Matt Shumer on his 50M-view AI essay: 'I originally wrote this for my parents'

Let me tell you about CrowdStrike first. Your business is AI. Their business is securing it. Crowd Strike secures AI and stops breaches. And without further ado, we have Matt Schumer in the recent waiting room. Welcome to the show, Matt.

The Schuminator,

the man of the hour. Welcome to the show. Thank you so much for taking the time.

The most viral essay ever, maybe. Has there ever been a more viral?

Would you have won the million dollars if you posted this a week ago?

I think so.

I don't know, but I do wish I posted this a week ago.

Well, fantastic response. What has the response been like to the essay from non- tech people? I imagine that you have friends, family friends that have been texting you screenshots or something like walk me through the experience of this.

It's been crazy. Uh I I wrote this with that in mind, but still I didn't expect it to actually happen. You know what I mean, right? I I originally wrote this for my parents. I was trying to explain to them what was going on while I was actually home for Super Bowl Sunday with them, some family, friends. And frankly, I couldn't find anything that was both clear and accurate, but also understandable for them.

Right? I was looking at Dario Mod's essays, and when you read them, they're perfect. They're amazing. But you have to be in tech to understand them. And I just wanted to find a way to communicate this in a way that allowed them to actually understand it without having to know all the jargon, the industry dynamics, everything. because I think it's important for people to know this, but there's just not much out there for the person that's not in tech. Um, so yeah, it did take off. It's been surprising and exciting. Um, frankly, I've been getting texts from friends who I haven't heard from in years, and they're like, "My boss is sending this around the office. It's it's it's surreal. I thought it would go a little viral. I didn't think it would go mega viral, and certainly not this level." Um, so I'm kind of taking it all in. I've slept like two whole hours since I posted this. I'm exhausted, but it's been it's been crazy and I'm glad that the message was received. I think that's the most important part, right? People did not happen the way I'm expecting, but it's good that people know that there's a chance

and the message is just something big is happening. You just want people to pay attention because I think people are going to debate like how how you presented different things. Is the COVID comparison really that accurate? But, uh, is that is that specifically you're just trying to wake people up? Exactly right. I if I knew how viral this was going to go, I think I would have spent more time thinking through some of the parts and how I presented them. But

is part of is part of why you why you you're maybe sort of worried or you want people to pay attention is is your experience as as an entrepreneur here because it feels like in many ways hyp hyperight like you know uh probably made more sense as a product two years ago and as all the different LLM have gotten better and the products have gotten better I imagine uh it has been great uh and sounds like maybe you're you're moving on to focus on investing But uh is part of this is it's been kind of very personal experience of just kind of waking up to the to the power of the the base models and and and the labs in general.

It has and I I'll share more on the company side soon. Um but I've been in this industry since 2019. I got in when I was in college working on a VR startup and I realized that like if I didn't drop everything to jump into AI, I'd regret it for the rest of my life. And I'm glad I did. Um but seeing the progression is is kind of freaky. I mean, you know, it couldn't write a sentence accurately in 2019, much less a line of code.

Now, we're looking at models that can build autonomously for many hours. I had early access to GPD 5.3 codecs. And that was a very eye openening experience getting used to that. At first, I was prompting and using it like I use any other model, little things at a time, iterating back and forth until I realized, wait, this isn't screwing up. I started pushing it forward and pushing it further and just saying, "Hey, here's the big spec I want. Don't come back to me until it's perfect and get it deployed on a server and it it did."

Yeah.

Um, and I'm not saying every industry is going to go just like that. I don't think it's going to look like that in every industry. But if that's a sign of things to come, I think people should be paying attention.

Yeah. Yeah. I mean the the VR example is so interesting because like completely agree with you on just the power of the models like that it's it's so real. It's it's so legitimate and it does and truly a lot of people are like oh it still hallucinates or six fingers in the images. It's like that that that is antiquated like you need to get up to speed. But at the same time like if if I was you know we see Mark Zuckerberg like cutting back on on virtual reality spending and plans there like and it doesn't and it feels like if AI is like going to advance everything you'd be like great like I I can invest 10 times less in reality labs and the headsets are going to get 10 times better. But there's a lot of these things in like the they're not even it's not even physical world. I'm not even talking about doctors. I'm talking about VR headsets. But even that doesn't feel like maybe like AI doesn't really pull that forward. And I'm just wondering like when I go to my friends and family, how many of them will actually be like, yes, this is going to change my life in whatever I do versus like, yeah, software will get better, but software is a small piece of what I do. But how how are you thinking about about people that where software is like a more minor piece of their world? Yeah, this is something that again, hindsight is 2020. I wish I had spent more time on knowing what was to come.

