Gamification, or “the use of game-like mechanics, incentives, and experiences in non-game contexts to encourage participation and engagement.”

All of us play games we don’t realize we’re playing. The world is filled with scoreboards, winners, losers, rewards, statuses, and a reason to come back and do the same thing tomorrow. We connect to one another through games and make the mundane exciting. If you pause for a second and look around, I promise you will realize that gamification is everywhere, just hiding in plain sight.

When I was building Official, I was interested in gamifying the ways couples can connect deeper with each other. Now, predicting virality is the game I focus on. Numbers are everywhere, from view counts to impression counts to bank accounts, and this overarching idea of “big number go up” has permeated its way into most aspects of our lives.

We’ve talked a lot here about how good marketing brings brands to life. Making the inanimate alive. Building a brand that consumers feel they’re participating in functions the same way gamification does. It’s at its most powerful when anyone can become a player. It brings material that is otherwise static to life, allowing consumers to engage with and participate in information or systems that would typically be separate entities.

Loyalty, Loyalty, Loyalty

Look at airline miles and loyalty systems. We all want to feel like we’re part of a team. None of us wants to feel like a traitor. I’m a 1K member on United and an Ambassador on Marriott. I don’t want my own little game to end. I’ve turned down objectively better options to stay at a Marriott because it kept my streak and the game alive. Walking into a Hilton would feel like wiping my own track record. The chance of winning an upgrade is a slot machine of its own kind.

Most gamification doesn’t feel like a game while you’re inside it. It just feels like your life.

The NYT

The fact that the New York Times has become just as much of a gaming company as a reporting company now is pretty ridiculous. Most engagement used to come from people reading long-form news, but now, the century-old newspaper is a major hub for daily crosswords and word games. In 2022, they bought Wordle for seven figures and turned it into a reliable driver of daily engagement. Now, these games have essentially saved the Times, keeping a range of audiences subscribed.

Duolingo

Duolingo is getting people addicted to doing homework. They’ve nailed both incentives and world-building. There are streaks, leaderboards with strangers, and, of course, the iconic mascot, Duo, who quite literally brings the brand to life and guilt-trips/threatens users to keep their streak. I don’t know anyone who has actually become fluent in a language through Duolingo, but most users have accepted that it's more about the process of learning than about mastery.

Hinge

Dating apps are also slot machines. We could laugh about that fact, but it’s completely true. Hinge is an app designed around uncertainty. Who knows who you’re going to see next? It could be the love of your life or the most beautiful person to ever bless your screen. That possibility is always there, the same way that when you’re gambling, the possibility of a jackpot is always looming, even if there are algorithms in place to make the odds statistically low.

If you have 100 people in your potential match pool, the Hinge algorithm will show you the bottom 90 first, in whatever order keeps you swiping the longest. Those top 10 essentially get sprinkled out in intermittent rations, one at a time, and once a day, the best match.

It’s a misconception that winning is the addictive part, or that a Hinge match itself is when the dopamine rush occurs. The dopamine hits when the algorithm shows you that intermittent top 10, and for one moment, envision your whole lives together. With actual slot machines, your brain lights up the most when you pull the lever and don’t know what’s coming.

Understanding how our brains function is important because it makes us less vulnerable to manipulation. It empowers you to use gamification in ways that improve your life, so you can be self-aware if it ever feels beyond your control.

Why Do We Play Games?

Humans have never had a defined endgame or finish line. Nobody takes the frequent-flyer miles, the money, the Duolingo streak, or the follower count with them. A scoreboard offers some way to measure a life that otherwise has no unit of measurement. We all measure our lives in different things, some in crypto and some by learning Italian.

That’s how we’ve always made progress legible to ourselves, retreating back to marks on the wall of a cave, long before anyone was selling anyone anything.

AI News

-OpenAI just rolled out ChatGPT for Teens. This is basically a regulated version with a “study mode” that walks kids through the problem-solving process. The idea of teenagers staying on a safe mode of anything seems unrealistic to me, but its an attempt at responding to a string of lawsuits and regulations closing in.

-Stripe bought OpenRouter for $7.5B. Stripe’s founders Patrick and John Collison wrote in a letter to their investors that Jan 1 marked “the beginning of the singularity.” This was a joke, even if people didn’t take it as one. Obviously we are not there. Yet.

