Craft's Sacks on AI Innovation, New AI Platform Glue

Summary

David, a venture capitalist and former head of product at PayPal, discusses the creation of Glue, a new enterprise messaging platform aimed at integrating AI into workplace communication. He highlights the platform's use of the latest version of Chat CBT, emphasizing its impressive performance and how it allows for a unified experience for both AI and human conversations. David also touches on the evolution of communication tools every decade and the importance of embracing new paradigms. He shares his experience using Glue internally at Kraft Ventures, where AI has been utilized to write investment memos and analyze team contributions. David expresses confidence in Glue's approach to AI integration and its potential to improve with advancements in underlying models. He also briefly mentions his political involvement and the importance of discussing current events, including the upcoming presidential election.

Takeaways

• 🚀 **Agnostic Approach**: The company aims to work with all major role models, starting with Claude, to support both A GB t four and Claude three.
• 📈 **Latest Technology**: They are using the latest version of chat CBT, Chargeability for Omni, which was launched recently and offers terrific performance.
• 🔄 **Unified Communication**: The goal with Glue is to create a native chat app that consolidates AI and human chats in one place, addressing the issue of channel fatigue from platforms like Slack.
• 🤖 **AI Integration**: AI is incorporated as a full participant in the chat, using live chat history as context to provide answers that regular chats could not.
• 💼 **Market Gap**: David identified a need for a new communication tool that reacts quickly to the disruption in the market, leading to the creation of Glue.
• 📊 **Productivity Boost**: Glue has shown significant internal success at Kraft Ventures, with AI writing investment memos and summarizing arguments based on chat history.
• 🌊 **Riding the Wave**: The strategy is to harness the improvements in underlying model innovation, ensuring that as the models improve, the features of Glue also get better.
• 🤝 **Partnerships and Integrations**: Glue plans to integrate with all major labs, allowing users to choose which model to use, and potentially making the choice for them based on the query.
• 📉 **Investment Insights**: David's background includes being the head of product at PayPal and creating Yammer, which informs his perspective on the evolution of enterprise communication tools.
• 🏆 **Competitive Edge**: There is a debate about where AI will live in the enterprise, and Glue's view is that it should be integrated into the chat for a seamless experience.
• 🗣️ **Political Engagement**: David may get more involved in the political process during the election year, possibly hosting events and contributing, though it's not his main focus.

Q & A

What is the primary goal of the Glue platform?

The primary goal of Glue is to create the first AI-native chat app that consolidates chats with AI and humans into one place, aiming to avoid the need for separate apps for these interactions.

Why did the team choose to start with Claude as their major role model?

The team chose to start with Claude because they wanted to work with major role models and Claude was their starting point, although they aim to be agnostic and work with all major models.

What is the issue with chatbots that the Glue platform aims to solve?

Glue aims to solve the issue of chatbots not having a multiplayer experience, as they typically do not allow co-workers to be included in the chat.

How does Glue address the problem of channel fatigue in platforms like Slack?

Glue addresses channel fatigue by implementing a new threaded conversation model that does not rely on channels, providing a more streamlined communication experience.

What is David's background prior to becoming an investor and podcaster?

David was the head of product at PayPal and then created Yammer in 2008, which was one of the first enterprise messaging platforms.

What is the new conversation paradigm that David refers to?

The new conversation paradigm that David refers to is based on chat GPT, which represents a shift from the previous paradigms of feeds and channels.

Why did David feel the need to build Glue instead of investing in an existing company?

David felt the need to build Glue because he didn't see any existing tools in the marketplace that were reacting fast enough to the disruption brought about by the new conversation paradigm based on chat GPT.

How does Glue incorporate AI into the chat experience?

Glue incorporates AI as a full-fledged participant in the chat, allowing it to use live chat history as context and provide answers that regular chats could not.

What kind of results has Glue shown internally at Kraft Ventures?

At Kraft Ventures, Glue has been used to write investment memos, summarize arguments for and against investments, and identify top contributors to deal flow, demonstrating its ability to provide valuable insights based on chat history.

How does David view the role of partnerships and M&A in the tech industry, especially with the rise of generative AI?

David sees innovation happening at every level of the tech stack and believes that being positioned correctly in the application layer is crucial. He suggests that instead of competing against the models, one should harness them, allowing the application to improve as the underlying models do.

