Sean Chen: From Google to the Streets - Startup Hustler & Content Creator

In this episode, we sit down with Sean Chen - a former data scientist at Google and one of the first AI product managers at Walmart/Sam’s Club - to unpack his whirlwind journey through tech and startups. 

Sean reveals how VC pressure can steer or stall your vision, and how he and his co-founder built “Shelfer,” a voice-driven inventory app for SMB retailers, in just two days on London’s shop floors. We also break down the most realistic agent-AI use cases (of course, how could we miss the buzzword), dive into authentic content creation (and handle criticism like a game), and explore why scrappiness often trumps perfectionism. 

Tune in for a candid conversation about hustling, pivoting, and staying true to yourself in the fast-moving world of entrepreneurship.

Transcript

Rui (00:22)
Hi Sean, how are you doing today?

Sean (00:24)
Hey what's up? Doing pretty good, how are you?

Rui (00:27)
Great, I mean, it's early morning for me, but I'm super glad to see you again, even though like technically we're just chatting online.

Sean (00:35)
Happy to have deep discussions I'm ready for all your questions.

Rui (00:39)
maybe let's start with a little bit self introduction, because I know you wear many hats and you do many things throughout time. And you are one of the few people have like super high energy and just like keep trying new things and moving around a lot and just do a lot of experimental ⁓ stuff. So I will just give you the space to actually make the

Sean (00:55)
Ha ha ha!

Rui (01:05)
self-introduction because I honestly don't know how to pick from all of your different experiences.

Sean (01:12)
Okay, sure. So I'm Sean. I'm from China. I studied in several different places. I studied in Australia, UK, in the US, and I mainly worked in the US. My career started as a data scientist because I studied statistics, machine learning. Ray, you and I met at MIT studying machine learning optimizations. And before that, I was actually in London studying at University College London for stats and econ.

and I started my career as an AI scientist at ⁓ several different companies, initially at CVS and then at Google. And then later I pivoted into product management. I became one of the first AI product managers at Walmart slash Sam's Club. Last year I quit my job, moved back to London in the UK and joined this startup incubator called Antler, which is a Singapore-based VC incubator.

that has an existence in Europe as well. And we built a startup called Shellfer, ⁓ which we're not continuing anymore. But it was a really wonderful experience. My co-founder and I had a great relationship and ⁓ have been hustling quite a lot with that idea. And it was wonderful. We onboarded like 70 different users in the retail industry. And it was basically helping SMB retailers to set up

inventory system with voice AI. And then later this year, I was on and off teams, sometimes being a solopreneur, launching random products, ⁓ and sometimes working with some friends, trying to tackle the next billion dollar idea as we claim it to VCs. But in reality, we're just like hustling all the time and struggling. ⁓ Yeah, this is who I am. Happy to

dive into anything that you feel curious. ⁓

Rui (03:09)
Yeah, for sure. I mean, I'm curious a lot of the things that you have mentioned, but we can dive into one at a time. So maybe since the last time that I talked to you,

It sounds like you do have the plan to come back to the US. ⁓ But let's chat about your experience actually in UK because you've done a lot. Like, I see your activities on LinkedIn, your YouTube channel. Yeah, on top of everything that Sean is doing. He's also a great content creator. You're also a very prolific one. Like I saw that you have about like 65 videos on YouTube channel and every video is

Sean (03:24)
Ahem.

Okay.

Rui (03:51)
from like 10 minutes to 30 minutes. That's quite a bit in the past like two, three years, right? ⁓ But anyway, I mean, just in your experience, in UK and trying to figure out what exactly is the type of product that you were trying to build. I know that you joined Entler, which is this well-known ⁓ tech incubator in UK, London.

Sean (03:58)
Thanks.

Yeah, sure. I actually applied to Antler right before I left the US ⁓ fun fact, I got their interview before I crossed the border in the UK. I think that's like a very interesting milestone. ⁓

And then, just after several interviews, we got in and I met my co-founder there. ⁓ We started actually several ideas at the beginning. ⁓ My initial attempt was in education tech. I actually built something for some customers based in some Australian schools. this is one thing that I feel like incubators

might influence you in a not necessarily positive way, I would say, is that they want you to build something that's targeting a much bigger market than ⁓ something that they don't think is going to be as big. So a lot of VCs don't think that education is that big of a market. So they kind of like forced us to drop the idea, to move into something that they think has a better founder fit for us. My co-founder and I both had a strong background in retail. I was at Google shopping, YouTube shopping.

