Related articles:

• HY4 Preview Is Here. Plus, WorkBuddy Gets 3 Updates (2 of 4) (Tech Strategy)
• More High-Tech Flex by Huawei R&D (4 of 4) (Tech Strategy)
• 6 Big Events in AI Agentic Ecommerce (Tech Strategy)

From the Concept Library, concepts for this article are:

• GenAI: LLM
• AI Agents

From the Company Library, companies for this article are:

• WorkBuddy
• Hy4
• Tencent: Marvis, Ardot, Miora, Ima, Toast

———transcribed

00:06

Welcome, welcome everybody. My name is Jeff Towson and this is the Tech Strategy Podcast from Techmoat Consulting.  And the topic for today, Tencent WorkBuddy just went international, which means it’s available pretty much everywhere. And it’s great.  I mean, it’s really good.  And there are several other buddies, this is what they call them, that are…

 

00:30

I think going international. So anyways, I’m paying a lot of attention to this. buddies kind of become my go-to thing now. There’s a couple others I go to, but let’s say it’s in the top three I’m using.  Impressive and available everywhere and way cheaper than the stuff in the West. I’ll talk about that.  Okay, that’ll be the topic for today. The WorkBuddy, code buddy, the other buddies.  Pretty cool. That will be the topic. Let me do my standard disclaimer here.

 

00:56

Nothing in this podcast or my writing or website is investment advice and numbers of information for me and any guess may be incorrect. The views and expressed may no longer be relevant or accurate. Overall, investing is risky. This is not investment, legal or tax advice. Do your own research.  Another standard disclaimer, I have had a consulting relationship with Tencent within the past 12 months. that’s, you full disclosure that’s out there, but this content is all me.  OK, with that, let’s get into the topic.

 

01:24

Now the big concept for today is obviously agents, really AI and agents. Those things kind of blur together, but man, agents are just a huge deal. It’s becoming the number one, two, and three thing I’m focused on. I’m incorporating it everywhere. It’s amazing, but it’s not simple.  AI is powerful, but it’s kind of relatively easy to integrate into a business and scale up. Give everyone AI, they chat.  Agents are 10 times more powerful than AI.

 

01:54

question and answer, that sort of thing, generate an image for me. But it’s much harder to deploy into an enterprise and it’s tricky to scale up because as most people have discovered,  the cost can get away from you very quickly.  So deploying agents into businesses, particularly e-commerce, that’s kind of my number one priority now. And if you can do that successfully, you can scale it up and it’s like superpowers. It’s really amazing. So that’s kind of the general topic for today.

 

02:25

But the context for this is, OK, AI is moving fast. Agents, that’s kind of the next phase of this. Robots will come later,  faster than I thought they were going to come. But the choices for agents have, you know, it’s put people in a bit of a bind.  Like, do we use Claude or OpenAI? We use their agents. They’re very powerful, frontier-level business models. But they get very expensive.

 

02:51

especially when you start to deploy them at scale. Agents can use up a ton of tokens very quickly.  And also there’s this question of,  am I giving away my alpha to a company that is then going to turn around and offer that as a service? Which we have seen happen.  Like a bunch of lawyers turn around and say, we’re all going to use Claude or something like that for our main agents. Basically, we’re going to integrate that into our workflows of our business.

 

03:19

And guess what? That service is going to get really smart at the law and then are they going to turn around and offer law as a service? So you basically trained your competitor by integrating this vendor into all the soft stuff that makes your business special. That’s a real concern. So that’s one option, which is problematic. The other option has been, okay, let’s go open source,  downloadable, we’ll run it locally. DeepSeek.

 

03:49

Right, and most of those models have been Chinese, but now Nvidia is making a major move there, which is really compelling actually. That’s great. That’s a good option too, but it’s a lot of work. You got to run up the servers.  That’s a step and I think a lot of people will have that.  Most people seem to be going for sort of this hybrid approach. We’re going to use frontier models,  subscription cloud-based for some things, and then we’re going to use on-premise.