What is David's stance on the current political climate, and will he be more active during the election year?

David has been vocal about his criticism of Biden and has indicated that he might get more involved in the election cycle by hosting events and possibly contributing in other ways, although he clarifies that politics is not his main focus.

How does Glue plan to integrate with large language models like Claude and others?

Glue plans to integrate with all major labs, allowing users the choice of which model to use. The platform may even make recommendations on which model to use based on the query, aiming to provide the best experience regardless of the underlying model.

Outlines

00:00

🚀 Introduction to Glu: Bridging AI and Human Chats

David discusses the creation of Glu, an app designed to integrate AI chatbots with human conversations within a single platform. The app aims to overcome the limitations of current chatbot technology, which lacks a multiplayer experience, and the channel fatigue associated with platforms like Slack. Glu is built on the latest version of chat CBT (Chatbot Technology), which was recently launched by Omni and praised for its performance. The goal is to provide a unified communication tool for enterprises that can evolve with AI advancements.

05:01

💼 The Market Gap and Founding Glu

David, a venture capitalist with a background in Internet product development, identifies a need for a new communication tool in the market that leverages the latest in chat GPT technology. He sees a gap where existing tools are not adapting quickly enough to the changes brought by AI. This prompts him to co-found Glu with Evan Owen. The discussion highlights the importance of positioning Glu within the application layer to harness improvements in underlying AI models and enhance the user experience.

Keywords

💡Agnostic

In the context of the video, 'agnostic' refers to the intention of the technology to be neutral and compatible with various systems or platforms. The speaker mentions wanting their technology to work with 'all the major role models,' indicating a design that is not limited to a single system or standard, allowing for broader application and user base.

💡Chatbot

A chatbot, short for 'chat robot,' is an AI-powered computer program designed to simulate conversation with human users. In the video, chatbots are central to the discussion as the technology aims to support both AI and human interactions within a single platform, addressing the limitations of current chatbot experiences by including a multiplayer aspect.

💡Channels

Channels, as mentioned in the script, refer to the organizational structure used in platforms like Slack for conversations. The speaker criticizes the 'channel fatigue' experienced by teams, suggesting that the new tool aims to offer a more streamlined and less cluttered communication method that doesn't rely heavily on channel categorization.

💡Threaded Conversation

Threaded conversation is a messaging format that allows multiple messages to be organized into a hierarchical structure, making it easier to follow discussions. The video discusses the desire for a new model of conversation that is threaded, without the complexity of channels, to improve team communication.

💡Enterprise Messaging Platforms

Enterprise messaging platforms are communication tools designed for business use, facilitating internal and sometimes external communication within an organization. The speaker's background includes creating Yammer, one of the first enterprise messaging platforms, which sets the context for their expertise and the evolution of such platforms discussed in the video.

💡Chat GPT

Chat GPT, or chat-based generative pre-trained transformers, refers to AI models that can generate human-like text based on given prompts. The video mentions a new conversation paradigm based on chat GPT, indicating the integration of advanced AI into communication tools to enhance functionality and user experience.

💡Investment Memos

Investment memos are documents that outline the rationale for a potential investment, including arguments for and against it. In the video, the speaker shares their experience with using AI to write investment memos, demonstrating the practical application of AI in business decision-making processes.

💡Large Language Models

Large language models are AI systems that have been trained on vast amounts of text data, enabling them to understand and generate human-like language. The video discusses the impressive capabilities of these models in providing detailed and contextually relevant answers when given access to chat history.

💡Enterprise SaaS

Enterprise SaaS (Software as a Service) refers to software applications designed for use by businesses, delivered over the internet on a subscription basis. The video talks about how every Enterprise SaaS app is looking to incorporate AI into their product roadmap, highlighting the trend of integrating AI into business applications.

💡Stack

In the context of the video, 'stack' refers to the layers of technology, from foundational models to application layers. The speaker discusses innovation happening at every level of the stack, emphasizing the importance of positioning a product to leverage advancements across these layers for continuous improvement.

💡Commoditized Version

A commoditized version of a product or service is one that has become a standard offering with little differentiation between providers. The video mentions the potential for a commoditized version of large language models, suggesting that the technology could become widely available and accessible to various applications.