⁓ Sam's Club Walmart. My co-founder was a ⁓ channel manager in one of the major furniture sales team based in West Africa before. So we both had a strong background in retail and we built this product called Shelfer. As I mentioned earlier, it's a voice AI based mobile app that helps SMB retailers to set up an inventory system from scratch and then that mobile app could track

the life cycle of a product. Say for example, if you're a family-owned business of five different retail shops in London, and you import a ton of ⁓ certain type of sauces for some South Asian dishes, and ⁓ traditionally they just like take down notes on their notebook or some more advanced retailers, they have a ⁓ beep system that can like scan things, but then they'd have no idea about

how much is their shelf value, what kind of products are expiring very soon, and how much waste they're wasting, like they have no idea if they should be procuring certain products ahead of time, diversifying their product categories, all these kind of things, they don't have these digitization solutions. So we decided that, let's build something easy for them.

And we started with mobile because most people have a mobile phone and they don't have a computer. And even they have a computer, they don't really use computer for checking the dashboard. So mobile phone is very handy for them. We started in September, ⁓ actually I built a prototype in two days and then went out there and then started selling. And then some retailers actually downloaded in the first day or the second day.

even if the product is not even live on Apple, not live on Android, it was just like a QR code they have to scan and download this sketchy version from a website. And some people were like, are you going to hack my phone or something? Some people were really, really suspicious about that. yeah, starting from there, we iterated quite a few different versions. And I think in a month or two months time, we had like 70 retailers using it.

Rui (07:07)
you

Sean (07:21)
So if you were in London back then in November and October, you'll be seeing me and another big guy. We're like hustling around and walking around the streets, speaking to every retailer we could find. It was a really fun experience. ⁓ Yeah, to be honest, my whole Antly experience is just like that with my co-founder and I made a ton of good friends there. Basically started a whole new...

ecosystem of friends in startups in the UK because back then most of my friends are in banking or consulting and not a lot of startup folks. So Antler really helped me to start my network in startup ecosystem.

Rui (08:01)
well, first of all, I love the story. you mentioned the part about how VCs, the investors actually almost directing and shaping what you're trying to build. ⁓ And you mentioned the downside of it, right? Because maybe they believe that they have

Sean (08:15)
Yes.

Yes.

Rui (08:21)
a wider lens and that's why they believe that certain area would be a lot more profitable and also because of the addressable market size. It's just a lot bigger. I actually don't really know how to feel about that because sometimes I feel like in order for someone to be truly successful, you have to believe in the problem space that you're solving and not every investor actually has that type of building experience or have that really, really deep.

domain experience to feel the customer's pains and everything. So how do we really balance between your own conviction versus someone who might have a wider, let's say, experience and insights into the industry trend?

Sean (09:01)
That's a really good question. I don't have a good answer for this because apparently I didn't continue with my previous idea. Even if there was already a product, there was a design partner, a school was using our product. But I think if I look back and then face the same person who told me to stop, ⁓ how would I do that differently? Or how would I think differently or more thoroughly? I think a few things. Number one,

Is this area or this direction that you're building for or you're solving a problem for, ⁓ how much passion do you actually have for it? Or are you just picking an area that seems like there's something interesting there? Or pick an area that you feel like you can relate to it a lot, like education is something that a lot of people can relate to. You need to really make a judgment of ⁓ how much passion you have for an area because if you decide to build something,

for an industry, you 100 % are going to receive all sorts of criticism, cynicism, from all sorts of directions. So it was a good test actually that the partners at Antler were pushing us and I actually backed down. So that probably says that maybe I'm like 1000 % passionate about it. Now I think about it, maybe that's the reason, but I think that's a good test as a first point.

that as a founder, you need to think very thoroughly if you want to do this, even though nobody believes in you. So a good example is that back in 2013, 2014, nobody's going to believe that there should be another note-taking app. But Notion was born from that time, and it worked. So you got to do something that you think is... You're solving something that's fundamentally broken, and you believe that, and you think everybody else is a fool.

