Stream of Conciousness
Motivation
It’s officially time that I start this thing that I have been thinking about for so long now, a blog. In the past few years, it’s become obvious to me and many other people that building a following is going to become increasingly valuable as time goes on. Because of this, rather than starting a shitty TikTok channel or a brainless and faceless Youtube channel like others suggested, I figured that to start, a blog would make the most sense. I hope that this blog will give me an outlet to improve upon my writing, while connecting with blog readers (a community I believe to be much more interesting than tik tok watchers). With this blog idea lingering for the past year or so, you may ask yourself why hasn’t he started earlier, and the truth to that question lies in the fact that I am lazy and fearful of putting skin in the game.
Anyways back to the value of building a following: I think that I could give a list of ten justifying reasons to support this idea that followings will become much more valuable in the future, however in hopes of keeping things short I’ll only mention a few. Firstly and most simply, I think that we have already seen this happen over the course of my lifetime and most obviously to me in the last 10 or so years. Things like streamers and influencers had some early adopters before this 10 year window, however the sheer increase from then to now is something that everyone has felt. Along with this, the amount of money that these creators are generating in total has gone up by quite a lot.
In layman’s terms, more people are consuming content, thus more people are creating content, and thus more value is being generated by these content creators.
Therefore, it follows from this trend over the past decade or so that a similar trend will continue and creators will only become more prevalent.
Another reason that followings will be more and more valuable as time goes on is because of technology and the rate at which it is improving. As will become clear in the posts I make on here later, I am an AI doomer of sorts, and more generally a doomer as a whole, but AI and other technology is affecting society quite a bit. Because of this effect now and the hypothesized effect soon to come, people have proposed things like UBI that aim to level the playing field for humans and their quality of life as they begin to be substituted out of “the loop”. If such a thing as UBI were to happen, or if humans come out of the loop in some large way, the only real avenue of meaning that will be left is to please other humans, with the most sensical way to do that being content creation. In this case, the demand for content creation rises (rather marginally I believe) however the supply of content creators rises more drastically therefore making it much more valuable to have a following as the size of your following relative to others has gotten bigger. Now I know I know, this dystopian fantasy seems specific and super far fetched, but I also could see many adjacent cases in which a similar thing happens, and the value of followings also increases.
Lastly, I think it has to be said that as creators become more and more plentiful, the amount of garbage “brainrot” content becomes more plentiful as well. Because of this, I think that building meaningful communities of non brainrot interaction comes at such a premium, even if the monetary or numerical compensation isn’t as great.
Anyways, enough for my spiel about followings and why I want to blog, let’s get into some more interesting stuff, like a review and interpretation of some interesting things that I saw today and some housekeeping stuff about my personal life.
Cool Things and My Interpretations
In this section, I hope to present some interesting things that I have come across over the period of writing this post as well as exploring what each of these “things” made me think about and my interpretations of them. I hope to do this on a consistent basis so that eventually, I can get a feel for my content consumption habits on a larger scale than what I see and remember on a day to day basis.
Cluely and Roy Lee:
Roy Lee, where do I even begin. The keen readers of this post will note that a portion of my “why followers are important” spiel was heavily inspired by some of the things that roy lee spoke about on TBPN a few days ago. He said that “it’s just so easy to go viral these days” because the supply just doesn’t match the demand ever since people started going crazy for short form content. While I definitely agree with him a little bit, I also think that it’s clear that he has a knack for this whole virality thing. Maybe its god given talent, maybe its hard work, but whatever it is, he knows what the fuck he’s doing.
This past week or so, some major developments have happened in the whole Roy Lee saga. First and foremost, the guy raised $15m from arguably the most prestigious VC fund in the valley (a16z). Secondly, he seems to be going more viral than ever, mostly capturing the momentum from this insane seed round. His announcement video had maybe the best balance of actual substance, and shock value/ragebait. I mean in my own head I am trying to decide whether or not this kid is a Mark Zuckerberg type, or if the “will Roy Lee serve jail time in 2025” market on Polymarket is the key to my financial freedom. Ive seen so many comparison clips of Roy’s interview style and young Zucks interview style, but if I had to make a prediction, I would say that in the end his business will fizzle out although I think he will be able to hold the spotlight for a long time, and potentially think of a more interesting and long term business (no offense to Cluely).