 

04:18

open source for the majority of the workflows because there’s a lot of volume there. Okay, that’s been kind of the choice.  Well, now there’s a bit of a third option, which is Alibaba and Tencent. These are frontier level models. Qwen from Alibaba, Hunyuan for Tencent. They’re offering sort of the same idea of, you can get this on the cloud. You can get this by subscription,  but it’s way cheaper than the Western models.

 

04:47

They’re going and that’s already been available in China for some time, but it’s going international now So maybe I can sort of have three options now. Let’s use these models as a frontier level subscription way cheaper and I mean this is kind of where you got to get into a judgment of what are the intentions of this company Alibaba in particular has a 20 plus year history of never competing with their partners

 

05:17

Amazon in e-commerce.  Amazon,  no. They have merchants who work on the marketplace and then Amazon has its own products including private label. So they’re your vendor, your marketplace platform and your competitor.  Alibaba has always had a strict line. They are a pure marketplace. They never compete with their merchant and brand partners. Okay, so if they’re offering Alibaba Cloud,  Qwen and now Qwen Work, which is

 

05:44

The agent version, is going international, literally last week, the beta rolled out. I’m very comfortable trusting them in my workflows. I think there’s a long history of them never competing with their customers or their partners. I’m pretty comfortable with that actually.  Tencent, would pretty much say the same thing. Do I trust Claude and OpenAI? Not in the slightest. Absolutely not.

 

06:13

And if I watch videos by Sam Altman or Dario, at the end of the video interview, I feel less comfortable than when I started.  You know, my standard joke has been that Sam Altman makes Mark Zuckerberg seem trustworthy. Well, I’m amending that joke.  Dario makes Sam Altman seem trustworthy.  I don’t trust them in the slightest.  So I am not integrating them into my business full stop.

 

06:40

I’m really quite comfortable with Tencent. I’m really quite comfortable with Alibaba. Now there’s other players out there, Baidu. There’s a lot of players within China, but these are the two that are going international right now. So that’s an interesting option. So based on that, yeah, I’ve got WorkBuddy running on a laptop and I’ve got, you know, the sort of the Western models running on another one and it’s pretty great. It’s great.  So anyways, that’s what I want to talk about today. We’ll talk about WorkBuddy. I’ll talk about Alibaba.

 

07:10

which is Qwen Work. They keep changing the name, but now it’s called Qwen Work. That’s their comparable.  It just went international literally last week in beta.  WorkBuddy has been for several weeks now. You can go get an international version of it. You can download it.  Prices are super cheap because they’re going for international expansion right now. And yeah, it works great. You log in with GitHub or Gmail, or you can log in with a Twitter account.

 

07:36

That’ll be how you set up your account. you’re using the domestic version in China, you’re going to log in with WeChat or something like that. But internationally, it’s one of those three. It’s real easy to do, easy to download, easy to use.  take a look. Okay, so let me go into sort of the basics of WorkBuddy and why I think this whole buddy system they’re rolling out is actually pretty cool. Now, before I get to that, there was some actually related news, which is HY4.  Hunyuan, people have trouble…

 

08:06

pronouncing Hun-Yuen, so now they’re saying HY3. Well, HY4 preview just came out recently.  That’s their upgraded version.  HY3 was kind of important.  was sort of end of last year Tencent basically revamped their whole approach to models. They changed the people, they changed the processes, they changed what the goals are.

 

08:32

And the goal really shifted from frontier level performance metrics to real world usability. They’re focused on creating things that are super useful in the world and super easy to deploy. Now that’s actually, I like that strategy. And the first version that rolled out was HY3, which came out a couple months ago, did pretty well. And now they’ve got HY4.

 

09:01

their preview right now and they’re putting that into everything. It’s going into all their use cases, all their products. So it’s getting useful in the real world very quickly. So anyways, HY4 preview just came out. It’s a pretty big performance upgrade. This is their flagship AI model. They have a ton of products, right? That’s 10 cent.  HY4, HY, this is their flagship AI model.  Their flagship agentic product is WorkBuddy.