not believing in you. I think that's the kind of craziness that founders really need to continue on something like a note-taking app back in 2014 when everyone was using Apple Notes and Evernote and all sorts of other note-taking apps. That's the first thing. I think the second thing is

it depends on the team because my co-founder and I, even though we were both interested in education, at that time my co-founder mentioned that he was not believing in using software to help students learn because students didn't want to learn. He was actually an education tech startup founder before. And then we had a discussion about this and I think there's some fundamental misalignment.

in this direction between us. So that pivot was a very natural choice with or without the push from Antler if you want to stick with this founder, if you want to continue to build a business together,

you have to either convince each other or break up or pick something else. And we decided to pick something else and that worked out well. So I think that was a good decision. It was a critical one to make.

And lastly, I think the education market or whatever idea you choose, like if VCs don't believe in you.

To be honest, sometimes VCs don't know anything that an area that you spend perhaps two months researching, building and testing about, they probably just read some old three summary from deep research, or from their own impression that this area is not going to scale. It's very possible. ⁓ So I think if I were to...

jump back to that time again, the most important thing is for me to do much wider research instead of listening to one voice only and understand that who has been in the shoes of the customers. And if the advice that I'm taking from was not originated from customer originated ⁓ experience or pain points, then ⁓ that's not the kind of advice that we should be taking.

Rui (13:02)
think this is an extremely thoughtful answer. And actually, especially your first point because I was going to follow up with another question, which is how much do you actually believe in the problem that you were solving? ⁓ Because for me,

Sean (13:05)
Thanks.

Rui (13:16)
a huge struggle is to even find the problem space that I feel like, yes, I can sink hours and hours of my life into the space because I just truly believe in the version of the world that I really want to create. ⁓ and the lack of problem statement has been the ⁓ barrier for me to really

dive into a startup mode. But at the same time, I understand sometimes the problem statement doesn't just jump out right? Like you have to keep iterating and pivoting until you finally feel like maybe you are clicking with something. So this is very interesting. also I love your answer around

how sometimes the VCs are not necessarily correct. A lot of times they're also just making judgments which may or may not be well funded or trained. that's a really incredible thoughtful for answering a very short period of time. So thank you so much for that.

Sean (14:12)
Thanks, thanks. You can see how much like, how much struggle I have had. So that's why I was like, okay, man, I have a lot to say about this.

Rui (14:20)
I want to take two step backs because the first answer that you were giving me has this really fun story about, yeah, we actually built the product in two days. And then the customers are downloading them from the websites and then thinking maybe this is a scam. I really love that bit.

Sean (14:31)
Mm.

You

Rui (14:38)
It sounds like you were extremely scrappy. And I find that to be a trait that I strive to have a little bit more, I think probably all of the education systems push students to get the A, you know, the top scores. And so you're striving for perfection. You're striving for solving the hardest problem possible. But now it seems like you have to retrain yourself in some way Is that true? Like how have you changed? Or maybe that's always been part of you, being scrappy.

Sean (15:07)
yeah, great question. I think it's always part of me. It's just, you decide when you're going to show that side of you and when you're in the right ecosystem to let that side of you to make you shine. In a big company like in Google, if I say, okay, I built something quick and guys, please try it. If you want to say like, oh, have you done a code review? Oh, have you done the like, I don't know, whatever, security checks and take like quarters.

Similarly, in Walmart, in Santa Claus, same thing. We tried to build something scrappy, and then we showed the rest of the team. And we got questions from this guy who was the CIO at that time of the company.

the reaction from them were that you know, this is this is bad ⁓ You're gonna leak so much information, which is right like for companies like Walmart They can't afford to lose any bit of their customer information Without you know the proper security checks like that. He's right and it turns out later that this guy was from FBI He was a retired FBI ⁓ top leader

who became the CIO of my function at that time. So with this kind of resistance, to be honest, you can't do anything. Yes, you can build something innovative, but then you know it's going to get killed at some point because nobody's daring to take that risk. And in a startup world, it's a completely different school of thoughts. I'll tell you something funny.