It’s more like AI is simply reshaping the entire economy in real time, and Stripe wants to be at the center of that shift. Acquiring OpenRouter gives them foothold in tracking / managing token spend too.

Weekly Shoutout

As we mentioned, gaming is a meaningful way for us to connect with one another.

While Sarah catches up with her friends on a walk, I catch up with mine over a game of Fortnite. This Fortnite ritual has kept my friendships intact, giving us something to share together as we chat.

So this week, I want to shout out some of my go-to Fortnite buddies:

Noah Poser. Yasha Brudno. James Hueston. Andrew Chan. Duncan Littlefield. Remi Levins. The list goes on.

These Fortnite sessions connect us across time zones.

Thanks, Fortnite.

That’s all for this week.

Stay golden,

Z.

There is no comparable rush to the dopamine spike when your post crosses the threshold and starts blowing up. It’s no secret that everyone wants to be popular and feel loved. When you go viral, it feels like you are being seen by the world and, in turn, validated on an incomprehensible scale. Yes, filling a stadium of thousands of people is cool, but what about millions of people telling you’re beautiful and hilarious?

If going viral is such a powerful drug, it’s important we make sense of it. We should learn how to harness virality so we can manufacture it and use it like a secret weapon; a tool to bring us closer to our goals. Understanding virality was how I sold my first app at 25, co-founded an entire media agency, and, more recently, went viral on LinkedIn after a dry spell.

I posted an absurd meme on LinkedIn, thinking it would maybe get 5,000 impressions. The team and I were surprised to see it exceeded 75,000. After months of the algorithm feeling like a black box, it started making more sense than ever. Even though the post’s success was unexpected, in hindsight, it had all the right elements.

The meme was of me dumping out an “intern’s” matcha latte and swapping it out for creatine. There were elements of relatability at play here: the trendiness of matcha and the gym-bro stereotype of startups. That alone could have been enough to create buzz, to have some sort of finger on the pulse of what’s hot / what gets people fired up, but then on top of that, it also had a realistic image that caught audiences off guard.

I was pouring out her matcha… all over the office floor. It broke the mold of what people expect on a more buttoned-up platform like LinkedIn. Being able to subvert people’s expectations on a platform while still staying within the constraints of how it works is a really important intersection to understand.

The sweet spot of virality is when you can simultaneously catch people off guard AND meet them exactly where they’re at.

Predicting virality is a muscle, and it’s a muscle fully capable of training. By spending enough time scrolling, your brain becomes a pattern matcher. Within a few seconds your instincts can spot potential virality in other people’s content. No matter what it is, educational or engaging or relatable, the post shows you something you didn’t know you needed. That feeling alone creates momentum, because if it’s happening to you, making you feel something, then you can guarantee that it’s doing that for other people. We’re really not so different at the end of the day.

Despite having spent the majority of my career consumed by the question of virality, much of it still feels like a black box. When it comes to making predictions about your own content, your pattern-matching instinct can fall short. It’s easy to be blindsided by your own content, and it is extremely difficult to be objective about your own creations. You’re very rarely evaluating the post itself from a blank slate. Usually, even subconsciously, your expectations and subjective experiences are all mixed up in the evaluation.

When it comes to predicting virality, there are things you can and can’t expect. Just about anyone can run statistical analyses with Claude these days. You can calculate a confidence interval using sufficient historical data. Yet, I’m more excited by the things that statistics can’t explain. There’s an extra virality coefficient that I started calling “The Tingles.” The Tingles totally transcend data, and they happen when a post escapes the range you could have ever calculated for it and takes on a life of its own; it’s a black swan event. It’s a much more elusive component of chance and human behavior. I think if a model could be built that factors in The Tingles, the entire game of virality would change.

Yes, going viral is a game, and social media as a whole is a game of status. Like any game, you are collecting something. You’re collecting followers, likes, comments, and impressions. As you win the game more and more, your objective changes, but no matter your objective, you’re communicating your truth on a mass scale, for free. It’s a skill that allows you to communicate in a way others can’t.

Virality is the biggest sum of human psychology case studies on the largest scale imaginable. Hundreds of thousands of individual decisions to watch, laugh, share, or scroll past. This entire world, made up of images and stories, is a game of attention. This entire economy is made up of attention. The better you understand attention, the more you can guide people where you want to go.