 

09:31

close second but work buddies arguably number one and sure enough HY4 is rolling into WorkBuddy right now anyways the preview model was released on August 28th so a week or so ago the basics yeah it’s a mixture of experts architecture not a big surprise so is most everything coming out of China 770 billion parameters 49 billion activated per token that’s really double

 

09:57

from HY3, which was about 295 billion parameters, 21 billion activated. So they doubled.  The context window is up to a million tokens. Pretty good. That’s four times up from what HY3 was, which is like 256K in terms of context window capacity.  The pricing, mean, pricing is always…

 

10:24

These are never these are never really apples to apples comparison when you hear these numbers. So the number you’ll hear for API pricing is about 90 cents per million input tokens,  about three times that for million output tokens. Again, keep in mind how you put things in and how you take them out can change radically depending on what you’re doing to those numbers. But generally speaking, it’s way cheaper from stuff out of the West. Not a big surprise.

 

10:54

Yeah, anyway, so it’s open-sourced, it’s going into WorkBuddy, it’s going into CodeBuddy, both the Chinese and the international versions. CodeBuddy is already international, now WorkBuddy is. It’s in Yuanbao and some others. So anyways, basic numbers, what does that mean?

 

11:10

Couple things got my attention. They’re going for performance improvements,  productivity improvements in real world scenarios. I like that, I think that’s great. So what does that mean? Software engineering. Yeah, that’s a big thing. Office and analysis activities. So you can see those line up pretty closely with CodeBuddy and WorkBuddy right there.

 

11:39

The third one, which they don’t talk about as much is game development.  Big surprise, Tencent’s the biggest gaming company in the world. They’re putting HY4 Preview into all their sort of gaming prototypes, gaming development, all of that stuff. And you would expect them to do quite well in all three of those categories. I’d expect them to be a major world leader in those three. Outside of that, there’s a lot of the stuff, but those are the big three.

 

12:08

What I’m kind of looking for, I’m looking for when you can put an agent in, whether a coding agent, a gaming agent, or an enterprise agent, that’s WorkBuddy, when you can sort of put that in end to end and it functions well. You really got to get it end to end. The entire workflow is covered, and usually there’s humans in certain parts of there, there’s the agent in certain part, there’s traditional software in certain parts.

 

12:37

You got to get the whole thing end to end and you have to get the quality right. Because if you can get that done, then you can start to scale up from one agent to two agents to 200 agents. But you can’t scale up the agent numbers until you get the quality performance right end to end. If your quality’s kind of sketchy where you have to have humans jumping in there and fixing things and checking things.

 

13:03

Well, when you scale up to 200 agents, the whole thing’s going to be a disaster. So it’s kind of like you got to get it done at the one to two agent level. It’s got to be seamless. The quality’s got to be trustable. Trust is a big deal. The reliability has to be there. And that’s actually pretty difficult. But when you can get that to work, then you can start to pull the lever on scaling up. That’s when you get an organization with 200 agents running all the time. And that’s a game changer.

 

13:32

You’re just different than every other business you compete with. So that scaling up question is very, big. That’s a tipping point. When can a company scale up their agent activities? The other tipping point is when the AI and the agents start to improve themselves without your input. That’s another game changer. And we can see that kind of happening in some cases like

 

14:00

How much of HY4 preview was created by HY3? Was it 20 %? Was it 80 %? It looks like it’s in the more than 50 % range. But once you get that to work, then the agents go to town and they start self-improving at a rate beyond what humans can do. Those are kind of the two tipping points I’m looking for in these models.

 

14:25

When can you scale from a couple agents to 200 without the whole system crashing and burning and your cost structure going through the roof? And when can they self-improve and how much? That’s what I’m looking for.  Okay, so that’s kind of, let me switch over to WorkBuddy now. Because I don’t spend a lot of time thinking about CodeBuddy because I don’t spend my days coding.  I spend my days thinking about integrating tech into businesses. So that’s WorkBuddy or Alibaba.

 

14:55

know, Qwen work.  Okay, so what’s going on with WorkBuddy?  Domestically, depending what numbers you believe, domestically I mean China, it’s arguably the market leader in China for enterprise agents. Now, it looks like that’s true, it’s hard to know the numbers, and also keep in mind, we’re in such early days, these are very, very small numbers, but it looks like it.