I had a colleague back then at YouTube that I was working with and she was telling me later that she was trying to find this job at this YCE startup with scaling and she got rejected immediately from them and the reason is that you worked at Google for too long. You worked at Google for 10 years. Nobody's going to take that risk anymore because it's risky for a startup to hire someone who's been in an unrisky environment for a long time because you don't know how to be scrappy. ⁓

Rui (16:57)
Wow.

Sean (17:00)
That's their stereotype, of course. My colleague was a very, very talented person. She has a lot of experience, 10 years at Google, very, very smart, ⁓ and a good mentor of me. But then the kind of advice, the feedback that she got shows how much like stereotypes people have. But stereotypes come from somewhere, right? Stereotypes don't just emerge from nowhere. And I think the true part of that stereotype is, yeah, in the startup, you just have to like forget about a ton of things that you were... ⁓

you have to learn as rules of behaving well in a big company. You just have to drop that. if you were in my shoes, we had like two months and a half or three months to incubate something. Am I going to like spend two months to build something? No, I only have two days. I need to pitch next week to show people what we've got. And in order to show people what we've got, want to...

We want test something, right? We want to get some real feedback from the customers. And the traditional way of showing people a Figma design doesn't work with the retailers. Like nobody's going to check your Figma. Nobody understand what a Figma is. If you show them something that you can click, jump to a different page, they might ask you, why does it not work? What are you going to say? Oh, this is a design Figma. Nobody's going to understand what you're saying. You have to show them a real thing and it works. Even if it's sketchy, there's no security checks, it got to work.

So that's how life is. And back to a question, did I change? I think it's always part of me. I always tried to hack the system when I'm in bigger corporations. But ⁓ the system doesn't reward that. The system actually could potentially discourage that.

Rui (18:39)
So what exactly is the problem that you're working on now?

Sean (18:44)
I haven't really consolidated the idea yet. There's no product yet. There's a concept. I believe that sellers, could be manufacturers, could be retailers, basically sellers who have a lot of customers to manage, they traditionally use a ton of tools called a customer relationship management tool, CRM, for a deal from the beginning.

of the customer coming in and asking for the part that they're purchasing to the end when you finished selling and ship the whole thing. It's called a deal flow. That deal flow has a ton of very manual processes And it's causing very high cognitive burden on the salespeople, on the marketing people, on the business owners. And I identified that problem through

open source project that I launched in the past and some SMB business owners reached out to me and we had a conversation and realized that ⁓ there's a pain point there. recently I've just been like really trying to double down and explore the details of the problems there. And you can think potentially the solution is an agentic workflow to replace some of the very manual work like

checking what kind of deal is stuck in the flow and why is it stuck in the flow? Why is it not moving from negotiation stage to signing the deal stage? Why is it still stuck at quoting

If you're a sales or if you're a business owner, if you have three customers, you don't need any software, you're fine. But if you have like 100 customers or 100 deals waiting for you, your day is going to be like a total mess and you have no idea who to prioritize and what information should you be looking at. And I think LLM is going to do a great job at helping these people to...

reduce a lot of the cognitive burden so that they could be focusing on being better sales,

Sean (20:35)
And I researched a ton of existing AI tools, either these big companies trying to add AI to their flow, like Hopspot, Salesforce, they're all working on AI, there are a ton of automation tools ⁓ claiming that they're using AI agents to automate ⁓ a lot of the workflows.

I think we're at a very early stage and I don't think that AI agents will replace everything.

I think ⁓ it's going to take a lot of really cross-disciplinary knowledge between software engineering, using AI, agentic thinking, as well as business thinking, sales thinking. for example, in retail, There's much higher frequency of ⁓ replenishing your stock,

Versus like manufacturing takes a year to deliver the 5,000 parts that you ordered. So these kind of scenarios really differ. it will be difficult to have a one for all AI agents solution to manage these kinds of sales deal flows. But there's enough pain for people to jump in and try something. I think that's the exciting part of this direction.