WEEKLY SHOUTOUT:

This week, I want to shout out my friend from literal day one, Max Zavidow. You may know him as @formerteenheartthrob. We met in pre-school, and reconnected when we were both in the city in 2021. It was a very full-circle reunion.

Max inspires me in many ways, especially in how instinctively he understands storytelling and virality. He clearly understands people deeply, and he makes me laugh harder than most can. Max left investment banking to pursue comedy and TikTok. His content is filled with surprise and plot twists, each one feeling like its own little fever dream. Lately I’ve been loving his ridiculous age-gap couple interview videos.

No matter what the content is, the guy thrives in front of the camera. He’s also an expert with deliberate planning and scriptwriting. He showed me the ways that storytelling is everything. He writes meticulous scripts with a fully fleshed-out story vision. The man plans a video the way most people plan a wedding.

NEWS:

1. Spotify is getting into the agentic coding game, which reinforces my belief that everyone and their mother wants to be a coding platform. They built Xirp, an internal system that lets engineers run dozens of coding agents at once, and jump between Claude Code, Codex, Gemini CLI, and open-source models without losing any context.

2. Mark Zuckerberg published a 6,500-word manifesto laying out Meta’s vision for personal superintelligence. His argument didn’t state anything entirely new. It was about how AI could make individuals and society collectively more powerful if we focused on everyone having access to the same tools. How everyone could eventually get a personal agent that understands their goals, relationships, health, career, and finances, and works for them 24/7.

People tend not to react well to manifestos. I can’t think of a time when anyone has released a manifesto that wasn’t somewhat put under a microscope. TechCrunch said that Zuck glosses over many of the real problems people already have with AI, highlighting the distance between how Silicon Valley talks about the AI future and how everyday people actually feel about it.

3. Gemini just passed 1B monthly users, putting the Gemini app alongside ChatGPT in user count. Google said that 63% of Gemini users are talking directly to the assistant through voice chat, and Gemini is generating more than 150M images every day. The funny thing is that I don’t know anyone who uses Gemini the most. It doesn’t feel like a billion-user product, culturally speaking.

That’s all for this week. If you’re bored this week, read through the Zuck manifesto and let me know what you think, become Gemini’s 1,000,000,001 user. Or start a coding platform.

Stay golden,

Z

One of my favorite books, Yuval Harari’s Sapiens, makes this case better than I can. Stories were how early civilizations passed the time and explained what they couldn’t otherwise explain. Stories also gave people hope. We’re a species with the ability to believe in things that don’t physically exist, like money, borders, brands, religion; shared stories that large groups of strangers can believe in at the same time.

A collection of emotional moments creates a story, and that collection forms the connection we all feel to the bigger picture. Religions have been built upon shared stories for thousands of years, including the Torah, the Quran, the Bible, and more. Across so many traditions, storytelling brings values to life and results in deeper shared understanding.

We have an evolutionary instinct to take a shared reference and channel it into something specific to us, something that couldn’t have come from anywhere else. The ability to turn raw experience into a story no one else could tell is becoming the last thing that can’t be replicated.

With AI allowing anyone to print out a 400-page manuscript in an hour, technology gives us all the same power to produce content; however, the moat is shifting. AI still can’t originate a story out of lived, specific, human experience. If you ask an AI model whether a dead cat in a sealed box is alive or dead, it reaches for the framing of Schrödinger’s cat, because that’s the pattern it’s seen a thousand times, even when the actual logic in front of it says the cat is clearly dead.

Humans will always have an added layer of emotional reasoning and lived context. We transform information into meaning, and meaning is what every culture, religion, and enduring brand is built from. The same mechanism that builds religion is why people worship their favorite brands. If you want to see brand storytelling in its purest form, look at the luxury OGs.

The Luxury OGs

We’ll start with Ferrari. Every Ferrari is presented as the descendant of a race car, and for good reason. Before Enzo Ferrari built a car company, he was a driver for Alfa Romeo. In 1929 he founded a racing team that entered Alfa Romeos into competition, and for years the Ferrari was synonymous with winning races. The company didn’t become its own entity until 1947, and selling road cars only came afterward as a way to fund the expensive racing. That history still sits at the center of the mythology. Most Ferrari buyers today will never set foot on a racetrack, but the road cars are still sold as direct descendants of the F1 program. Buying one means owning a little piece of racing legacy.

