23 July 2024

I Built An App With Claude 3.5 In Less Than 10 Hours

1Frameworks

10Insights

Frameworks in this episode

Entrepreneurship5 steps

Content-to-App Validation Loop

Test an idea through content, then turn proven demand into a lightweight app

Read →

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Hot Take· 3

Hot Take15:30

AI Could Make One-to-One Subscription Software Profitable

The hosts argue that AI can collapse the cost of implementing customer-specific features. Instead of choosing only the requests favored by a majority, a software business could maintain customized variants with different subscription prices while keeping marginal development costs low.

Traditional SaaS prioritizes features wanted by the majority
AI can reduce the cost of customer-specific changes
Different customers could pay different prices for customized variants
Custom software and subscription software may converge

“you're essentially getting custom subscription apps”
Co-host · 16:00

“the marginal cost is near zero”
Kieran · 17:00

#custom software#saas#personalization#marginal cost

Hot Take19:00

Claude Could Consume the Lightweight Website Builder Market

Claude can generate and publish simple websites from natural-language instructions, directly addressing the basic use cases served by lightweight site builders. Kieran argues that adding forms and other basic web capabilities could let Claude take substantial market share from this profitable category.

Users can describe a website instead of assembling it manually
Claude generates the code and publishes the result
Personal-brand and single-use sites are strong initial use cases
Forms and basic interactions would expand the competitive threat

“I don't see how Claude can't eat up that market”
Kieran · 19:30

“they can eat up a ton of market share that is a very profitable space to live”
Kieran · 20:00

#website builders#claude#no-code#market disruption

Hot Take20:30

AI Could Produce Ten Times More Websites Within a Decade

The co-host predicts that radically cheaper technical work will lead people to build ideas they previously would have filtered out on cost grounds. This increased experimentation could produce both valuable innovations and large amounts of low-quality or harmful material.

High development costs historically forced extensive pre-build filtering
Cheap AI development encourages people to try more ideas
The number of websites could increase tenfold
Higher creation velocity will have both positive and negative consequences

“in 10 years we will have 10x the websites on the internet”
Co-host · 20:30

“now everything is just going to get built”
Co-host · 21:00

#future of web#ai coding#experimentation#prediction

Explainer· 3

Explainer00:30

Claude Apps Turns AI-Generated Code Into Shareable Web Tools

Claude's publishing capability lets users move beyond generating code or discussing an idea with an LLM. They can create a lightweight application, publish it, and immediately share a usable web experience.

Users can create lightweight apps through conversation
Publishing turns generated code into a shareable web artifact
The feature lowers the barrier between an idea and a usable product

“now you can create lightweight apps click publish and immediately have something shared on the web”
Kieran · 00:30

#claude#ai apps#publishing#web development

Explainer06:30

Why the MVP Generates a Prompt Instead of the Final Output

The demonstrated app cannot yet call an API from within Claude, so it produces a complete prompt rather than the finished LinkedIn post. Users copy that prompt into Claude manually, creating a functional but partially manual bridge until API access is available.

The app assembles the topic, post template, and writing style
It generates a detailed prompt rather than the final post
Users manually copy the prompt into Claude
A future API connection could return the output inside the app

“because I cannot it cannot call apis yet”
Kieran · 07:00

“what the user can do now is just copy this to clip board and go run it themselves in Claude”
Kieran · 08:00

#api#prompting#mvp#claude

Explainer12:00

The Economics of a Ten-Hour AI-Built App

The hosts calculate how a small application can be financially worthwhile when AI reduces initial development to roughly ten hours. One hundred customers paying $10 would produce $1,000 in the first month, equating to about $100 for each initial development hour before ongoing costs.

The estimated total build time is eight to ten hours
One hundred users at $10 produce $1,000 monthly revenue
Low initial development time changes the viability of niche products
Retention can increase the return without repeating the original build effort

“let's say 8 to 10 hours all in”
Kieran · 13:00

“right off the bat you're making $100 an hour in month one”
Co-host · 13:30

#saas#unit economics#revenue#ai development

Story· 2

Story08:00

Claude Adds an A/B Testing Feature From Three Sentences

During the demonstration, Kieran asks Claude to let users select two writing styles and generate competing prompts. Claude modifies the interface and labels the outputs A and B, showing that it understood both the requested functionality and the experiment's purpose.

The request was expressed in plain language
Claude changed the app to support two writing styles
The generated interface used A/B testing labels
Small product changes can be implemented during a conversation

“we want the user to be able to select two writing styles to create two competing prompts so they can a B test the post”
Kieran · 08:00

“I'm basically creating new functionality for an app and I wrote three sentences”
Kieran · 08:30

#ab testing#ai coding#product iteration#claude

Story18:30

A Parent Builds a Seventh-Grade Math Quiz Without Coding

A showcased Claude app gives a seventh grader a customized math quiz. The example illustrates how non-developers can create narrow educational tools for a particular child rather than searching for a general-purpose product.

The app targets one child's learning need
A non-developer can create the tool
AI enables highly specific educational software
Personal usefulness can justify an app even without a large market

“this is someone creating a matths quiz for a seventh grader”
Kieran · 18:30

“you can actually build apps to help your you know child learn whatever they they're trying to learn”
Kieran · 18:30

#education#personalization#parents#math

Takeaway· 2

Takeaway05:30

Ask AI for Strong Phrases Instead of a Finished Post

Kieran finds that Claude's full social posts still require heavy editing, but its ability to produce voice-specific phrases is useful. He asks for several talking points and selectively incorporates the strongest ones into his own writing.

Full generated posts may require extensive editing
Voice-specific phrases can be more useful than complete drafts
The human writer retains control over structure and judgment
Generating multiple phrases creates options without outsourcing the post

“I've started to ask it to give me talking points not the post”
Kieran · 05:30

“I basically ask it for like seven phrases and then I try to figure out if I can integrate them into the post”
Kieran · 06:00

#ai writing#linkedin#editing#content

Takeaway21:00

When Everyone Can Build, Ideas and Distribution Become the Advantage

As software creation becomes broadly accessible, coding ability becomes less differentiating. The hosts expect competition to shift toward choosing better ideas, reaching the market quickly, listening to customers, and developing strong marketing and sales capabilities.

AI makes more people capable of building software
Idea quality becomes more important
Speed to market remains a competitive factor
Customer listening guides useful improvements
Marketing, sales, and distribution become central advantages

“everyone is the Builder anyone can bring their ideas to life”
Kieran · 21:00

“best ideas listen to customers really good at distribution Marketing sales exactly right”
Co-host · 21:30

#distribution#marketing#sales#product strategy