Rui (21:42)
when you talk about engineering AI, because this is such a buzzword to a point where I don't even want to talk about it, but at the same time, it's such a buzzword for a reason. And I have to say, I do have a lot of skepticism over it, I'll just give you a very simple example. I was trying to ask Chet GPD to help me modify a spreadsheet.

simply filling some dates based on some patterns that I have in mind, right? The first try, it actually

got 80 % right. So I was like, oh, OK, But then I do need to get it right. So for the maybe like 20 % of the correction that I needed to do, that took a significantly longer time for me to finish. One, it starts to hallucinate. It just randomly deleted columns.

Sean (22:13)
Hmm.

Rui (22:32)
from the spreadsheet when I never even mentioned the word delete. So that's one. Another thing is

I'm writing requirement essentially for GPT in a way that needs to be a lot more specific than a human being, right? Because somehow humans, can infer what the requirement is, like build X and then it's pretty straightforward to a human being, but somehow it's not as straightforward to a machine. And you actually need to break that X down to even smaller units.

And that actually makes me realize that, agent AI, at least in the first iteration, how much it could infer from ambiguous context accurately that's extremely difficult. And then I was trying the product Zapier, you know, which is this software that basically enables you to transition a workflow from one software to another, let's say from Google Gmail to WhatsApp.

I was trying to see whether Zapier could take the date from the spreadsheet and send an automatic notification to the WhatsApp message between me and that person to say, hey, this is the due date and we need to do this. And then please check the task. This is a very

specific workflow. And it's very simple because there are only two endpoints and then one tool that you choose to connect between these two. even that, there is a little bit like hiccups. But it's still the best workflow that are ripe for this type of agent AI, because the task is very well defined.

Sean (24:09)
Yep.

Rui (24:11)
I just don't know exactly what people really have in mind when they talk about agent-gk-ai because to me, at least for now, those extremely specific and repetitive and not super fancy workflows is the first one to go.

And that's the most realistic one. anything fancier than that, I think people are going to run into a lot of frustration and disappointment.

Sean (24:33)
Yep. You just answered your own question, by the way. That's exactly how you use agents. And you just gave the best summary of what pain points AI agents currently are, and how do you use agents smartly and not shooting yourself in the foot, What kind of tasks agents are really good at, and what kind of tasks agents suck. I think in this nutshell, firstly, like...

For the tool that you're trying, I think it was opinion shared by Sam Altman, the host was asking him similar questions. And he mentioned that ⁓ it really depends on what the startup is really betting on. If the startup, some startups are betting on say, okay, it's currently LLMs are making a ton of mistakes. So let me try to use LLM to do things that are

very, very, very specific. And there's not a lot of room for, ⁓ even if you make a mistake, it's fine. Like conversational ⁓ agents for you to talk to, right? It's for fun, And there's some ⁓ apps for kids for Christmas that you can talk to. These are like fun apps that ⁓ could catch some attention and then can make some quick money. That's one type of companies. The other type of company is that they're making their products

that are barely working and they're betting on the vision that LM is going to be so much better five years down the road so that maybe tomorrow when they wake up, GPD-5 is out and everything works with very, very rare mistakes being made. These are all possible. I think ⁓ Sam himself really is betting on the second one because he is working on the foundational model research. He believes the fundamental model is getting better because he has seen how GPD-4.0 ⁓

how much the improvement was compared with GBD 3.5. I think currently we're kind of stuck at GBD 4.0 because we haven't seen GBD 5. people are starting to think that, OK, so maybe this is where LLM is. ⁓ And then you come back and think about, so if this is the situation, how can we build around AI that can actually provide value?

And I was watching this YouTube video by this engineer called Barry Zhang from Anthropic. He shares some very interesting insights I'd like to share ⁓ on your channel as well, ⁓ because I really agree with that. I think he mentioned three points. I'll try to see if I can quote. The first point is ⁓ don't use an AI agent if you can just use workflows. Because of the fact that LLMs essentially...

are really good at processing text information and give you back text information or give you like structural information, but it's not very good at like automating the whole thing. Even if like a ton of agents, they will tell you that it will be smart enough to use tools. But I, as a founder personally, I wouldn't let my agent to use tools for me unless I feel like I'm very a hundred percent confident that I can do it very well. So I probably just do like very, very simple tasks that

an agent could do very well. example, when I was building Shelfer, the agent that helped me save a ton of time was turning the voice of a retailer who has, say, a Pakistan accent, turn that into a form that logs the information in the structure that we already predefined. What's the part name? How much is it? What's the discount? What's the expiration date? All these kind of things.

