The episode argues that weak daily engagement is an even more serious problem for AI products than one-month retention. Leading incumbents command frequent usage, while even the best-known AI products often remain occasional tools.

Daily-to-monthly active usage reveals product habit strength
The median engagement figure cited for leading AI apps is 14%
Low engagement indicates limited intrinsic recurring value
Initial user acquisition can conceal a weak usage habit

“The big problem is not in churn, it is an engagement.”
Kieran Flanagan · 04:00

“And so there's this a whole slew of apps that are getting next to no engagement.”
Kieran Flanagan · 05:00

#engagement#ai-products#product-market-fit#usage

Hot Take· 4

Hot Take06:00

Everyone May Build a Personal Team of AI Coaches

The hosts envision people creating personalized AI advisers modeled on themselves or trusted experts. Products such as Whoop's conversational interface demonstrate how private data can support specialized wellness, cooking, fashion, or performance coaches.

AI companions can embody trusted experts or improved versions of the user
Personal data makes coaching responses more relevant
Specialized coaches could cover wellness, cooking, fashion, and performance
Multimodal responses may extend coaching beyond text

“Everyone's gonna have a version of themselves, everyone's gonna have a version of someone that they feel they can learn from.”
Kieran Flanagan · 06:00

“So you have a wellness coach, right? Like you actually have a wellness coach.”
Kieran Flanagan · 06:30

#ai-coaches#personalization#wearables#companions

Hot Take09:30

Multimodal AI Is Creating Copy-and-Paste Software

A demonstration using GPT-4 vision shows an AI recreating a SaaS application from a screenshot. The generated result reportedly reproduced much of the interface, data, and functionality before any manual code editing, pointing toward radically cheaper software imitation and prototyping.

A screenshot can serve as a software specification
AI can describe an interface and generate implementation code
The cited prototype reproduced functionality and financial dashboard data
Fast replication could commoditize many point-solution products

“He went from showing the AI the screenshot of the SAS app to just copy and paste in the code it gave him and then…”
Kieran Flanagan · 10:30

“We are getting to the point where you can basically just tell the app to replicate software that you give it, and it will just…”
Kieran Flanagan · 11:00

#software#multimodal-ai#code-generation#commoditization

Hot Take11:00

When Software Is Easy to Copy, Audience Becomes the Moat

The hosts predict that creators with established distribution will increasingly launch software products. If building software becomes cheaper and products become easier to imitate, trusted access to an audience may matter more than the product itself.

Lower development barriers let creators launch B2B software
Existing audiences provide immediate distribution
Product commoditization increases the relative value of trust and reach
Small teams may build highly profitable software businesses

“Product gets commoditized, audience value goes up tenfold, if not a hundredfold.”
Kieran Flanagan · 11:30

“what you're about to see in this era of copy and paste software is anybody with an audience is gonna have a software product.”
Kit Bodner · 13:00

#audience#creators#software#distribution

Hot Take16:30

Designers May Move From Figma Directly to Working Software

The hosts expect multimodal code generation to give designers a larger role in building software. A designer could define the experience in Figma, have AI produce the implementation, and then use light editing to prepare it for release.

Figma designs can become inputs for code generation
AI narrows the gap between interface design and implementation
Point solutions are more immediately accessible than complex enterprise systems
Software teams may become more designer-led

“there are great examples out there of people going from Figma files to software, right?”
Kieran Flanagan · 16:30

“And so I do think you're gonna see a lot of companies as well go much more designer-led in how software gets built.”
Kieran Flanagan · 17:30

#designers#figma#code-generation#product-development

Explainer· 3

Explainer02:30

Why First-Generation AI Apps Struggle With Retention

The hosts compare one-month retention for established consumer apps with leading AI-first products. Even prominent AI apps trail incumbents, suggesting that early adoption and publicity have not consistently translated into durable habits.

Retention determines whether growth can compound
Established consumer apps show stronger one-month retention
AI-first apps have struggled to become recurring habits
High churn makes customer acquisition progressively less effective

“If you want to build a remarkable company, you need to retain your customers.”
Kieran Flanagan · 02:30

“The media here is about 42%. So the median on the kind of non-AI companies in terms of usage, one month retention and the first…”
Kieran Flanagan · 03:30

#ai#retention#churn#growth

Explainer08:00

Why ChatGPT Added Suggested Prompts

Suggested prompts are interpreted as product-led growth infrastructure rather than simple interface decoration. They reduce the effort required to discover useful tasks, accelerate onboarding, and encourage users to return more frequently.

