MMarketing Against The Grain
← All episodes
20 November 2025

Why We’re Leaving ChatGPT for Google’s Gemini 3

1Frameworks
9Insights

Frameworks in this episode

Communication6 steps

Sports Coach Video Review

Turn a recorded episode into specific issues, fixes, and tougher follow-up questions.

Read →

Insights & moments

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

Hot Take· 1

Hot Take00:30

Why Gemini 3 Feels Like a Bigger Release Than GPT-5

The hosts argue that Gemini 3 is a larger and more useful release than GPT-5, particularly for marketing and business work. They contrast it with ChatGPT's unreliable automatic model routing and say that OpenAI has opened the door for competitors.

  • Gemini 3 is presented as the strongest model release of the year.
  • ChatGPT's automatic routing may prevent users from receiving the model they expect.
  • Claude and Gemini feel more compelling to the hosts than the current ChatGPT experience.
  • Google is benefiting from dissatisfaction with GPT-5 and GPT-5.1.

“I really do think this was bigger than GPT5.”

Kipp Bodnar · 00:30

“Jet GBT has has kind of opened the door for other competition and Google's walking right through it.”

Kipp Bodnar · 02:00
#gemini#chatgpt#ai models#model routing

Explainer· 3

Explainer03:30

Google Search Is Starting to Return Apps Instead of Blue Links

Gemini 3 lets Google's AI Mode generate interactive interfaces for individual searches. The hosts believe these lightweight applications can explain complex topics, keep users inside Google, and create new surfaces for monetization.

  • AI Mode can generate a custom interface around a search query.
  • Interactive explanations can replace static documents and lists of links.
  • Generated interfaces help Google retain the user's attention.
  • Composable interface elements could eventually contain advertising.

“It will actually return lightweight apps,”

Kieran Flanagan · 05:00

“This is the future of search.”

Kieran Flanagan · 05:00
#google search#generative ui#ai mode#search marketing
Explainer18:00

Gemini Analyzes Ad Pacing, Visuals, and Psychological Triggers

A generated creative-director prompt evaluates an Ed Sheeran Heinz advertisement across its hook, narrative structure, pacing, and psychological triggers. The hosts emphasize that observations about pacing and imagery demonstrate analysis beyond transcript text.

  • A reusable prompt can define a complete creative-director review.
  • The analysis covers hooks, narrative structure, triggers, and scoring.
  • Visual and pacing observations require more than transcript processing.
  • The model produced the analysis quickly during the recording.

“a complete prompt of hook analysis, narrative script structure, psychological triggers, scoring system, everything.”

Kipp Bodnar · 18:00

“This is how you know it's not looking at a transcript.”

Kipp Bodnar · 19:00
#ad creative#multimodal ai#creative analysis#advertising
Explainer20:30

Why a Ketchup Bottle Became the Ad's Most Compelling Frame

Gemini selects the 50-second product reveal as the strongest frame in the Heinz advertisement and explains the visual contrast. The saturated red bottle cuts through the restaurant's muted luxury palette, while cheap plastic clashes with chandeliers and white tablecloths.

  • The model identifies a precise frame within a 79-second advertisement.
  • It describes the shot and explains why it carries the narrative reveal.
  • Color contrast directs attention to the Heinz bottle.
  • Class and texture contrast reinforce the joke and product positioning.
  • Frame-level analysis could be applied across a larger collection of ads.

“the most compelling frame occurs at 50 second mark, the hero reveal of the ketchup bottle.”

Kipp Bodnar · 20:30

“The cheap plastic industrial design of the bottle sits in direct opposition to the crystal chandeliers, white tablecloth, and fancy vegetables.”

Kipp Bodnar · 21:00
#visual storytelling#advertising#color theory#frame analysis

Tool· 3

Tool07:30

A One-Prompt App Builds a Sourced Marketing Strategy

Kieran demonstrates a prototype that researches a company and turns its growth problem into a marketing plan. The generated application includes progress feedback, strategic opportunities, competitive data, source citations, prioritization, and a roadmap.