I haven't seen mistakes, to be honest. 99.9 % accurate. And even if it's not accurate, the human could edit it later. So that was kind of the scenario when we thought that agents was okay, it was doing a good job compared with a human trying to type it down and put it into forms. And I think a lot of the so-called AI companies are doing similar things. And there are some other like very, very fancy AI companies like workflow builders like Zapier.

Or some generic agents like Manus. Manus was claiming that they built the generic Asian platform that could do anything for you and they're using tools and stuff. But they're doing research. Doing research means that sometimes it doesn't matter if you're making some small mistakes. There's not even a criteria of what a mistake is. You're just doing research. You're searching around. You're using tools. You're using maps.

That's fine. So I think it really depends on the use case in a nutshell. And it depends on what the founder is really betting on. sorry, back to Barry's point, use a workflow as much as you can, unless you have to use an agent. And the second point was that if you're using an agent, try to make it do a very simple task, as simple as you can. Or...

in other words, as specific as you can, which really aligns with what you said. You probably have seen some browser agent or some agent that can just help you click on some software and then finish the task for you on the software with the mouse. That was like...

a technology that was ⁓ not even new, it was an old technology. It was basically taking screenshots of your laptop every five seconds or every two seconds and then locate where the button is and then use that operator thing to move your mouse over that location and hit click and then jump to the next page and take a screenshot again. that it will, it's trying to observe this world but then it's very slow and then

If you don't give it very strong, very specific instruction, it's not going to do anything for you.

Rui (30:07)
Hmm.

so to close up the story and to piggyback on the point that you made about how do we really use the agent right now. The end my exploration in the story is I give up on the Zephyr altogether.

Sean (30:29)
Hahaha

Rui (30:30)
⁓ it was like maybe 50 bucks with something a month, ⁓ for the service, that's very expensive. I don't really want to pay for that. ⁓ second, because of that price tag, it forced me to think, wait a minute, do I really need something so fancy you know, making a reminder specifically in the WhatsApp chat that I have? And then I just came to the conclusion. You know what?

I'm just going to set a calendar reminder, block out, let's say, Sunday morning time, and then invite that person to the calendar and making sure that there is notification every 10 minutes to remind you that please go check this thing. And then you probably just quickly glance through the spreadsheet and see which roles are actually about to be due. And that's it. That's how I solve.

Sean (30:59)
There you go.

Rui (31:24)
the giant problem that I had at that point in time in a very simple way. Because at the end of the day, you probably could come up with some very simple solution, but somehow the fancier to trick you to think that, oh, my life could be much easier if I can just solve it with this specific tool. So that's the end of my personal tiny exploration.

Sean (31:44)
Yep.

A lot of companies are claiming they have really high ARR, monthly ⁓ recurring revenue, annual recurring revenue, but they're like multiplying their monthly revenue by 12 to show that this is my annual recurring revenue. But in reality, most people are just curious about it and they're going to cancel it two months later, like what you did with Zapier. So right now, a lot of the fancy tools that people are building are...

Rui (32:10)
Mm.

Sean (32:16)
⁓ riding on the waves of this new AI curiosity economy where people are happy to try new things and then after two months they're like, this is a piece of shit, I'm not going to continue. that's the fun part of building anything right now because you don't know which one that people are curious about are going to last for a very long time and then be like, holy shit, this is actually good. Like cursor is actually really good.

I started using Cursor in August or something last year. And then people around me were asking me like, dude, I prefer to just do raw chat GPT. I prefer just copy paste code into my VS code. I don't buy in this Cursor shit. a few months later, I think there was a YC partner on X at posting that they were asking the batch of Y Combinator at that time who's using Cursor and 95 % of people were raising their hand.

So I think some products stay, some products die, ⁓ which is giving this game more fun. You never know which one is going to stay. ⁓ There's always going to be skepticism at the beginning. But it turns out so far, the best use case is using AI to help you code. It's the most productivity boost compared with any other use cases in the world for using AI.