I try to address it a little bit in the article and I'll hopefully provide more color here. Um

my dad's a lawyer as an example.

Yeah.

Like it's not going to stand in court uh anytime soon. Yeah.

Right.

It can help with writing. It can help with reviewing things that he's he's working on. It's essentially an associate and it's getting better every year.

Yeah.

But it's not going to stand in court. I think the ideal thing here is people read this and they kind of take it in their own unique way for their own unique needs um for what they're doing right if you're somebody who's very very exposed and I think there are a lot of people who are who kind of are denying it

I think it's worth paying attention and saying okay what is the world going to look like in two three years how do I prepare myself because it's it's worth it just in case there are industries where I think it will take longer though even in those industries for example let's say legal those that are just getting started today might have a bit of a rough time whereas a partner might do quite well actually.

So, I think it's very contextual for each person and that's what I wish I kind of got across a bit better. Um,

maybe I'll do a follow or something like that. But

yeah,

yeah, I I I I think it's really tricky to give sort of like a silver bullet one-sizefits-all answer for everybody. Totally.

The goal is to just say, hey, look, this is happening. Can you try to think of it about it a little bit for your industry? Because even if it's going to change an industry and there's like like the ability for it to happen, it's not going to happen overnight even if technology is there. Though I do think for a lot of these industries the technology will be there relatively soon. It's just a matter of how long it takes to proliferate proliferate through society.

Yeah. Yeah.

Lawyers don't tend to post as much as software engineers.

Uh true. But you can imagine

lawyers out there thinking I don't even write contracts anymore. I just review. Which is like kind of what you're seeing.

Yeah. most engineers saying,

but at the same time, so so part of the I I will I'll be interested to see what happens with like the general legal job market,

but we've just been hearing examples from companies being like, I need to hire hundreds of interns that are just going to come and work

in a new way and and use the tools sort of natively. And you can imagine that happening, but but again, there's just

I don't know.

Yeah. what what advice uh or what what recommendation would you make to a new grad right now?

It's really tough and it's one of those areas in the article that I don't think I had a clear answer on but that's because there is no clear answer. I have a lot of people in my life that have asked me this and part of my thinking on this article was like

can I answer some of these questions and there are parts of this that got clear parts that didn't. This is one of the parts I don't think I have full clarity on. I think knowing the tools is better than not knowing the tools.

Yeah.

Focusing on industries that maybe will take a little bit longer to be disrupted so you can entrench yourself a bit first is valuable and there are people better than I at knowing which ones those might be.

Um but really I think it's just getting used to using this because this is the new world. It's like you know if a calculator is is available you should probably be learning how to use it. This is this is that times a thousand.

Yeah.

Um so I think it's just learning to be adaptable. I think there probably is. We'll look back in 10 years and there will be clearly like this is what you should have done. Totally. I don't think anybody knows today.

Yeah.

But it's also been very contextual. It's going to depend on what your goals are as a person.

Yeah. I mean obviously like there's those posts about like it's the last moment to join a lab before the takeoff and like it'll be fun to join and and that makes a ton of sense. At the same time, I I love the idea of being a young person becoming a a hacker, like lightly technical, but really fluid in the tools, and then going and working at like an oil and gas company and just being like the most cracked, most productive person at this company that has like assets, and you're going to be able to just do so much more across the organization and and be that like, you know, thousandx intern basically. Uh Jordy, where do you want to go?

A lot of uh comments from people asking if AI wrote the essay. Oh, yeah. What's your what's your writing process actually like these days?

It didn't stick out to me as AI, but

you definitely I didn't you probably if you did use AI, you pulled out a lot of the M dashes.

Yeah.

So,

it did help a lot and I think that's kind of point a lot of people are like dunking on me for saying that and I posted that I I used AI.

Yeah. It's not really a gotcha if if if it

did insane numbers like whatever whatever whatever the like if you're upfront about it, it's fine. Yeah.

It's like proving the point. It's like people are saying this thing sucks at X Y and Z, but like look, it's 50 million views. Like there's a reason that I'm talking about this stuff.

