MMarketing Against The Grain
← All episodes
09 November 2023

Massive GPT4 Upgrades: ChatGPT Store, GPT Turbo & A LOT More

4Frameworks
13Insights

Frameworks in this episode

Insights & moments

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

Hot Take· 7

Hot Take01:00

Why the GPT Store Could Rival the Original ChatGPT Launch

The hosts argue that a marketplace for user-created AI assistants could unlock a major new category for entrepreneurs and consumers. They temper the comparison with OpenAI's unsuccessful plugin store, noting that the opportunity depends on the marketplace finding product-market fit.

  • Users can create, buy, and sell specialized AI assistants
  • The failed plugin store shows that distribution alone does not ensure demand
  • A successful marketplace could create a new entrepreneurial category
  • Non-developers may participate more easily than in traditional app stores

I think the GPT store is the better version of what that was potentially going to be

Kieran Flanagan · 01:30

this to me is as big of a release as the original chat gbt for consumer consumers

Kieran Flanagan · 03:30
#gpt store#marketplaces#openai#entrepreneurship
Hot Take05:00

OpenAI's “GPTs” Name Hides What the Product Actually Does

The hosts criticize OpenAI for calling its assistants “GPTs,” arguing that consumers do not naturally associate the acronym with task-completing agents. They believe the early category leader missed an opportunity to claim a simple, intuitive brand before competitors arrive.

  • Product names must communicate the consumer use case
  • “GPT” does not clearly imply an assistant that completes tasks
  • The word “apps” succeeded because users understood it immediately
  • Weak category branding leaves room for a competitor to define the market

gpts is awful for consumers

Kipp Bodnar · 05:00

people do not understand they do not equate GPT with like the AI assistant part

Kieran Flanagan · 06:30
#branding#gpts#positioning#consumer ai
Hot Take09:00

The Coming Data War Will Reshape Company Valuations

The discussion predicts escalating competition to protect, acquire, and build valuable datasets for AI products. As businesses recognize that differentiated data improves specialized assistants, existing data assets may become more strategically and financially important.

  • Companies will compete to protect and acquire useful datasets
  • Public sources such as transcripts may become less accessible
  • Distinctive data can strengthen AI product differentiation
  • Markets may revalue companies that control valuable data assets

we are going to be in a data War over the next several years

Kipp Bodnar · 09:30

the valuations of a lot of existing companies could change as data becomes a more valuable asset

Kipp Bodnar · 09:30
#data#ai assets#valuation#competition
Hot Take17:30

A Screenshot May Soon Replace Most Software Support Tickets

The hosts predict that vision-capable assistants will diagnose software problems from screenshots and explain fixes conversationally. They expect this to reduce the cost of support dramatically and automate a large share of troubleshooting work.

  • Users could submit screenshots instead of describing errors manually
  • Vision models can interpret interface states and error messages
  • Speech and chat can guide users through corrective actions
  • Routine software support may become substantially cheaper

I can just grab a screenshot of the error put it into the chat the chat can parse it and then actually speak it back…

Kieran Flanagan · 18:00

software support is going to be become a hundred times cheaper and automated most of it automated almost all completely automated

Kieran Flanagan · 18:00
#customer support#vision api#automation#software
Hot Take23:00

Software Interfaces May Collapse Into a Natural-Language Layer

The Zapier demonstration shows an assistant accessing calendars, resolving conflicts, and communicating through Slack using natural-language instructions. The hosts argue that conventional applications may remain as back-end systems while users increasingly interact through a conversational layer.

  • Assistants can invoke actions across external applications
  • Natural language lowers the learning curve for software
  • Traditional applications may persist primarily as back-end services
  • Conversational interfaces could reduce manual CRM and workflow administration

there is a natural language layer being built on top of software

Kieran Flanagan · 24:00

the learning curve goes to zero deal I just tell the assistant what I actually need to do

Kieran Flanagan · 24:30
#natural language#software ui#zapier#automation
Hot Take23:00

Software Interfaces May Collapse Into a Natural-Language Layer

The Zapier demonstration shows an assistant accessing calendars, resolving conflicts, and communicating through Slack using natural-language instructions. The hosts argue that conventional applications may remain as back-end systems while users increasingly interact through a conversational layer.

  • Assistants can invoke actions across external applications
  • Natural language lowers the learning curve for software
  • Traditional applications may persist primarily as back-end services
  • Conversational interfaces could reduce manual CRM and workflow administration

there is a natural language layer being built on top of software

Kieran Flanagan · 24:00

the learning curve goes to zero deal I just tell the assistant what I actually need to do

Kieran Flanagan · 24:30
#natural language#software ui#zapier#automation
Hot Take26:30

Natural-Language AI Makes Every Application Global by Default

The hosts argue that multilingual models can let creators build in one language while customers interact in another. This could expand cross-border commerce quickly while disrupting translation-heavy international marketing structures and increasing pressure on global payment infrastructure.