 

15:24

It looks like they’re sort of the AI office agent that’s moving. The numbers that come out from mostly Tencent, 20 million monthly active users, that’s June, 2026.  ByteDance is a player. There are probably 10 to 12 million monthly active users. Again, these are numbers reported from others. I don’t know how true they are. I suspect they’re roughly accurate.

 

15:50

Alibaba, which used to keep changing their name. used to be called Coder Work. Now it’s Coin Work. $8 million, something like that. So in theory, and this is from,  I think Pony Ma said this in their earnings report. Basically, they’re kind of bigger than the other two combined.  But early days, you’re talking about $20 million. mean,  in China, AI usage is up to $500 million monthly active users. So these are small numbers. But that’s compelling.

 

16:20

And the gross margins look overall, this is reported by them,  seem to be overall comparable to 10 cent cloud. Now that’s important because the cost structure for these things is significant. So if you’re not getting a healthy growth margin, if you’re not getting incremental revenue, these things are not sustainable. And that actually applies, you know, that’s for 10 cent.

 

16:45

That actually applies to anyone implementing agents in your business. You’ve put in one to two, it’s working end to end. You have to look at the incremental revenue and the cost structure because if it’s not covering the significant costs for these things and you scale up, you’re going to bleed red really, really quickly. So before you scale up from one to two agents and pilots to 100, you got to have the margins covered and you got to have incremental revenue and all that crap.  So, okay.

 

17:16

That looks like that’s what’s happening.  WorkBuddy looked like it’s doing well. If you’re in the Shanghai Metro, you’ll see WorkBuddy billboards everywhere.  Like they’ll cover, I took pictures of maybe 10 of them. They covered the whole section of the Metro and they’re everywhere. So they’re pushing hard. They look like they’re the market leader. They’re pushing for market share and to show incremental revenues such that the cost structure is covered. And ideally you’re making some margin.

 

17:45

Okay, that’s update number one.  Update number two, there is WorkBuddy Enterprise Edition. I’m using sort of the personal edition.  There’s an enterprise edition, which is basically when you go from obviously one or two people using it to teams using it, teams of humans in a business,  and you’re dealing with lots of types of agents you can deploy. That’s a pretty powerful upgrade. It’s a pretty serious one.

 

18:14

So their WorkBuddy Enterprise Edition has basically a three-layer agent team or a three-layer model.  Layer one, you can use expert agents. Layer two, you can use assistant agents. And layer three, that’s teams of humans working together and AI agents. So when you’re talking about a team in this enterprise, you’re talking about a mixture of humans and agents.

 

18:43

which is interesting.  So the three layers of the Enterprise Edition, Layer 1 Experts. These are kind of your role specific agents with very specialized skills like legal,  sales,  design,  software development and coding. Okay, you can call on these agents. They have them in a library and you’re going to use them sort of short term and on demand for whatever project you’re working on.

 

19:14

and they have pretty decent specialization in certain activities like coding or graphic design or something like that.  But your overall architecture is not going to access these by default. You have to sort of turn them on or off, use them on demand, turn them off or whatever,  and they’re not going to know everything about your business. They’re only going to know what context you provide in the conversation or by upload.

 

19:44

So you’re basically going to have to connect in with Skills or MCP to access your files or to access external services.  Okay, the assistant layer,  no, these are your full-time 24-7 digital employees. They’re operating all the time. You can turn them autonomously or on-demand. I you can say how much they work. They’re going to be doing ongoing tasks without intervention. They’re going to access…

 

20:13

Databases of files you give them all the time. So these are quasi digital employees That run all the time they can operate independently as one assistant or they can operate you can spin them up as teams and You can spin up one assistant who’s going to spin up several sub assistants This looks more like your employee base The first bit the expert that looks a little bit like your consultants Pretty cool

 

20:43

Third layer is teams. This is your collaborative project space. You’re going to have human employees. You’re going to have AI assistants all working together in complex workflows.  They’re going to do task handoffs. They’re going to centralize knowledge, skills.  This is your ongoing organizational collaboration layer. Pretty cool. That’s all. I’m totally convinced with that architecture. I think that’s great.