Rui (33:33)
Yeah, fascinating. ⁓ I know that I would have spent a significant amount of time in chatting about startup, agent AI and things like that. But I'm actually also curious about some of your life experience

talked a little bit about your YouTube channel. And by the way, check out Sean's YouTube channel. great. And it's very authentic.

you're pretty prolific and you're active on so many different social media and platforms and honestly I'm not that.

type of content creators. For some reason, I want to hide behind the screen. Maybe it's just another barrier for me to really cross at some point. I'm actually curious, what type of topics do you usually avoid or

Sean (34:03)
Hahaha

Rui (34:16)
hesitate to post in public space. Because you seem to be very open about everything, and you're very genuine in your answer. But I'm actually curious whether there's any hesitation on your end on any topics.

Sean (34:29)
I don't want to post complaints, if that makes sense. Because I feel like there's so much negativity on the internet and I want to share more positivity. And I like myself better when I'm sharing positive things, Like, you definitely probably have watched some YouTubers who share how bad things are, how...

Rui (34:32)
Mm. Mm.

Sean (34:57)
things have changed. I don't want to share that kind of thing. Of course I have negative thoughts. I have negative feelings, but I prefer not to share it publicly because I see my contents as a place to form a really high quality community. Like people who reach out to me after watching my contents, reading my contents, some of them become really good friends. So I want to pass on the positive mindset of

mine to the rest of the world so that people who share similar feelings or people who find it helpful will be naturally like coming towards my direction. then, and I'm not sure if it, if it, if it's accurate to put it this way, but it helped me to save a lot of time socializing because

Oftentimes new people that I meet mostly came from LinkedIn, came from Twitter, came from YouTube, came from my Discord channel or Xiao Hong Shu anywhere. there's pretty high density of, you know, like people with shared values.

So I don't post complaints. don't know. Maybe if I post complaints, I'll meet different friends. yeah.

Rui (36:07)
Interesting. You attract people who respond to the message that you're trying to put out there.

what do you wish that you would say a little bit more on your platform?

Sean (36:23)
yeah. I want to do more tutorials and demos actually, because I realized that that was really helpful for a lot of people and I enjoyed it. I enjoyed teaching things actually. I enjoy sharing the new cool tech that I learned. It's like when you were a kid, you had this really cool toy and then you want to show people, hey, this is a cool toy. This is how you play it.

And when there's a new tech, I share it. And then there are people like really find it helpful. And that gave me a ton of incentives or motivation to continue to do that. I really enjoyed it, to be honest. And I know like I shared more of my thinking process back then. Until recently, I share more tutorials and tech demos. But I feel like this is a kind of a pivot of me.

in social, in my contents. And I want to do more of that because I actually feel a very strong sense of enjoyment of giving

I just find that super rewarding.

Rui (37:33)
fascinating. And the one thing that you did with your social media content that I really liked and also respect is how you just record the videos in a very, you know, ⁓ natural space, either in a random like booth or on the street. And then you just sit there and talk for like 10 minutes or 20 minutes or 30 minutes. Maybe you have some like talking bullet points. But how did you get so comfortable just like showing up?

Sean (37:51)
Yeah.

No, you just keep doing it, right? When I watched my first videos, it was super awkward. And it just gets less and less awkward. And they get less and less editing, to be honest. Like, I really care about me saying ⁓ and then I spend hours cutting the ⁓ And then later I realized that who gives a shit? I don't care anymore. Let me just say the And that...

Rui (38:23)
You

Sean (38:27)
actually gave me less pressure because you know when you're looking at the camera and you're outside there people like passing by and some people are staring at you you have this pressure and if I'm still worrying about myself saying these these filler words I can't say anything so I decided that okay if I'm talking about something if I'm showing some some demo I just talk whatever I want to talk and then I won't edit I just let it go and then later you realize that you get better and better and better so

That's how it is. ⁓

Rui (38:59)
you know, when I'm doing editing, I realized that filler words are actually necessary. especially if you're talking a fast pace. I think you're

your talking pace is actually very reasonable, but if you're talking a faster pace and if you don't have any filler word, I feel like the information density is way too high. You almost need to give the listener and yourself a break in that gaps because not everybody paying attention all the time.