Another interesting development in the Roy Lee story arc, is the large group of people I see on twitter, trying to gotcha him. It’s important to note that while I say that they are trying, a large majority of them are succeeding, like in the video attached to this section, where @aryamantitan finds some way of jailbreaking the Cluely assistant totally discreetly. These gotchas as a whole serve as a method of proving my point that while Roy certainly is an interesting founder, what he’s founding isn’t necessarily my cup of tea (that is, I dont think it’s worth shit). Another Interesting gocha moment that happened in this past week or so was the leak of the Cluely system prompt, which is quite interesting seeing as the system prompt is worth some large percentage of the $15m he raised. However, despite my uncertainty in the long term success of Cluely, Roy Lee used this moment to produce possibly the most underrated tweet that I have ever seen on my timeline. As of right now this tweet has ~90 likes. For context, this post came in the comments of a thread exposing the system prompt, where Roy and some onlookers explained to the discoverer of the system prompt, that realistically, any company whose system prompt can be leaked, has had it leaked. Both the distributors of these models and the companies using them all face the same issue. Here Roy is explaining some small design choice about how he stored and retrieved his system prompt (I believe he hard coded it into the client or something like that).
That being said, this tweet just blows my mind because it shows just how sophisticated Roy Lee’s self proclaimed “rizz” is, but also just how gigabrain his marketing strategies are. In all seriousness this could’ve been a totally unplanned development internally at Cluely, but the fact that they see it as a positive way of “marketing” their product really makes me think that this shock value/brainrot marketing is gonna become the new meta.
I could go on and on about Roy Lee, but for now, I think that’s probably enough (probably until my next post).
TBPN
Next thing I want to cover is TBPN which has completely changed my content consumption habits over the past 1 week. I always have some podcast that I deem to be the meta, but TBPN really blows them all out of the water. I mean every little detail about their production is perfectly done in my opinion. Full transparency, I actually think that both of the hosts are pretty mid, but the amount of information that you get in such a small time makes it my new go-to podcast, beating out some past legends like My First Million, The Lemonade Stand, How I Built This, NPR pods, etc. The thing that I like most about the podcast is that everything is very fast paced, and somehow, because of this they seem to cover all of the elements of a podcast that I previously had to go from source to source to get. They cover the heavy hitting daily news and recent developments in the startup world. They meme a bit and make jokes throughout. But yet they still seem to cover the quintessential part of podcasting that I love the most, where they bring on guests with world class topic lists to pick their brains about the things that their guests would know better than anyone. With all of that, I also think that they reach the perfect amount of depth with each of their guests where they don’t go too into the weeds, but they pick the guests brains just enough to get the viral stories right out of them.
Enough with the glaze session but if anyone who reads this needs a new podcast to check out and is interested in big tech/startups/entrepreneurship, I think this is your endgame podcast. *No affiliation to TBPN btw but holy fuck would that be a sick goal to strive for as this blog improves.
Tesla and Robotaxi vs Waymo & Geohot
Next on my topic list is a little discussion of the tesla robotaxi rollout in Austin this past week. From what I’ve heard on twitter it seems like they killed it. Don’t get me wrong, Twitter is probably the most biased source of Elon related news, but seeing videos and live streams from the event, my mind sort of got blown. I think through it all, the only limiting factor about this launch that prevents me from yoloing my networth into tesla calls is the fact that there was a person sitting in the front seat (although it was unclear what their real purpose was) and the fact that the area of service was so small even compared to waymo. The person in the front seat thing shouldn’t be a long term issue for the scalability of the robotaxis, as long as there is no foul play with human in the loop developments, like what went down at the humanoid robot event in Silicon Valley earlier this year (they were totally human controlled). Along with this, the small area of service begs a similar question for me, where I am unsure how manicured the actual software was to handle that small sample size of roads rather than using a more general computer vision model. Only time (and leaks) will tell with these issues but don’t get me wrong if these aren’t too bad, I think that Uber and Waymo are in for a tough next few years battling Tesla in the scale wars. I Also think that if those potential issues are not catastrophic that this event will be the beginning of the answer in terms of the Lidar vs Computer Vision debate.