 

21:11

So WorkBuddy Enterprise Edition, really pretty impressive. The thing that goes sort of hand in hand is that is something called Agent Suite. This is basically, okay, you have WorkBuddy, but then you have all these other Tencent tools. I mean, they’ve got ad tools, they’ve got marketing tools, they’ve got document, they’ve got media, they’re the whole suite of products.  Okay, so they’re sort of re-engineering a lot of their productivity tools with AI.

 

21:38

And then they’re integrating them right into WorkBuddy. So no more APIs. They’re going to be sort of inherent to the product itself, which is called Agent Suite. And that basically was launched in June alongside the Enterprise Edition of WorkBuddy. Yeah, that’s pretty powerful. OK. So that’s Tencent Docs, Tencent Drive,  List Young, and all that stuff. Yeah, it’s all going right into WorkBuddy natively. No more APIs. That got my attention.

 

22:07

Alright, upgrade number three, update number three, this will be the last one. This is for WorkBuddy. And then I’ll talk about some other buddies. They have a tool called Human Agent Dual Writing.

 

22:19

Yeah, this, I’m probably going to use this. I have resist, I like to do my own writing.  I don’t like to use AI for writing.  I find it makes me dumb. I sort of resist doing that. So I will use AI for research. I will use AI to go look stuff up. I will use AI to fact check me.  I don’t let it write my content. I force myself to do it myself.

 

22:47

My sort of experience has been I can read stuff and I feel like I know it to one level. I don’t really feel like I know it until I can write it myself. And I usually write longhand with a pen and paper. I don’t let myself do it on computer because I can generate the articles but then I feel like I don’t really have it in my head. But if I can write it longhand on a piece of paper, I’ve got it. So I have resisted the idea of using AI to write.

 

23:16

And I still probably won’t do that somewhat. But they’re rolling out basically human agent dual writing, which means, you know, people use AI, then they copy and paste and they put it in an article or whatever. Okay, this is no, this basically runs alongside as you’re writing, it runs alongside another window as sort of your co-author.

 

23:42

Within your document, you select part of it, part of Word, part of Excel, part of PowerPoint, mark down files, and you basically give the agent, which is running in parallel to you, a natural language command to change that bit, and it will modify the selected bit, but won’t change anything else.

 

24:05

Okay, that’s kind of compelling. I could say, look, I’m looking at my article, I like this. Paragraph two, I don’t think I really made my point clearly. Clean up paragraph two, give it some bullet points and make the points sharper, but don’t change the overall argument. And then it just rewrites that section for me and I don’t have to type.

 

24:28

I’m tempted to use that. I don’t know. I know human agent dual writing.  Okay, that’s compelling.  Now, you could do this in a couple collaboration modes and they have three. One is human-human collaboration. Two humans just sort of type or just give natural language commands and it changes sections of the document.

 

24:54

one off, one on.  I could say, I don’t like this document we’re working on, change section two, it’ll just change the section. My co-author, a human, could do the same thing. You could have human-AI collaboration. Well, now my co-author’s an agent, an AI. Same process, but I’m not dealing with a human.  And then you could have AI agent to AI agent collaboration, no humans, two or three agents all working on the document together. Okay.

 

25:23

three types of collaboration to generate this stuff.  I’m not sure how I feel about that, but I accept that it’s super powerful. And I accept that that is probably the way that articles and documents and PowerPoints and Excel and Markdown files, I accept that that’s the way that most of that is going to happen in the future.  Okay, anyways, that’s kind of the upgrade.  Human-agent dual writing, that’s impressive.

 

25:51

The Enterprise Edition plus Agent Suite within WorkBuddy, that’s impressive.  then, yeah, all right. So that’s kind of the summary for WorkBuddy.  I’ll let you know in a week. I’m using it every day at this point. We’ll see how I feel about it. But yeah, so far I’m on board.  It’s pretty great. I put it up there with like Claude Cowork and all those, but way cheaper. And I would say more trustworthy, because I don’t trust those two particular companies at all.