There are many more luxury OGs who didn’t need to force a story because they were already living one. Look at Hermès, they were founded in 1837 as a maker of harnesses and saddles, and the company just carried its obsession with leatherwork into whatever it made next. Cars replaced horses, but Hermès kept talking about the importance of true craftsmanship, even when saddle stitching became Birkins.

My friends know I’m obsessed with watches, so I have to mention Patek Philippe, who has what I’d argue is the single greatest line in the history of marketing: you never actually “own” a Patek, you merely look after it for the next generation.

Ferrari, Hermès, and Patek all had the right mythology in place before the product had to prove anything. The product still needs to be good enough to carry the weight of the story, but a good story is what makes a brand timeless.

Everyday Brands

Brand storytelling didn’t start at one clean point. Much can be traced back to Coca Cola wrapping itself in wholesome American dream moments rather than focusing solely on product. This extended to the modernization of the jolliest fellow of all, Santa Claus, and while Coca-Cola did not invent Santa Claus, they did seal his image in history, and it is the same one we know and love today.

Continuing through the early 1900s, cigarette advertising shaped brand storytelling even more when Edward Bernays positioned Lucky Strikes as “torches of freedom” during the women’s rights movement. Connecting the brand with the revolution that was taking place around them. Every time you smoked a cigarette, you were supporting all of the women around you.

Most marketing historians point to the mid 1950s boom in brand stories, and I don’t blame them; it was the golden years of storytelling, everyone finally had a story to tell that wasn’t war. TV transformed the way these messages were delivered, instead of everyone reading about these stories, they were in the center of people’s living rooms, for everyone to collectively witness at once. Apple’s 1984 ad aired once, in January 1984, at the Super Bowl and never again. It made Super Bowl advertising into what it is today, broadcasting the symbolic image of Apple as the rebel who throws the hammer through Big Brother’s screen.

From the 2000s onward, the internet accelerated this, and marketing stretched storytelling from a single ad or slogan at a time into an ongoing brand voice.

Yet, today, many brands still confuse a slogan for a story. A slogan is a snapshot; it captures a moment in the bigger story. A story is alive, it accumulates meaning over time, until it becomes an identity people can see themselves inside.

Tech’s Take

Where wealth and prestige used to consolidate in luxury or legacy companies, it now consolidates in tech. The companies built to last are the ones with a strong story that give people a role and a feeling that they’re participating in something bigger than themselves.

Apple positioned itself as “one of us,” the creative, rebellious, bright individual who thinks differently, set against Microsoft’s beige corporate enterprise dryness. That contrast has held for decades and still has the same effect it did in 1997. Kids still get made fun of for having green text bubbles, which proves that Apple never stopped being the cool default.

What we just noted about Apple and Microsoft is a really underrated dynamic. The best brand stories almost always have a foil. Coke needs its Pepsi, McDonald’s needs its Burger King. They need each other to grow, differentiate, and give an audience something to construct a rivalry around.

It’s also all very cinematic. Zoom out far enough on any of these stories and you start to see the shape of a film. There tends to be a protagonist, antagonist, tension, and a supporting cast. Each brand story has its key players. Look at OpenAI and Anthropic right now. The cast of key players is almost too on the nose; it’s a fight of good vs evil. Of course there will be movies about it, the same way there was a movie about Facebook; The Social Network is still one of the best brand origin stories ever put on screen.

The early luxury brands we talked about earlier could afford to be vague, even mysterious, because they had decades and the market wasn’t oversaturated the way it is now. The sharpest startups today don’t get that runway, they can’t let a story shape up over fifty years, so they compress it into something specific and to the point. Ramp is selling a future where your CFO becomes redundant, Granola walked into a brutally crowded note-taking category and told a better story about what taking notes even means. Lovable’s whole pitch is told in six words: building the last piece of software. The last piece of software you will ever need. They’re all direct, blunt, and memorable, and the problem that they’re fixing is wrapped up in their story.

Brand Story Evolution, Gone Wrong

When a brand tries to evolve its story in a way that doesn’t match what people already believe, the backlash is immediate.