Um, and you know, I think there's like little details I get wrong here because this has been a crazy couple of days. But

the way that I approached it was again, I wanted to write sort of like a Daario style essay, but for the average person. So what I started doing is I have like lists of like all the things that have like sort of made me think over the years that I agree with and I've sort of points on what I agree with in each of those articles, what I disagree with, what I think has well and what hasn't. And I've kind of dumped this all in and I used Claude for this. Even though I think Codeex is way better for engineering, Claude is better for um this sort of stuff.

I dumped all those articles in and I said, "Okay, here's what I agree with. Here's what I don't." Um, and then I basically spoke to it for like an hour with my own thoughts and my own feelings and how I wanted to present it and what I think actually makes sense because a lot of these things past kind of are somewhat accurate, somewhat not. And I iterated with the AI for for many many hours going back and forth until I finally felt like, okay, it's time to write. And I kind of had to put out a whole sheet on how to do this.

And then I actually went and I uh took that and wrote the first draft myself. And then there were obviously parts I struggled with. I had the AI come and help.

Um, I hel I had it kind of like work on the wording with me. And then once I had a first draft, I brought it back and I had it critiqued it. Um, and that's kind of there's there's obviously more details than

I mean I I I think when people lob the like this is AI, they think that you that your prompt was like write a essay that will get 50 million views and and it just like oneshots it. It's like no, this is a collaborative back and forth. Yeah, of course. Um Yeah. Yeah.

We tried that on X.

Yeah.

There's a lot of hashtags.

Yeah. Yeah. Oh yeah. It can be can be a little rough.

Um yeah. Yeah. Fascinating. Um Jordy, anything else?

Uh no. What are you What are you What are you going to do next?

Yeah. Yeah. I mean, I I've noticed from your portfolio that there's a lot of uh chip companies, hard tech investments, like what are you most excited about just in the economy broadly?

I mean, in terms of what I'm investing in, and I intentionally tried to leave this out of the article because I wanted the article to be just purely my message, not sort of like kind of like have to deal with the sort of interplay of that. But in terms of what I actually believe in, I think the chip companies are very interesting. Um, I was, you know, I don't know if you saw a couple years ago, I kind of helped Grock blow up and that that's worked out quite well. Um, I'm in Etched a few others like that. I also think the rails are going to be very important. Um, I tend to invest in, you know, I just started my fund around this thesis that there are going to be sort of these things that AI needs to proliferate, right? not just the the model itself and the weights itself, but it's the

uh you know the rails for them to communicate with each other. And I just invested in a new company uh agent relay that's about to come out that I think is going to be really exciting there. Um I also did the agent mail so very very similar. Uh Daytona

um

SF Computes a little different, but I think they're going to be Evan is a killer. Um, I'm sort of interested in that sort of stuff. Like how do you take the models and then bring them out of their boxes because this is obviously happening and you just kind of want to put your money where where uh it's going to grow. Uh, obviously.

Yeah. It feels like there's whenever something crazy happens, there's a lot of like h it's almost like nerve-wracking panic, but I I do think we're in this moment right now where there's going to be a re-evaluation of how value acres in tech because tech has become everything. Like, you know, car companies are tech companies now. Everything's a tech company. Uh and but the business models are wildly different and you know you know the the software agents are going to accelerate some companies. They're going to damage some companies and uh we're we're going to be following it all right here. But we appreciate you taking the time to come chat with us.

Yeah. Get some sleep.

Congratulations on the massive article. I mean it it it I definitely agree with the point of like we we're not talking about the the latest updates to AI uh broadly enough.

I do think it's interesting some people will read this let's say they're a firefighter or a school teacher.

Yeah.

And be like wait I'm going to have to do less paperwork.

Oh yeah.

There's actually a lot of people that should be

but that's but that's a good thing there. There are some there are some beautiful white pills. There's some there's obviously some quagmires and some quake sand you want to avoid. But this is how technology is is rolled out and adopted and so you want to you want to skate where the puck is. But we appreciate you coming on the show and breaking it down for us. Thanks so much, Matt.

Yep.

We'll talk to you soon. Have a good one. Good to see you.

Let me tell you about Gemini 3 Pro, Google's most intelligent model yet. State-of-the-art reasoning, next level vibe coding, and deep multimodal understanding. And without further ado, we have Vlad Tennv from Robin Hood in the reream waiting room. Let's bring Vlad into the TV Ultra. Vlad, how you doing?

Whoa. Cactus.

Cactus.