 

26:20

But I do kind of trust this one. But we’ll see. That’s where I am on that one. OK, let me move on to the other topic, which is choose your buddy, which is I was at the Tencent exhibit at World AI Conference in Shanghai. And when you walk into the Tencent exhibit on the wall was a big billboard that said actually didn’t say choose your buddy, said find your buddy.  And they showed one, two, three, four, five, six, seven, nine buddies, all with little

 

26:50

you know, cute mascots that you could use. And one of them was WorkBuddy, one of them was CodeBuddy, one of them was Qclaw, which I’m not sure the status of Qclaw right now, but their argument, which I’m a believer in,  you know, these should be super useful in daily life and daily work, and they should be incredibly easy to deploy and use.

 

27:17

You don’t need to code, you don’t need to set up servers, none of that nonsense.  That’s their approach to all this AI and agent stuff. And their first iteration of that is basically nine different buddies you can just choose,  download, start using it. It’s really easy. WorkBuddy? Yeah. Go to the international edition, download it. It’s super, you don’t have to install or do anything. It works right out of the box. It’s super easy. I like that overall approach. And I know the buddy thing is kind of cute.

 

27:47

So, okay, WorkBuddy, code buddy, fine. I’ll give you a couple others. I won’t go through all of them, because some of them are,  eh, some of them are, I’m going to use some of them, I probably won’t. There’s one which is basically, the mascot is a piece of toast. I don’t quite get why it’s toast. Tu Su is the Chinese name.  It’s basically a vibe coding tool for…

 

28:15

individuals who aren’t coders and who aren’t really enterprise. So what would a regular person who doesn’t code, who’s not doing this for work do? Well, they would create a mobile app. And that’s what this one does. If you want to build a functional mobile app using just natural language,  no coding required, this is it. Look for the toast icon.  You basically describe in plain language

 

28:44

I want to create an app that helps me manage my luggage packing as I travel around the world. Because actually I have to have a long list of things. I travel a lot.  This week was Manila and Honjo. I’m always forgetting stuff. OK, maybe I code up an app of all the things I have to do. Turn off the power, turn off the water, make sure you have a QR code for arrival in the Philippines, which is a new thing.

 

29:13

Make sure you did the QR code for arriving in China, which is another new thing. Don’t carry a portable battery in China domestically anymore that doesn’t have the 3C approvals.  If you have a regular battery from outside of China and you go into China and you try and fly domestically, they’ll take your battery. I’ve lost one.  Anyways, I could think up a long list of things I keep forgetting as I travel on the world. I want to create a simple app.

 

29:39

that will tell me everything I need to do and make sure I put in my flights, my countries, it tells me everything. Okay.  I describe it in plain language. It will spin up three agents,  a product planning agent, a design agent, and a development agent, a coding agent. They’ll all work together. They’ll complete a complete app for me.  And if I want to share it with friends, fine. If I want to use it myself, if I want to put it up into the Apple iOS store, it’ll do it all.

 

30:08

As of July 2026,  Toast has generated 50,000 apps with 30,000 users. Most of them are kind of lifestyle tools, education tools, a lot of mini games. If you want to spin, actually, this, would be compelled to do this actually.  Every now and then I think about creating a little mini game. I’d like to do that.  Like a board game and stuff like that. And I never do it. I could do it on this.

 

30:35

Okay, kind of interesting. I don’t think I’m going to spend a lot of time there, but not bad. So that’s the Toast app. There’s an app called Marvis,  which is the logo is a horse with a red scarf. Okay, I like that better than the piece to toast.  Marvis is literally named after Jarvis from the Iron Man movies. It’s your AI assistant that you can talk to, but it lives on your device. So it’s something you’d put on your laptop.

 

31:04

and you would give it free reign for your laptop. And it would tie into your phone and it would tie into other devices, which is kind of how Iron Man uses it in the movie, right? Tony Stark, he sits in his office and he talks to Jarvis.  It’s the same thing. This was launched May 20th, 2026.

 

31:23

Okay, how is that different than chat GBT and things like that? Well, this is dialogue based AI and it basically operates your laptop. That’s pretty much it.  And how much permission you want to give? I don’t know, do you want to give it operating system capability? You want to give it administrator role? You want to,  I don’t know. And it works cross device.  So personal devices, your personal document.

