A failed voice interaction with Claude leads to a broader criticism of current AI products. Despite rapid capability improvements, rough edges and unreliable interfaces still make the tools harder for mainstream users to adopt.

Advanced AI products still fail in routine interactions
Voice interfaces remain inconsistent
Product reliability matters alongside model capability
Dedicated dictation tools can outperform general AI interfaces

“All these AI tools are still really, really buggy.”
07:00

“Like the amount of bugs and stuff on the edges, like we're wondering why there's not mass AI adoption. That's certainly one of the reasons.”
07:00

#ai adoption#voice ai#product reliability#claude

Explainer· 3

Explainer01:30

What the Creator Hub Is Supposed to Automate

The hosts define a lightweight tool for identifying and engaging creators who can distribute a company's products or services. The intended workflow covers company research, buyer identification, platform selection, creator discovery, partnership economics, and outreach.

Research the company and its products or services
Identify the audience and buyer
Rank the platforms where buyers consume information
Discover relevant creators with genuine distribution leverage
Support outreach and partnership proposals

“what we mean is a tool that's gonna let you identify creators to partner with, do the outreach, probably propose like the economics and rough…”
01:30

#creator marketing#automation#distribution#creator discovery

Explainer03:30

Turning a Company Domain Into a Creator Shortlist

The proposed app begins with a company domain and researches the business, its offering, and its likely buyers. It then identifies where those buyers consume information and finds creators whose topics and audiences align with them.

Use the company domain as the initial input
Infer products, services, audience, and buyers
Determine which platforms matter to those buyers
Find creators who discuss relevant topics
Display creator audience sizes by platform

“Once they have entered the domain, the tool will research that company, the product and services, and it will establish who the audience is and…”
Kieran · 03:30

“Where do those buyers go to consume information?”
Kieran · 04:00

#creator discovery#buyer research#audience research#platform selection

Explainer18:30

Deep Context Produces a Completely Different App

One host gives Perplexity Computer a book chapter containing detailed thinking about creator partnerships. That source material leads the system to generate a richer creator skill and a campaign-oriented application with scoring, contact discovery, proposals, and email drafting.

Supply domain-specific source material as context
Convert detailed written expertise into an executable skill
Use that skill as the basis for an application
Expect richer context to change product structure and features
Preserve deep thinking so AI can extend it repeatedly

“And because it had the chapter and context, this is why context is so important.”
18:30

“It actually built an entire skill around creators.”
18:30

#context engineering#knowledge reuse#creator marketing#ai apps

Story· 1

Story21:30

Perplexity Builds a Working Creator Hub in About 30 Minutes

After Claude's voice and specification workflow causes delays, Perplexity Computer produces and deploys a functioning creator-hub prototype. The app analyzes HubSpot, recommends platforms and creators, builds a shortlist, and drafts personalized outreach, although the creator rankings still contain obvious gaps.

The app was conceived during preparation for the episode
Perplexity deployed a functional web prototype
The prototype generated a buyer profile and platform recommendations
It discovered creators and drafted outreach
Missing relevant creators exposed the need for better APIs and iteration

“Creator Hub is live, baby.”
21:30

“I could probably do a couple of revs and sell this. And you made it in in 30 minutes of perplexing computer.”
23:30

#perplexity computer#creator hub#live build#mvp#ai agents

Tool· 1

Tool09:30

Why Perplexity Computer Feels Like an AI Work Hub

Perplexity Computer is praised for synchronized desktop and mobile access, cloud execution, persistent files, and understandable task visibility. The hosts see its ability to keep working after a laptop closes as a major advantage over desktop-heavy AI workflows.

Tasks synchronize across desktop and mobile
Cloud execution continues when the laptop is closed
Files remain available as reusable context
The interface exposes task files, usage, credits, and to-dos
Potential integrations could turn it into a central work hub

“it has a mobile app and a desktop app, and they sync perfectly and can access everything in the cloud”
09:30

“you can literally just shut your laptop right now and it'll just keep building.”
20:30

#perplexity computer#cloud agents#mobile ai#productivity

Takeaway· 5

Takeaway01:30

Why Kieran Starts Every AI Build in Claude

Kieran explains that he begins with Claude even when another product will execute the build. He uses the conversational model to clarify the idea and develop the instructions before handing the task to a coding or agentic tool.

Use a conversational model to shape the task first
Iterate on the idea before invoking the building tool
Separate specification work from implementation
Use cheaper conversational tokens where practical

“everything for me starts with Claude even when I'm using other tools”
Kieran · 01:30

“I weirdly use Claude Opus to kind of go back and forth with and then I prompt Claude Code.”
Kieran · 01:30

#claude#prompting#ai workflow#vibe coding

Takeaway08:00

Align With the Model Before Asking It to Execute

Kieran argues that users should discuss the desired outcome with the model before asking it to produce the final artifact. This alignment step helps expose assumptions, improve the plan, and reduce the chance that execution heads in the wrong direction.

Work backward from the desired result
Discuss the approach before generating the artifact
Ask the model to confirm a shared plan
Move to execution only after alignment

“I work back to front.”
Kieran · 08:00

“I think one of the best things you can do when working with these models is to first go back and forth until you align…”
Kieran · 08:00

#prompting#alignment#ai collaboration#planning

Takeaway08:30

Give AI a Role and Concrete Product Constraints

The discussion highlights role prompting and explicit product parameters as useful ways to focus an AI model. In this build, the prompt positions Claude as a CTO and emphasizes simplicity, strong UX, and short time to value.

Assign a role suited to the task
Describe the target user
Specify desired UX qualities
Define the required functionality
Ask for sequential implementation phases

“The other thing is, yeah, I give it a role.”
Kieran · 08:30

“I try to give it some parameters.”
Kieran · 08:30

#role prompting#product design#ux#ai prompts

Takeaway13:30

Prototype First So Everyone Can React to Something Real

The hosts describe a cultural shift among startups toward rapidly building prototypes and MVPs before committing substantial time. A working artifact makes the idea tangible and lets a team decide whether the concept deserves a full implementation.

Build a rough version before making a major commitment
Use prototypes to make abstract ideas concrete
Let stakeholders react to a working artifact
Invest in production only after validating the direction
Treat the podcast build as an MVP rather than a finished product

“The clear cultural shift that's happening with most of these companies is that they are just prototyping and MVPing stuff very quickly so that everyone…”
13:30

“And if so, then we'll go invest the time in building it.”
14:00

#prototyping#mvp#startups#product development

Takeaway23:30

The Next Iteration Needs Platform APIs and Connectors

The live prototype demonstrates that an agent can establish the basic product experience without extensive integration work. For better creator coverage and more accurate rankings, the hosts would connect platform APIs and other data sources in the next iteration.

Use the prototype to validate the workflow first
Add platform APIs after the core experience works
Improve creator coverage with dedicated connectors
Treat weak LinkedIn results as a data-access problem
Iterate rather than expecting a production system from one prompt

“We would give it access to APIs.”
23:30

“the X connector, these different connectors that allow it to pull from different platforms.”
23:30

#apis#creator data#product iteration#integrations

Frameworks in this episode

Align Before You Execute

Campaign-First Creator Partnership Workflow

Context-to-Skill-to-App Pipeline

Domain-to-Creator Discovery Pipeline

Prototype-Then-Invest Decision Gate

Role, Parameters, Functionality Prompt Stack

Target-Platform Prompt Adaptation

Insights & moments

◆Hot Take· 1

Bugs Are Still Blocking Mass AI Adoption

✶Explainer· 3