Podcast Summary: "Claude Broke. Perplexity Built the App Anyway"

Marketing Against The Grain
Hosts: Kipp Bodnar (HubSpot CMO), Kieran Flanagan (SVP Marketing)
Date: March 26, 2026

Episode Overview

In this engaging, hands-on episode, Kipp and Kieran take listeners through the real-time process of building a "creator hub" AI agent using Perplexity Computer and Claude Code. Their goal: to demonstrate how marketers can use the latest AI tools to prototype innovative solutions that help brands efficiently discover, vet, and reach out to potential creator partners. The conversation is candid and practical, highlighting both the potential and the current limitations of AI-driven marketing tools.

Key Discussion Points & Insights

1. Why Build a Creator Hub?

• Main marketing trend: Brands are increasingly working with creators for distribution and partnership, especially in B2B.
• The gap: Existing tools don't fully meet marketers' needs for creator discovery, outreach, and management.
• The solution: Custom-building a lightweight AI agent to automate creator research, outreach, and campaign management.

2. Live Build Approach (00:34–07:36)

• No prep, full transparency: The hosts build the tool live, with candid explanations and comments about their process, including mistakes and learning moments.
• Tech stack:
  • Kieran: Starts with Claude for ideation and specifications, then uses Claude Code.
  • Kipp: Champions Perplexity Computer for its flexible, cross-platform use and intuitive UX.

Notable Quote

"Everything for me starts with Claude, even when I'm using other tools... I prompt Claude Code. It saves you tokens."
– Kieran Flanagan [02:06]

3. Kieran's Prompt Engineering Workflow (07:36–09:10)

• Back-and-forth alignment: Iteratively clarifies the tool’s function, requesting Claude to act as a CTO.
• Best practices called out:
  • Assigning roles to the AI model
  • Breaking out desired features step by step
  • Prompting for functional specs and technical phases

Notable Quote

"One of the best things you can do when working with these models is to first go back and forth until you align on something… go back and forth and say, can we align on how we do this?"
– Kieran Flanagan [08:06]

4. Specs, Tools, and Context: Claude vs. Perplexity (09:31–11:07)

• Spec breakdown: The AI creates a functional spec for the tool, including company research, buyer personas, platform ranking, creator identification, and basic outreach.
• Tool selection reality: Hiccups with API keys and Claude's interface prompt a shift to Perplexity Computer for speed and usability.

5. Perplexity Computer's Power Features (15:41–19:02 & 21:09–22:32)

• Rich context files: Kipp demonstrates how carrying over research (like draft book chapters) as context turbocharges the AI’s effectiveness.
• Seamless cross-device experience: Perplexity's desktop and mobile apps sync with no friction, providing true flexibility.
• Lightweight MVP: The team is able to get a functional prototype built live within the episode.

Notable Moment

"[Perplexity Computer] has a mobile app and a desktop app and they sync perfectly and can access everything in the cloud. And it just works..."
– Kipp Bodnar [09:51]

6. Iterative Development & the MVP Mindset (12:25–14:16; 23:12–24:08)

• Modern prototyping culture: The hosts stress rapid MVP cycles over lengthy memos and delayed builds.
• Real-world AI limitations: They encounter bugs and edge-case failures, reinforcing that “prompt and play” doesn’t always yield perfect results.
• Team takeaways: Frequent iteration, adding context, and customizing prompts are crucial.

Notable Quote

“You really do have to, like, work on this stuff and go back and forth. So this is really an MVP.”
– Kieran Flanagan [13:27]

7. Demoing the Creator Hub Agent (21:45–27:04)

• Step-by-step AI performance review:
  • Buyer persona generation, platform discovery (LinkedIn, newsletters, etc.)
  • Creator identification and fit scoring
  • Outreach email drafting and JSON export options
• Two different builds: Kipp and Kieran’s apps differ due to the context files provided to Perplexity—a live demonstration of the impact of context on AI output.
• Integration vision: Future plans include linking the tool with HubSpot for seamless campaign management.

Notable Moment

“We might actually get this done, which is pretty amazing... I could probably do a couple of revs and sell this. And you made it in 30 minutes.”
– Kipp Bodnar [24:03]

8. Tangents, Startups & AI Possibilities (18:00, 19:10, 24:08)

• Startup ideas: Marketplace for out-of-copyright books parsed for AI, AI-assisted content creation from Kindle books, leveraging public domain for new educational products.
• Reflection on the AI future: Frustration at the number of tools but excitement about combining features for a dream workflow.

Notable Quote

“Free startup idea for anybody who wants to start a startup... do the Kindle Marketplace, but for allowing books to be accessed by AI.”
– Kipp Bodnar [19:10]

Notable Quotes

• “We're doing this live with no prep. We're going to compare and contrast Perplexity Computer... and Claude Code.” – Kieran Flanagan [00:34]
• “Assign a role to your AI, break out the functions, and go back and forth iterating—that's really key.” – Kieran Flanagan [08:06]
• “Perplexity Computer... it is about a hundred times easier to understand than cloud code.” – Kipp Bodnar [16:43]
• “If they allowed you to host apps, I would probably never use any other tool.” – Kipp Bodnar [16:58]

Segment Timestamps

• 00:00–02:06 – Introduction; building a marketing agent for creator partnerships
• 02:06–07:36 – Prompting techniques and workflow alignment
• 07:36–09:10 – Value of iterative prompt engineering and MVP thinking
• 09:31–11:07 – Reviewing functional specs, API considerations, and tool comparisons
• 11:46–14:16 – MVP realities, culture of rapid AI prototyping in startups
• 15:41–19:02 – Kipp on leveraging Perplexity Computer’s context features
• 21:09–22:32 – Cross-device cloud functionality and practical uses
• 21:45–27:04 – Live demo/analysis of the Creator Hub agent
• 18:00, 19:10, 24:08 – Startup ideas, content AI possibilities, reflections

Conclusion: Key Takeaways

• Building with AI is process-driven: Expect iteration, experimentation, and occasional bugs—but rapid MVP cycles unlock real efficiencies.
• Prompt engineering matters: Role assignment, context, and back-and-forth alignment yield drastically better results.
• Perplexity Computer shines for marketers: Its file system, cross-device support, and ease of use make it a standout for quick AI app prototyping.
• Creator partnerships are the new growth engine: Developing tools to automate and streamline this process is the next frontier for marketing teams.

“If we can do this in under 40 minutes on a podcast, imagine what you could do with a full week.”
– Kieran Flanagan [12:50]

End of summary.