This analysis of the Tesla robotaxi soft launch leads me into my next big topic which is Geohot (George Hotz) and his segment on the TBPN podcast network this past week. Geohot has long been my favorite person in the tech world because of his give no fucks attitude and contrarian opinions that I often find to be more logical predictions than that of the crowds. This week, he said a few things that really itched my brain. The first, and most applicable to the Tesla vs. Waymo’s argument was that Waymo actually has ~1.2 humans per car remotely controlling the actions of this car. This totally changed my robotaxi/driverless car worldview because for so long I was under the impression that driverless cars almost “just worked”, however it’s now clear to me that we’re a bit farther from the solution than I thought we were. This also leads me to believe with even more conviction that Elon has always known what we are doing with the robotaxis and the whole game is really his for the taking. I think this because firstly, the tesla robotaxis we saw this past week cost ~40k so it seems, and from what people in the cars saw, they had no alterations or cameras attached to them, that is, they were simply using FSD. However, these Waymos that have ~1.2 people in the loop per car (MORE THAN AN UBER) cost closer to 250k and do the same thing. Because of this, when we talk about scalability, even if Tesla is also remotely controlling their vehicles with a human in the loop or even if there is a major accident in the news, I still think that no one else has a chance.
It’s important to note here that while I haven’t done my due diligence on any of these numbers that George Hotz cited, I definitely put a lot of trust in him because of his track record. For example, in this appearance he had on TBPN, they asked him “How are things at comma going” and to that he responded quite happily, “We are ~2 years behind Tesla”. I mean not only the honesty is impressive but the awareness to say such a thing so frankly just blows my mind. On a similar note, he also expressed his distaste in the whole “protect your company so that you can win” idea by encouraging anyone to challenge him. I mean he just knows that what he is doing is special and that really no one can do it but him. All of this together just makes him such a role model for me because he has such confidence in himself and yet with that confidence he also brings such honesty and novelty to this world of egotistical fucks looking to raise the next $1b rather than creating something actually good.
The next big thing that Geohot mentioned that was one of my biggest takeaways from this week of content consumption was the AI hype hierarchy. He first dove into the bullshittery that comes with words like AGI and agents which are constantly regarded as the future, with people saying “AGI NEXT YEAR” or “2025, the year of agents”. “What’s an agent” George repeated, “what’s AGI”. These questions that really no one has a good answer to should be the starting point of our understanding of AI, not the end. After this subtle rant about the ambiguity of these words, he got into the landscape of AI startups and companies. He broke it down into I believe 4 tiers. Tier 4 are the AI wrappers as people call them. These companies may be doing some subtle fine tuning or may be creative in the way that they prompt these models, but in reality, their success is 100% dependent on the functionality of the models, and the decisions made by the companies who make them (excluding open source for now). This specific realization was my favorite of the entire week. Now, when I think about companies like that of our favorite viral machine Roy Lee, I now realize that if OpenAI or Anthropic or Google or really any other significant private or open source player in the LLM game wanted to make something with similar functionality to Cluely (maybe because they realize that people desire such a thing), it makes his company obsolete. This moat issue can be generalized to so many more companies and really changes the way that I think about valuing companies focused on AI. The next tier George mentioned was tier 3, in this tier, he noted companies like OpenAI and Anthropic, the companies who provide these models. Interestingly, he reasoned that in the future, even these companies may lack a moat in his opinion. Next was tier 2 and 1, with tier 2 including companies like nvidia that make the hardware, and tier 1 being more low level things like datacenters and fabs (if I understood correctly). Anyways, without getting into the weeds about all of these tiers and what I think about them, I think that specifically, this thought experiment regarding the tier 4 companies reframed a lot of my prior beliefs that lead me to believe that certain companies were very valuable.
For my final “cool thing” I’d like to include the best ad that I saw during this blog creation period. This ad is from Google, where they use Veo 3 to create real life videos displaying comedic interpretations of classic fairy tales. I’m not gonna get into it just for time’s sake, but I think that if I had to invest in one company right now, it would be google. I realize that a lot of these things I deem valuable about Google are probably priced in, but despite this I just can’t see a future where Google doesn’t “win” AI. Their moat is just too big.