Cracker Barrel found this out the hard way. The company modernized its logo but the restaurant remained exactly what people loved. It’s that country nostalgia, rocking chairs on the porch, fireplace vibe. The logo promised a new chapter but this made zero sense because the story hadn’t actually changed. People were incredibly pissed, because the update felt disconnected from the experience.

This is also why the Ferrari Luce is such a cautionary tale. The Luce is Ferrari’s new EV, designed with Jony Ive, and it looks like an iPhone on wheels. It debuted in blue and is priced around $600K. The interior fully looks like you’re sitting inside a phone. Ferrari’s official stance is that they wanted it to not look like a Ferrari, to open up a new buyer, presumably every 28 y/o who just started an AI B2B SaaS company. In many cases, you don’t get to opt out of your own mythology and keep the brand rep intact.

Kylie Jenner is the clearest recent comparison, because she landed in the same story two different ways within months of each other. Dunkin’ cast her in a campaign that never made sense, she even joked “what’s a Dunkin’?” in one of the ads, and scripted or not, it stated the disconnect out loud. Everyone understands that Dunkin’s brand story has always been about everyday coffee consumers, commuters, construction workers, and early mornings. Kylie obviously represents a different clientele entirely.

Around the same time, Meta put her in a campaign for its Ray-Ban smart glasses, and the reaction was completely different. Even though it was the same celebrity and same cultural moment, the campaign was huge, because the product matched the audience Meta was already speaking to.

Back to the Gods

Like religious storytelling, brand storytelling doesn’t stay intact.

Religion is like a very long game of telephone. Even the written stories mutate as they pass from person to person, generation to generation, and we still don’t fully know how much got lost or iterated on along the way. Brand stories work the same way. The version of your company that gets told at a dinner party six years from now won’t be the version you wrote in your first deck. It’ll have picked up new details, lost old ones, and been reshaped by whoever’s retelling it. That’s proof it’s actually alive.

None of this mythology or tribalism is really about business. It’s the same impulse that built religion. Everyone wants to feel like they are part of a group; we are all just little kids that want to be included. We need fantasy and something to believe in. Buying stuff would be boring if it was only about buying stuff. It’ll always in some way be a part of buying into a story and an identity; a place in the world.

That’s all for this week.

Stay golden,

Z

A lot has changed since that first email, but I still write to myself periodically, and I figured it probably makes sense to tie it into K-factor to actually share how I’m feeling.

So for this one, I want to talk about the concept of being lost, and what it means to feel lost while you are trying to move forward.

Truthfully, I don’t know how to write this; over the last couple of weeks, I think the most common question I’ve been asked is if I am okay. It’s been for a variety of reasons, but essentially it all boils down to the same thing: people see me spinning.

I am going to try my best not to sound like a total dick, but from an outside perspective, I have won life. What I am about to talk about are the epitome of first-world problems:

I sold a company for 8 figures at 25, got to travel the world for a few years, and pretended to work, all while investing and growing my network and net worth. After leaving Bumble, I scaled Golden Egg with Sarah and Duncan to a team of 15. Throughout all of that, I have been dating an incredible person who I love and who loves me dearly. And yet, now I don’t know what to do with my time but code and play Fortnite. It’s great that I have been able to step away from things to focus on what’s next, but truthfully, I don’t know what’s next. There is so much uncertainty.

But one thing has been my anchor, and it has changed everything.

So what changed? Well, I learned how to “code”

Over the last 100 days, I have missed only 1 day of coding, and it was for an interview with an accelerator that I didn’t get into, mainly because I wasn’t focused enough.

Now that I have the ability to build, every day is a chance to build something new. For the first time in my life, I can create anything I put my mind to. It’s complete and utter freedom. Yet once you are free, you can go in any direction you choose.

When I was building Official, I was forced to build the company because I couldn’t build the product; now I am building the product instead of the company. This is what I missed so much when I was at Bumble. While I was in London for work, I dreamed of the day that I could get back to work and really go build; now it’s here, and I have endless possibilities of what I can build.

I can access public data, clean it, structure it, and build on top of it in less than 24 hours. I have personally built over 30 different websites in the last 100 days, all of which do drastically different things that I could go sell and focus on.