Highlights

The development of Glue, an AI native chat app that integrates with both humans and AI, aiming to streamline communication within a team.

Glue's use of the latest version of chat CBT (Chatbot Chargeability) for Omni, launched the day before the discussion, noted for its impressive performance.

The identification of a gap in the market for a unified communication tool that supports both AI and human interactions within a single platform.

The limitation of current chat CBT systems in not providing a multiplayer experience for co-workers.

The issue of 'channel fatigue' in platforms like Slack and the desire for a new threaded conversation model without channels.

David's background as an Internet product head at PayPal and the creator of Yammer, one of the first enterprise messaging platforms.

The evolution of enterprise communication tools with each decade introducing a new paradigm, from social networking feeds to chat GPT.

The decision to create Glue due to the lack of a tool in the market that was adapting quickly enough to the new communication paradigm.

The importance of incorporating AI as a full participant in chats, allowing it to use live chat history as context for more accurate responses.

Remarkable results from using Glue internally at Kraft Ventures, including AI-generated investment memos and analysis.

David's perspective on the strategic positioning of applications in the enterprise SaaS market to harness, rather than compete against, AI models.

The potential for Glue to improve with each new release of underlying AI models, such as Chachi Beaty.

The hints Sam Altman provided during a previous discussion that became clearer with the release of chat crypto by Open AI.

Glu's intention to integrate with all major AI labs and provide users with a choice of models, potentially making model selection for them based on the query.

David's experience with selling Yammer to Microsoft and his understanding of M&A dynamics in the tech industry.

The strategy of riding the wave of underlying model innovation to continuously improve the features of Glue.

David's potential increased involvement in the political process during an election year, possibly through hosting events and contributions.

Transcripts

David, if it's actually just going to start with the technology itself in

igloos built to support both A GB t four and Claude three.

And I just wanted wondered if you talk us through the approach you took in

building it. Right.

Well, we want to be agnostic. We want to work with really all the

major role models and we're starting with chapter, but we're starting with

Claude. And by the way, we're using the latest

version of chat CBT, Chargeability for Omni that they just launched yesterday

and it is pretty awesome. The performance is terrific.

But stepping back a second, what we're trying to do with with glue is create

the first Aid native chat app. And what it does is it creates one place

where you can have your chats with A.I. and your chats with humans, with your

team, because we think people we think employees want to do that in one place

instead of going to two separate apps. The problem with chat CBT, as awesome as

it is, is that it's not a multiplayer experience.

There's no way to bring your co-workers into the the the chat with you.

And the problem with Slack is really that the channels get in the way and we

hear from lots of teams that they have channel fatigue and really want this new

threaded conversation model without all the channels.

So that's what we've done with Glip. David, we know you is a venture

capitalist, a long time investor in technology companies.

What was the needs that you identified or the gap in the marketplace that that

you wanted to to start and build a company yourself and the tool rather

than back, I guess, or invest in another player?

Well, you know, before becoming an investor and before becoming a podcaster

actually was an Internet product. I was the head of product at PayPal, and

then I created a company called Yammer back in 2008, which was the first or one

of the first enterprise messaging platforms.

And back then it was based on the idea of feeds, which, thanks to social

networking, were kind of the conversation paradigm.

Then, you know, the next decade we had Slack, which was based on IAC and

channels, and now we have a new conversation paradigm that's based on is

based on chat GPT. I think every decade you get a new

paradigm like this and you need a new communication tool.

And that's what we've built with Glu. I just didn't see that tool in the

marketplace that was reacting fast enough to this disruption.

And so I partnered with Evan Owen, who's the my co-founder and CEO, in order to

create in order to create glue. Now, David, I know that you like to

start your pitches with a demo, so let's demo it and a little bit I want to look

at how it looks, it feels and how it's going to interact.

And as the founder, as well as putting the money into this, you're not worried

about, well, the competition that's out there because everyone's getting into

enterprise generative I. Yeah, for sure.

I mean, right now every Enterprise SAS app is trying to figure out how to

incorporate into their product roadmap. There's no question about that.

But I think there's still a pretty big debate about where A.I.

is going to live in the enterprise. Where are you going to go to ask those

sort of broad based questions about, for example, who in your enterprise has the

right expertise, who should be talking to those sort of core enterprise

collaboration use cases? And our view on it is that I should live

in the chat because again, you don't want to have one application for your

air chats and one application for human chats.

