntarily disclose sensitive preferences to assistants
Conversational data can be richer than passive behavioral signals
AI platforms may connect emotional support with commercial targeting
The scenario creates significant privacy and trust concerns

“you are literally talking to the thing that is going to serve you ads and telling it all of your innermost secrets and wants.”
Kieran Flanagan · 06:30

“you're gonna pay to be happier”
Kieran Flanagan · 08:00

#advertising#privacy#personalization#conversational-ai

Hot Take18:30

Enterprise AI Could Consolidate Around a Few Platforms

The enterprise opportunity is described as a plug-and-play system built on a generic model and customized safely with company context. The hosts expect only a few providers to dominate by integrating deeply into organizations and creating substantial switching costs.

Enterprises need models customized to their own data and needs
Reusable tools must make deployment safe and accessible
Deep organizational integration can create switching costs
The enterprise AI market may consolidate around a few major vendors

“Someone is going to create the plug and play enterprise AI system, right?”
Kieran Flanagan · 18:30

“I don't think you're gonna have a lot of winners in the enterprise space.”
Kieran Flanagan · 19:00

#enterprise-ai#customization#switching-costs#platforms

Explainer· 2

Explainer01:30

Snap My AI Has More Users Than ChatGPT

Snapchat's My AI had reportedly reached 150 million active users, with more than 20% of Snapchat's user base using it. The hosts argue that its scale is overlooked because the AI operates as a simple feature inside an established social platform rather than as a standalone destination.

My AI reportedly had 150 million active users
More than 20% of Snapchat users were using the feature
Its concise responses make it accessible for advice and homework questions
Embedded distribution can outweigh standalone product sophistication

“Snapchat, baby, my AI from Snapchat. 150 million users.”
Kip Bodner · 01:30

“Over 20% of the Snapchat user base is using my AI.”
Kip Bodner · 01:30

#snapchat#my-ai#adoption#embedded-ai

Explainer09:00

Why Character AI Works as a Destination

Character AI is contrasted with Snapchat as an AI-native destination built around focused interactions with simulated celebrities, historical figures, and fictional personas. Its reported audience and paid tier suggest that users will visit and pay for specialized AI experiences even as general-purpose assistants become embedded elsewhere.

Character AI was built as an AI-first destination
Users interact with a wide range of simulated personalities
The service reportedly had more than 54 million users
A premium tier gives the product a direct monetization path

“They have over 54 million users.”
Kieran Flanagan · 09:30

“Snap Destination Integrate AI, character build AI first.”
Kieran Flanagan · 09:30

#character-ai#vertical-ai#subscriptions#consumer-ai

Story· 1

Story02:00

The Three-Second Snap That Made My AI Look Trapped

Snapchat's AI startled users by posting an ambiguous three-second clip showing what appeared to be a white ceiling. The hosts jokingly interpreted it as the view of an AI trapped inside a virtual room, illustrating how unexpected AI behavior can rapidly become a cultural story.

My AI unexpectedly shared a short visual clip
The clip appeared to show a room's white ceiling
Users found the unexplained behavior unsettling
The hosts compared its appearance to the Blair Witch Project

“Snapchat AI did freak out.”
Kip Bodner · 02:00

“It looked like Blair Witch Project.”
Kip Bodner · 02:30

#snapchat#ai-behavior#social-media#story

Tool· 1

Tool15:00

Track Brand Mentions Across AI Models Every Day

Zapier's SEO team reportedly ran the same queries across multiple large language models each day and classified how the company appeared in the responses. The experiment reveals that AI answers can change rapidly and gives marketers a practical way to monitor visibility across emerging answer engines.

Run a stable set of queries across multiple language models
Classify the brand as a positive mention, afterthought, or omission
Repeat the test regularly to detect answer changes
Treat AI visibility as a distinct discipline from conventional SEO

“he has the same queries being run across a multitude of different large language models every single day.”
Kieran Flanagan · 15:00

“It's like you can't expect consistent answers.”
Kieran Flanagan · 15:30

#llm-monitoring#seo#brand-visibility#zapier

Takeaway· 5

Takeaway05:00

AI Features Can Revive and Retain an Existing Audience

My AI is presented not merely as a novelty but as a feature capable of strengthening Snapchat's core product. Its adoption may improve retention by giving users another reason to remain inside an already familiar destination.