  • Models can translate interactions between creators and users
  • AI applications may reach international markets without separate interfaces
  • Cross-border payment rails could become a bottleneck
  • International marketing teams may shift away from translation-heavy work

if it's natural language you know what it is it's Global by default

Kipp Bodnar · 26:30

every kind of application that you build is global by default

Kieran Flanagan · 28:00
#globalization#translation#international marketing#commerce

Explainer· 4

Explainer10:00

AI Assistants Could Package Years of Expertise for Instant Reuse

The hosts explore how an expert could encode learned knowledge and procedures into an assistant that others can use immediately. This could make professional networks more valuable because people would exchange not only advice but task-performing tools built from their expertise.

  • Expertise can be partially encoded into instructions and reference data
  • Assistants may distribute practical knowledge faster than traditional teaching
  • Sharing assistants could raise productivity across professional networks
  • Automated knowledge may compress years of learning for specific tasks

you've done the learning you've automated the knowledge but you can now share that rapidly with people

Kipp Bodnar · 10:30

we can take some segment of work and say here you go now you can do that to the same degree that I can do…

Kipp Bodnar · 11:00
#expertise#knowledge sharing#productivity#ai assistants
Explainer14:30

GPT-4 Turbo Expands Context While Cutting Cost and Latency

The episode highlights GPT-4 Turbo's larger context window alongside improvements in speed and price. The hosts emphasize that fitting an entire book or hundreds of pages into context materially expands the kinds of analysis and applications developers can build.

  • The model supports a 128,000-token context window
  • Larger context can accommodate an entire book
  • Lower cost and faster responses expand viable product use cases
  • Infrastructure improvements may matter as much as headline features

you can now have 128 128,000 tokens so you can now upload an entire book to gp4 Turbo

Kipp Bodnar · 15:00

we should not underestimate the bigger faster cheaper part of the story

Kieran Flanagan · 15:30
#gpt-4 turbo#context window#pricing#openai
Explainer18:30

Real-Time Vision Opens Sports, Webcam, and Security Use Cases

Examples include automatically narrating a football match in a personalized style and identifying objects held in front of a webcam. The hosts extend the idea to doorbell cameras, delivery detection, and automated messages based on observed events.

  • Vision and speech can generate personalized sports commentary
  • Webcam feeds can be interpreted in real time
  • Security cameras could identify visitors and deliveries
  • Observed events can trigger automated messages or workflows

you could use this to narrate your favorite team in your case it's Liverpool in whatever style you want

Kipp Bodnar · 18:30

you can use the vision API connect it to a webcam and ask GPT in real time what's happening in the webcam

Kipp Bodnar · 20:00
#computer vision#webcam#sports#security
Explainer25:00

Why an AI-First Device May Not Need a Large Screen

The hosts connect natural-language software control to speculation about an OpenAI device. If an assistant can coordinate services without users opening individual apps, an AI-first device may require less screen-based interaction than a smartphone.

  • App switching is a consequence of visual software interfaces
  • An assistant can coordinate services through a unified conversational feed
  • AI-first hardware may reduce dependence on large screens
  • The device concept challenges the smartphone's app-centric design

the AI version of that doesn't need a big screen because it doesn't need apps because it can do everything within the feed

Kieran Flanagan · 25:30

it doesn't need to actually click in and click out of these different apps

Kieran Flanagan · 25:30
#ai devices#hardware#interfaces#smartphones

Tool· 1

Tool16:30

How Vision and Speech APIs Turn Website Screenshots Into Audio Critiques

A showcased application captures website screenshots, sends the images through GPT-4 Vision for analysis, and renders the resulting critique through text-to-speech. The example demonstrates how multimodal APIs can be composed into a complete marketing experience rather than used as isolated model features.

  • The application converts a website into screenshots
  • GPT-4 Vision interprets the visual input
  • Generated text becomes spoken output through a speech API
  • Combining modalities creates new marketing product formats

in the back end it takes screenshots of the website and is able to par those things in through the gbd4 vision

Kipp Bodnar · 17:00

I'm taking the image feeding the image in get text and then pushing that back out through the text to speech API

Kieran Flanagan · 17:30
#vision api#text to speech#websites#multimodal ai

Takeaway· 1

Takeaway28:30

AI Could Produce Smaller, More Centralized Companies

The episode predicts automation across field marketing, early sales work, and other operational roles. As the cost and time required to solve problems decline, the hosts expect smaller and more focused teams to operate businesses that previously required larger organizations.

  • Early sales and some marketing tasks may be automated
  • International and field operations may become more centralized
  • Lower production costs favor smaller teams
  • Focused companies may achieve greater output with fewer people

smaller teams more centralized yes smaller more centralized teams

Kieran Flanagan · 29:00

smaller companies smaller teams more focused lower because the cost is coming down so low to solve a problem

Kipp Bodnar · 29:00
#team design#automation#sales#marketing