 

31:51

System level control it can check your battery health it can you know look at your operating system usage things like that it can optimize what you Run it startup cleaning files, so you want to give it system level control you can do that File management I actually find that pretty helpful I would actually consider using this just to manage my library of documents and books and that’s pretty good cross-device control can I

 

32:20

see my PC screen from my phone. Yeah, you can do that. I can unlock my PC from my phone. Okay. Can I run my apps on my PC if the app is on my phone? Yeah, I can do all that.  Okay.  Compelling. You can basically install it out of the box. You don’t have to do anything. You get six AI agents immediately.

 

32:48

I’m kind of half on the fence on that one. think I’d give that a try.  I’m not sure. I don’t know why it’s a horse with a red scarf, but yeah, that’s better than the toast. Okay, here’s one I actually, I’m going to use. This is called Ardot, Er dian in Chinese, I believe. The graphic is a little black square. I don’t know what that’s about. Okay, this is a design agent. This is a creative tool. You can create digital products. You can create apps.

 

33:18

pages. You can create websites. You can do posters,  illustrations, designs. You can do PPT presentations. Okay, I like that. How is that different than a normal? There’s a lot of these. Well, this one is meant to be sort of editable, reusable. So it’s giving you sort of standard ongoing assets, digital assets that you can use over and over and evolve over time.

 

33:44

as opposed to sort of a static output and then you’ve got to start over. So I could take, let’s say, one of my keynote presentations, which are usually PPTs, I could put my keynote in there and just use it as an ongoing digital asset that I can edit and change and upgrade,  maybe make it look a little better, because my graphics don’t look very good, generally speaking. I could give it my website, I could give it posters. I like the idea of having sort of

 

34:14

a core set of digital assets that I’m always upgrading and iterating on as opposed to starting over and over. I must have a 1,500 keynote presentations, pages,  in files going back 10 years. I would love to pull those all together into some ongoing digital assets I could upgrade over time. I like that one.

 

34:40

And that’s kind of the differentiation that you can change these things over time.  Related to that, there’s a buddy called Miura. This is for creative professionals who need visual content.  And this is full stack, everything. Create your brand identity, create your marketing campaign, create your videos.  It’s a,  like, if you were going to hire a design studio,

 

35:06

to look at your business, do the brand, do the marketing, do the videos, do the creative elements.  This is a software version of that. And you can do complete visual packages for a marketing campaign, all of that. This is for brand designers, for marketing professionals, for creative teams. Yeah, this is a real problem for agencies that used to sell this work using teams of professionals, and now the AI is going to do it. Yeah, that’s pretty…

 

35:35

That’s pretty powerful. I will probably start using this because I don’t spend a lot of time thinking about brand design and all that and I should because my stuff is pretty basic.  I joke about this.  My web pages look pretty basic and they look basic forever.  They don’t function awesome but people don’t really go to there for a nice web page. They go for some level of thinking and that’s what I compete on is level of thinking. But I should really do that. It’s lame of me that I don’t. So brand design.

 

36:03

e-commerce advertising, video and campaign assets, social media and content, stuff like that.  So Miura, think that’s on my short list. Last one, which I won’t. There’s another one which is called IMA. I-M-A. This has been a, this is nothing new. This has been a Tencent thing for quite some time.  IMA is basically like a library you build for yourself for all your knowledge. And

 

36:33

You know, that’s been a thing for them. And there’s an enterprise version and then that’s Lexiang, I think the name’s Lexiang, Lexiang. The personal one is called ima, I-M-A. Okay, that’s become really important because they’re basically putting AI and agents inside that. It’s called IMA Copilot. That’s the buddy. Yeah, as you start to do more and more stuff with agents, you’re going to have a library of repository of knowledge.

 

37:01

that you’re going to have your AI and agents draw on. And you’re going to build that over time and you’re going to give it access to your expertise. Now I’ve been basically doing this, when I do AI and agent stuff, I just point it to my webpage because my webpage already has like 900 articles and 295 podcasts, all of which are, so I can just sort of point to that as a repository of knowledge. But most people have not spent a decade of their life.