  • The prototype was created from a single prompt.
  • A progress indicator reassures users while research runs.
  • Recommendations are prioritized by impact and effort.
  • The report exposes its sources and the queries used to produce it.
  • The application proposes a marketing team and implementation roadmap.

“So what this basically does is it will custom create you a marketing strategy.”

Kieran Flanagan · 07:30

“It g me the methodology and but it's showing me all the sources.”

Kieran Flanagan · 10:00
#marketing strategy#vibe coding#lead magnets#research
Tool11:30

Map the Pain Points Behind Your First 1,000 Customers

A second prototype turns a target-customer description into an audience-intelligence report for startup founders. It develops an ICP, extracts real pain points, identifies where prospects discuss them, and surfaces possible buying signals.

  • The founder starts with a plain-language description of the desired customer.
  • The app produces an executive summary and lightweight ICP.
  • Pain points are linked to the channels where they were expressed.
  • Founders can enter those channels and test the product with prospective users.
  • Buying signals can help prioritize outreach.

“This is really for startup founders to get my first 100 to a thousand customers.”

Kieran Flanagan · 11:30

“Then it finds all of their mentioned pain points and it tells me where they're talking about those pain points.”

Kieran Flanagan · 12:30
#customer research#icp#startups#pain points#lead generation
Tool17:30

Gemini Finds the Best LinkedIn Clips Inside a Full Video

The hosts ask Gemini to locate two 30-second excerpts suitable for LinkedIn: one educational and one provocative. It returns proposed start and end points and identifies clips the hosts consider immediately usable, although it does not perform the final edit.

  • The model can analyze an hour-long video rather than only its transcript.
  • A platform and content objective make the clip request more specific.
  • Gemini returns start and end points for selected excerpts.
  • The model can distinguish educational material from a spicy take.
  • A human or editing tool must still cut the final media.

“All right. Here are two perfect 30 secondond clips tailored for LinkedIn. One is a spicy take, one is pure education.”

Kipp Bodnar · 19:30

“And so it doesn't edit the clip out for you, Kieran, but it tells you what the what it is and the start and end.”

Kipp Bodnar · 19:30
#video clips#linkedin#content repurposing#gemini

Takeaway· 2

Takeaway06:00

Personal Software Is Moving From Prediction to Production

The hosts revisit their prediction that AI coding would enable disposable personal software and real-time composable interfaces. They view Gemini-powered search as the first production-scale glimpse of software generated for a specific person and moment.

  • Cheap generation makes software viable for narrow, temporary use cases.
  • Interfaces can be assembled dynamically around an individual request.
  • Personalization will extend beyond search into broader internet experiences.
  • Gemini-powered AI Mode demonstrates the concept at production scale.

“we're going to move to this era of personal software where you can have this kind of disposable software that you use for use cases…”

Kipp Bodnar · 06:00

“it is the perfect like >> microscope into the future of the internet.”

Kipp Bodnar · 06:30
#personal software#vibe coding#personalization#future of internet
Takeaway22:00

Analyze Winning Ads, Extract the Formula, Then Generate a New One

The hosts propose combining multimodal analysis with video generation: collect top advertisements, identify each one's strongest frame, infer recurring creative patterns, and use those patterns to prompt a new ad. They suggest Gemini could create the prompt while a video model produces short clips.

  • Analyze a large set of successful advertisements.
  • Find and explain the strongest frame in each example.
  • Synthesize the repeated visual and narrative patterns.
  • Turn the resulting formula into a video-generation prompt.
  • Assemble generated eight-second clips into a new advertisement.

“Imagine having it go and find the best 100 ads of the year, find the best frame in each of those ads, and doing an…”

Kipp Bodnar · 22:00

“then have it use that formula to then build a net new ad in V3.”

Kipp Bodnar · 22:00
#ad generation#creative research#video ai#gemini#vibe coding