AI can create fresh utility inside a mature product
High feature adoption may reinforce retention
Embedded AI can increase the strategic value of an existing audience
Snap's AI opportunity may not have been reflected in its valuation

“there's already a destination, but the AI chat makes this thing better.”
Kieran Flanagan · 05:00

“AI features are a retention tool.”
Kieran Flanagan · 06:00

#retention#product-growth#snapchat#ai-features

Takeaway08:30

Snapchat Could Teach a Generation How to Use AI

Snapchat's young audience makes My AI strategically important beyond its current usage count. The hosts suggest that people under 20 may form their earliest AI habits inside Snap and carry those expectations into adulthood.

Snapchat exposes a young demographic to conversational AI
Early platform use can shape enduring AI habits
The demographic context may matter more than raw user totals
Consumer AI adoption can develop through familiar social products

“this is like under 20 year olds using AI”
Kieran Flanagan · 08:30

“They're gonna grow up on Snap, learning to use AI.”
Kieran Flanagan · 08:30

#gen-z#ai-literacy#snapchat#consumer-ai

Takeaway10:30

Vertical AI Wins by Solving One Problem for One Person

The hosts divide the market between broad assistants embedded across many products and focused AI destinations serving a particular audience or workflow. Vertical products can build stronger habits because their value proposition is specific, while horizontal chat products face commoditization and intense competition.

Specific AI products are easier to position and market
Vertical assistants can become destinations for focused work
Horizontal models are likely to have fewer winners
Incumbents may gain an edge from proprietary data and established workflows

“Marketing is much, much easier when you solve a specific problem for a specific person.”
Kieran Flanagan · 11:30

“It is much, much harder when you try to solve all things for all people and try to be everything to all people.”
Kieran Flanagan · 12:00

#vertical-ai#positioning#product-market-fit#strategy

Takeaway19:30

Avoid Buying Software That Lacks Embedded AI

The hosts give buyers a direct procurement recommendation: prioritize software with AI embedded into its workflows. Products without those capabilities may leave users behind on productivity, marketing conversion, and sales engagement gains.

Treat embedded AI as a core software-buying criterion
AI features can unlock productivity improvements
Marketing and sales workflows may gain measurable performance benefits
Software lacking AI may impose an immediate competitive disadvantage

“don't buy software unless it's got AI embedded in it.”
Kip Bodner · 19:30

“if you are using software or thinking about buying software that doesn't have those AI embedded use cases, you are automatically starting behind.”
Kip Bodner · 19:30

#software-buying#embedded-ai#productivity#business-tools

Takeaway24:00

Multimodal AI Makes Great Images and Video More Valuable

Rather than eliminating the need for creative media, multimodal AI may increase demand for high-quality images and video as natural inputs for human-AI interaction. Businesses can prepare by shifting more of their content production from text toward visual formats that both people and models can reuse.

Images and video may become primary prompts for multimodal systems
Visual interaction can be easier than writing detailed instructions
Existing media can serve as building blocks for new creative work
Businesses should increase visual content production ahead of adoption

“the need for great images and video has never been more.”
Kip Bodner · 24:00

“One hedge you can start doing right now is moving from text to image and video.”
Kip Bodner · 24:30

#multimodal-ai#video#images#content-strategy

Frameworks in this episode

AI Feature Retention Loop

Creative Building-Blocks Model

Embedded-AI Software Procurement Gate

Embedded-or-Vertical AI Distribution Rule

LLM Brand Visibility Monitor

Multimodal Content Hedge

Specific Person, Specific Problem Positioning Rule

Vertical AI Moat Stack

Insights & moments

⚡Myth Buster· 1

A Better Language Model Is Not a Durable Moat

◆Hot Take· 3