It all belongs together and the A.I.. What we found is that when you do

incorporate the A.I. as a full fledged participant into the

chat, it's able to use all of the live chat history as context.

And so your your your work chat app is able to give you answers that regular

chats but could not. We've had really staggering results at

Kraft Ventures using this product internally.

I've asked the A.I. to write investment memos for me,

summarize the arguments for and against investing.

The company has done an unbelievable job just assembling those arguments based on

the chat history. I've asked it who is contributing the

most to deal flow at Kraft and the is able to tell me.

It's really amazing the answers you can get when you give A.I.

access to your chat history. Amazing answers built by in many ways

amazing large language models that underpin it.

I'm interested in you having sold Yammer to Microsoft previously, having

understood the way that M&A works in the company that looks for exits is now

seeing all these interesting partnerships, shall we call them, being

done, whether it's aqua highs, whether it is opening, it's teaming with

Microsoft Anthropic, of course, going in with Amazon and Google at the same time.

How do you judge that as a VC? Well, there's innovation happening at

every level of the stack. I mean, you're seeing tremendous

innovation at the at the foundation model layer.

You've got open air, you've got anthropic, like you've mentioned, you've

got A.I. now launching that we just invested in.

Then you've got innovation happening at the application layer of the stack where

applications are able to take advantage of all the new capabilities created by

these stellar labs. And if you're positioned correctly in

the application layer, you don't want to compete against the models.

You want to harness them so that as the models get better and better, your

application gets better and better. And that's what we've seen with Glu is

that with each new release of Chachi Beaty or these other models that we've

incorporated, like Lord the The Glue chat tool just gets better and better.

So we're going to ride that wave. As the underlying model innovation

continues, the models get better and better.

The features of our application will get better and better, and that's how you

want to position yourself. Sorry to interrupt you, David.

You had Sam Altman on the all in part last week, and I think on social media

there was some debate which he participated in about whether he said

anything new that go to yesterday's presentation from Open AI.

How much of it then, surprised you based on the conversation you'd had last week?

Well, in retrospect, he gave us a bunch of hints last week, so I was.

I shouldn't have. I was being a little bit uncharitable,

but really, I wasn't criticizing Sam. I was just trying to ratio Jason.

But. But yeah, no, Sam gave us a bunch of

hints last week that now make a lot more sense in the light of chat crypto

yesterday, which really was an impressive demo.

And like I said, we've already incorporated it into glue.

And I can tell you the performance of it is way better than GP T for Turbo.

So kudos to the Open Air team. They've really done a great job with

that. David a quick question.

Just thinking about Croft's portfolio, did you try to get into the X air

around? Yeah, we did.

We. We've participated in that.

I'm from Ifollow. Will you integrate into Glu?

And how are you thinking about whether winner takes all or whether there will

be? Of course.

Well, a commoditized version of large language models that.

Yeah. I mean, I suspect that will integrate

with all the major labs. Our our view on it is that we're all

agnostic. We want to give our users the choice of

which allow them to use. And in fact, we're going to plug into

all the major labs and then we can actually make the choice for them in

terms of which model to use based on the query that they're trying to ask.

So we actually think that the more models, the better.

Do you think I could have got a large evaluation?

You know, I'm going to let them in, announced their own round.

I don't want to announce the round for them.

But, but if the question is whether we participated, we did.

David look, it's an election year, presidential election year.

You tell us about your background in technology and being a tech investor and

you've been vocal on the political front.

Will you be more active this year in this election cycle?

Well, you know, we talk about politics is one of the the topics on the all in

part. You know, we discuss current events,

business markets, politics, foreign policy events all over the world.

And so obviously, the election is unavoidable.

And it's no secret that I've been a critic of Biden.

So. So I may get more involved in the sense

of maybe hosting events, maybe contributing.

But but it's really it's it's not it's certainly not the main thing I do.

Let's just put it that way. It's the way that you get involved in

the political process is that you typically host events and you get to

maybe meet the candidates that way. And that's something we've tried to do.

Me and my co-host in the All in Pod, we've done that now for Robert F Kennedy

Jr. We did that for Vivek Ramaswamy.

We had him on the part and then we did events for them and we've extended an

offer to the the other major candidates and hope that they accept.