 

37:29

creating content on one particular topic.  I just happen to have done that so I can point agents to my repository, which I already have. Most people you’re going to want to build a knowledge base for yourself.  IMA Copilot, that’s a good one. I’m going to look about, you know, whether this makes sense and then if as you start running your models, your agents, let’s say you’re using Hunyuan or whatever, yeah, you’re going to use RAG to draw on additional knowledge as needed. Okay.

 

37:58

That’s IMA Copilot.  That one’s probably pretty powerful.  I expect this to become a more important part of just sort of WorkBuddy as an architecture. You’re going to want WorkBuddy, if you use WorkBuddy, you’re going to want that to sort of be your main tool, but it’s going to draw on a database, which is probably going to be IMA, and then it’s going to draw on a library of skills and agents that you can pull in, which is kind of what they’re doing right now.

 

38:27

That’s how I think it’s going to play out architecture-wise.  Those are the agents I think about. There’s the Toast, which is the Vibe-code and mobile app or a mini game if you want. I’m going to do that. I think that’d be fun to make a game.  Marvis, which is the Jarvis, let it run your laptops and all that.  I’m on the fence on that one. I don’t like giving anything access to my laptop.  R.dot, is the digital assets piece. I like that. I’m going to use that one.

 

38:57

Mura,  full stack creative design. I’ll probably have it go through all of my stuff and give it a makeover. Brand design, marketing, ad campaigns, all of that. And then IMA. Yeah, IMA is the big one for me. That’s probably, I’ve got to put all my knowledge into a more structured database. I can’t just keep pointing AI and agents to my web page. I’m going to figure out how to do that. Okay. That is it for today. Hunyuan, WorkBuddy.

 

39:26

five or six other compelling buddies, but I like this approach and I like that it’s real easy to use, the price is much lower than what’s coming out of the West and they are absolutely going international. So this is going to be available for everyone. I’ve been using WorkBuddy both in China and outside of China. There’s two different versions you can use. The outside of China one is not as robust yet, which is what you would expect, but it should be.

 

39:51

Anyways, that’s it for this. I’ll probably do another podcast tomorrow because I skipped last week. mean, it was an amazing week. I’ll just sort of talk through what I was doing.  I was in Manila finishing up this course, which is sort of a partnership between me and Alibaba teaching e-commerce strategy to entrepreneurs.  That went really well. I am going to roll this out. If you’re interested in having a course, here’s the way the course did. We did about

 

40:19

four to five weeks of sessions,  about seven modules overall, each module’s three hours of just like top to bottom, here’s how you think about e-commerce and e-commerce meets AI and how you can spin up a one-person e-commerce company now with agents. Like the content was, I thought it was fantastic and it was about 70 % Alibaba content and they’re the masters at this subject, full stop. So that was kind of phase one and then.

 

40:47

At the end of that, we took a smaller subset of that group to Hangzhou and they spent three days on the Alibaba campus doing classes, visiting the largest KOL studio in China, which was fun. I’ll show pictures of that. Digital live streaming where you can do 24 seven live streaming with no humans, just digital avatars, which works really well. I took them to G. Lee.

 

41:17

Zeekr, which people don’t really know as well. It’s the number three EV company in the world. It’s about to pass Tesla. That’ll happen in the next couple months.  Deep Robotics, which is one of my favorite robot companies. And we just had a great time. And that was sort of phase two of this program. And it went great. I’m going to, this is going to be my thing. I’m going to start rolling this out internationally to other markets. Looking at Brazil, looking at Southeast Asia right now.

 

41:47

And that would either be sort of my thing or in some cases it’ll be a partnership with Alibaba again. But that’s obviously up to them to approve so I can’t say we’re going to do that. But I’m going to definitely try and do that in a couple other markets. If that’s of interest, send me a note. We’re rolling out right now; it went really well. Big success, I think. So that’ll be the thing. I’ll talk about all this tomorrow. Had a really fantastic couple of weeks.

 

42:12

A lot of good content, a of good thinking. But yeah, I’ll go into sort of how we think about all that stuff. Okay, and that is it for me for the content for today. Another one, think tomorrow I’ll do it. I’m back home, so I’m sort of back on my content creation schedule.  That’ll be it for me. Anyways, I hope this is helpful and I’ll talk to you probably tomorrow.  Bye bye.

 

——–