Suggested prompts provide immediate examples of product value
Templates can support acquisition and onboarding
Specific use cases reduce blank-page friction
Prompt suggestions aim to increase recurring usage

“You see a ton of suggested things to do because they are trying to make it more specific to you.”
Kit Bodner · 08:00

“templates are the number one way that product-led growth companies try to acquire users, onboard users, and increase usage of that product.”
Kit Bodner · 08:00

#chatgpt#templates#product-led-growth#onboarding

Explainer13:00

Creator-Led Software Could Reshape Margins and Venture Capital

An explosion of easily created software could increase competition and compress traditional SaaS margins. At the same time, creators may prefer small, profitable businesses that require less outside capital, challenging conventional venture-capital economics.

More software supply could push gross margins downward
Creators can combine distribution with inexpensive product development
Small teams may generate substantial revenue without becoming conglomerates
Capital-efficient companies may have less need for venture funding

“Well, we're gonna have way more software than ever before. We're gonna be overwhelmed with software.”
Kit Bodner · 13:00

“It's going to be a very different industry when you have people who want to build a small company, a profitable company, and probably do…”
Kieran Flanagan · 14:30

#saas#venture-capital#creators#margins

Story· 2

Story04:30

Why Character.ai Stands Out Among Early AI Products

Character.ai is presented as a rare early AI product with unusually strong engagement. Its users create and converse with virtual companions for friendship, therapy-like conversations, and other forms of personalized interaction.

Character.ai leads the cited AI engagement comparison
The average user reportedly spends about two hours in the app
Companionship creates a recurring reason to return
Users apply the product to friendship and therapy-like interactions

“And the average user spends about two hours within this app.”
Kieran Flanagan · 05:00

“People use them for friends, people use them actually. I've seen on uh Twitter slash X for therapy.”
Kieran Flanagan · 05:00

#character-ai#companionship#engagement#chatbots

Story15:00

A Notebook Photo Became a Custom Web App

A developer photographed a physical notebook layout and asked GPT-4 vision to turn it into code. The example suggests that people and businesses could digitize familiar paper workflows without abandoning the formats that already work for them.

Physical layouts can be interpreted as digital specifications
Users may preserve familiar workflows while gaining software functionality
AI lowers the cost and expertise required for digitization
Paper-heavy businesses could adopt custom digital tools more easily

“And it instantly turned that notebook into code to make a digital version.”
Kit Bodner · 15:00

“Think about anything physical can immediately for an extremely low cost become digital.”
Kit Bodner · 15:00

#digitization#gpt-4-vision#paper-workflows#custom-apps

Takeaway· 2

Takeaway07:00

Customization Could Turn AI Tools Into Daily Habits

The hosts attribute weak AI retention to products that are insufficiently adapted to each user or business. They expect multimodal interfaces, customized models, better infrastructure, and cheaper compute to make the next generation more useful and habitual.

Generic AI tools often lack recurring individual value
Customization can align a model with a specific user or business need
Multimodal interfaces expand how users interact with AI
Cheaper infrastructure should make personalized models easier to scale

“Those failures are because AI apps have not found a way to create something that is valuable to users on a daily basis, weekly basis,…”
Kieran Flanagan · 07:00

“For the most part, that is because those models and those tools are not customized for that individual or business's need.”
Kieran Flanagan · 07:30

#customization#retention#multimodal#infrastructure

Takeaway22:30

Multimodal Interfaces Could Bring AI Its Next Wave of Users

The episode identifies multimodal interaction as a major transition from AI hype to practical adoption. Images, text, and eventually video make complex capabilities easier to access, while browsing and improved model reliability expand the number of useful real-world tasks.

Multimodal input reduces the effort required to use AI
Image-based interaction makes AI accessible to less technical users
Internet browsing improves access to current information
Real feature delivery marks a shift beyond the initial hype cycle

“multimodal really reduces the barrier to use and continue to use AI technology and will bring the next wave of adopters on board.”
Kit Bodner · 23:00

“It just makes it much more usable being able to use it through.”
Kieran Flanagan · 23:00

#multimodal#adoption#chatgpt#ai-products

Frameworks in this episode

Ambition-Action Leadership Matrix

Audience-First Software Strategy

Copy-and-Paste Software Workflow

Retention-First Growth Test

Template-Led Product Growth Loop

Insights & moments

⚡Myth Buster· 1

AI's Bigger Problem Is Engagement, Not Just Churn

◆Hot Take· 4