Anyways, fellow reader, please unwind and enjoy this retarded ad that somehow made me laugh. Housekeeping
For the last section I want to include in my daily or weekly or whatever the fuck blogs, I have some housekeeping regarding my life. Here I want to give some basic updates of what I am working on at the moment, what I am learning, how I’m doing athletically, and anything else that I may want to explore in the future. Feel free to steal any of my ideas, or shit on them and call me stupid. The main goal of this section is also for me to be able to look back at these posts over a long period and get a high level view of any evolutions that I went through.
Firstly, I have spent most of today and yesterday trying to figure out how to construct a secure multi platform app using react native, node, express, and postgres. With this, the main challenge that I am having is how to balance my use of AI (cursor) and actually putting head to book and understanding what I am writing. I feel like there is such a balancing act going on, and at times, I struggle to understand whether or not I am on the ground or still standing, because it’s so damn easy to get a “working” version of what I want without understanding why exactly certain lines of code are the way they are. Anyways, to manage this, I am trying to spend more time in the docs of express and react native, in hopes of setting up a sufficient base before I hop into vibe coding, but I just can’t tell whether I’m doing enough. Through this process, the only thing that’s somewhat reassuring to me is the fact that if I just keep doing projects similar to this one, no matter how much vibe coding I do, eventually I will get to where I want to be knowledge wise. Right? On the same note I doubt that theory because of a study recently read that I will link below. Anyways.
Before I move on from my discussion of this balancing act, I want to share one idea that I had to solve this issue. Ready? A chatbot that lets you: 1. Ask questions to the docs of a certain thing (like react native) and 2. Work through the docs at a faster rate with an integrated cursor type interface that will explain whatever you need to be explained. Realistically, this definitely exists or can just be made for myself with some basic tinkering.
Next I want to talk about a project that I believe would be a great project for someone trying to break into ML/AI or SWE fields (like myself). So about 2 months ago I started playing a game that I loved in my childhood called clash royale. I didn’t think much of it, but over time I felt the same relationship between me and this game developing as the relationship that I had with chess a few years prior. That relationship really being composed of the fact that I would watch videos about it studying the game to try to get better, and in both cases I would find myself spending 1-3 hours a day almost in this super locked in state to try to gain elo and prove that I was actually good at the game. Eventually I gave it up just because how childish and wasteful the whole thing felt, but I realized one interesting thing. I realized that this game didn’t have cheaters the same way that chess did. This took me down a bit of a rabbit hole that I am still in, where I thought a bit about why this game didn’t have cheaters. Was it the fact that the short term decisions that the player has to make makes the game too hard to cheat, or was it the sheer number of characters that one can play. I am still working out the answers to this, but I thought about it for a bit and came to the conclusion that at some point during this summer I was going to build a Clash Royale Engine, to cheat my way to Ultimate Champion (better rank than I have ever been). To do this the real main issue that I will have to solve is a data issue. In chess each position can be described with some concise notation, making it easier to construct and train models to learn how to play the game. However in Clash Royale, such a notation does not exist, and even if it did, you would need thousands of thousands of entries each game. Hopefully by next post Ill have an update about my progress, but I challenge anyone to make an engine quicker/better than the one that I am trying to make. Here is a reddit post that I found about someone with a similar idea who actually ended up making an engine (although it’s super shit).
Clash Royale Engine
In terms of what I am learning right now (past week), the main thing has really just been this app development stuff, although I have spent 20+ hours listening to TBPN as I mentioned before so hopefully that counts for some type of learning (definitely not).
Gym wise I really have been a little off for about the past month or two now. I haven’t hit my legs in decades and really am relying on playing sports for cardio. I am eager to get back onto the tennis and basketball courts in the near future. However, all of this being said, I have still been alright about hitting my protein goals each day although I have definitely been getting a bit fatter because I’m eating 2.5k+ calories pretty consistently. I just recently started tracking calories on myfitnesspal again so hopefully I see the effects of that in the coming months. Finally, I have been consistent with weight training about 4 times per week although the level of intensity I have in the gym is not quite the same as what I had in the winter and early spring.
If you really read this whole thing, I owe you a big hug. Be ready for much more of this to come because as I keep hammering out these blogs they will only get better and better (I realize this is probably rather shit). Anyways see you next time.