Sean (39:15)
Interesting.

Rui (39:30)
There is a level of information transmission density I think everybody, at least on average, of people can take. ⁓

speeding up that information, you have the illusion of having more information. But at the same time, how much your brain actually processes through, I'm actually not so sure.

Sean (39:52)
you remind me of something that I read on Twitter recently, which is some people complain that

A lot more tweets these days sound like they were generated by AI because people want to be perfect and it's possible to be perfect when you post something in text. And then there's this guy I really like, his name is ⁓ Levels. He's Twitter kind of Levels IO. He's probably the biggest indie hacker on the internet. He's Dutch and he said that I just prefer to talk my way with my own way of speaking English. I don't use AI because it feels more authentic.

On the one hand, as you said, it helps you to decrease the information density when you speak very fast so that people can find it easier to follow. On the other hand, it makes you more real because nobody can talk without a script right in front of you and talk perfectly like the US president, not the current one, the previous ones. It's impossible. ⁓

I feel like it makes you more human when you speak with filler words. And also I'm using, we're both using English as a second language,

Rui (40:59)
I'm still making English grammar mistakes, Sometimes, she and he are still a blur in my head just because, you know, in Chinese, when we talk about she or he, the pronunciation is the exact same one. So, sometimes I feel so embarrassed. I feel that...

Sean (41:01)
Same here, yeah.

Rui (41:18)
You know what, after spending 10 years in the United States, and I'm still making such rudimentary mistakes, and I feel so embarrassed about myself, and that's the part sometimes I'm really trying to edit out. not doing editing is a form, it requires a lot of courage, because it's truly showing exactly who you are, all the flaws laid out in front of like, in today's internet,

almost like naked for thousands and thousands of people. And then anybody can attack you, whether they understand you or not. that's why what you are doing, just unfiltered in an extremely simple environment, sometimes even like a little pressured one because you're just like on the street sometimes. ⁓ That's actually, to me, takes a lot of courage.

Sean (42:00)
Ha ha. Yep.

Thanks.

Rui (42:06)
So I really respect you for that and I actually am curious how do you really handle criticism Because I'm sure you also get some of that too

Sean (42:13)
Yeah, for sure, A few ways. I usually read through it. If it's actually making sense and I'm actually wrong, I recognize that. I say, you're right. And I think there's nothing to be ashamed of that because I can make mistakes. And if somebody pointed out, great, ⁓ I'll take it. you pointed out a mistake that I made, so you were helping me. Fine.

If there's someone who's like making personal attack, I attack back. I don't even hesitate. Yeah. I treat this as a game actually. So if somebody's going to offend you right in front of you and then you take it, I think it's dumb. And I know at the same time that I'm not going to be, you know, like so extreme with this person.

Rui (42:49)
Mm.

Sean (43:04)
So I attack back, I see if this person punches me back. I'm just playing the game and see how it goes. Right? It's like, yeah, it's fun. ⁓ I take it this way to be honest. sometimes people just have very different opinions and they're not telling you why you're wrong, but then they're telling you that you're wrong without any reasoning. And with these people, just thumbs up.

Okay, thumbs up. No words, thumbs up. That's it. Why do I thumbs up? Because replying to them increases my engagement activities. It shows to the algorithm that I'm engaging with the community, so more people are going to see it. So I'm helping myself, right? That's it.

That's it. Yeah.

Rui (43:53)
I love it. Spend minimum time to, you know, obsessing and feeling anxious about other people's thoughts and comments. And a lot of times, especially in some short video, people can misinterpret you very fast, right? Because there's a lot of context that we're not saying. And there's a lot of context that's like already projected in the audience head. And they're not really seeing what you're trying to say, like

Sean (44:09)
Very true.

Rui (44:20)
in the way that you want them to see. So I think there's just a lot of room for miscommunication. I mean, it's already difficult enough at work, let alone trying to communicate a message so well in a, I don't know, minute time window and then somebody can actually get what you're saying and respond in a very reasonable way. So I think there's a lot of expectation about Like how do we really expect people to walk away from it?

Thank you so much for your time. This is really fun to just like catch up with you and get your latest thoughts.

Sean (44:49)
Yeah, likewise.