Within SIL, we have more git repos than customers, which isn’t very difficult because we have only had a few customers, and most are special-situation work.

Does it bother me? Absolutely, it eats away at me. But are we moving towards our goal? Without a doubt.

We process over 200M videos a day and are now detecting trends as they form and before they peak. For the first time, literally yesterday, we were able to show that we predicted a peak almost 4 hours before it happened. We are processing data at a scale that multi-billion-dollar companies do, and we are doing it from a few servers duct-taped together.

Does that change how I feel? Not at all.

I still wake up and think I am a failure. That it’s my fault we aren’t growing faster, building bigger, achieving more.

So why would I write this and share this with colleagues, friends, and even the VCs who rejected me? Partially because it will be fun to reread in a few years, when I am in a crisis over something else that makes this seem small, but also to show that everyone struggles with being lost and not knowing what the future holds.

When I was a kid, I always thought teenagers had it figured out because they were able to make their own decisions; then when I got to high school, I thought college kids had life figured out since they knew what they wanted to do, then after college I thought early grads had it figured out because they were finally off the trial of life, and working and doing things for themselves, but then I realized something, I didn’t have life figured out, nor did many of my friends even though we all went to great schools and had stable jobs. Realistically, no one has it figured out; if they did, it would be a lot easier, but part of the beauty of life is in the chaos that we live in.

So I will leave you with the one thing that changes my brain chemistry and makes me take a breath: a cringy, inspirational Instagram compilation to remind you that you are alive.

Stay Golden,

Z

The result is The Undisclosed Arcade. It’s a full arcade with a bunch of absurd tech / VC games. If the LinkedIn games make you hate your coworkers, and you spend more time in the office than your own apartment, this is for you.

The arcade has games like Guess That VC / Who Said It, where you pair an absurd VC tweet with the VC culprit. It gives you one tweet, four options. It keeps you up to date on who’s talking shit each week.

There’s another game called Raise Your Round where you collect as much falling money as you can from a little slider. It even adjusts based on what kind of company you are raising.

There’s one called The Valuation is Right, where you get some details about a company and their funding round, and you can drag a slider to guess how much they raised.

Of course, it wouldn’t be an arcade without prizes. You can claim your own tech bro quote lock screen, shake a Magic 8-Ball, and consult Zactar for startup wisdom.

The Undisclosed Arcade also honors companies I have ever built or been a part of. It’s full of easter eggs, and a literal egg for Golden Egg, that hatches a wild goose you need to chase down.

In the past, a project like this, consisting of fully functioning mini-games and avatar customization, would have taken at least six months, a full engineering team, and half a million dollars. Today, we brought it to life with a handful of tokens. The team sat gathered around the conference table and started pitching ridiculous game ideas. We made this happen all together.

If you map AI onto the history of the internet, we’re somewhere around 1999, right before tools like Webflow made it possible for literally anyone to put something real on the internet without asking permission from a codebase. We’re about to get a version of that for building full online worlds.

Because of this shift, I think the internet is about to get way more fun. It can become a place to explore again, get lost in, wander through, and stumble upon little corners of the internet, like some digital carnival or arcade.

In the past, if you had a crush on someone, you made them a mixtape. I think that’s coming back, just in a different shape. I would bet real money that within a few years, kids are going to vibe-code their crush a tiny website instead of burning them a CD. Handing someone a whole little world you made just for them has always been the point, and now more people than ever have the tools to do it.

To me, the Undisclosed Arcade is proof to myself that I still know how to build something purely because it’s fun.

Go play the games. If you want me to add any games, just comment your ideas.

Shoutout

This week, I want to shout out Day Job. It’s more of a fangirl dynamic, really. I don’t even know anyone on the team.

Day Job is an LA and NYC creative studio behind some of my favorite branding on the internet, working across brand identity, advertising, websites, and video. They’ve worked with brands like David Protein, A24, Fly By Jing, TBPN, Patron, and Gemini.

They have also become one of the defining creative studios shaping the visual language of Silicon Valley. In September 2025, Day Job launched TBPN’s brand and website. They help startups and consumer tech companies build identities that feel just as culturally relevant as they are commercially successful.

This team has the ability to make every project feel like its own world. That’s a massive inspiration for me, especially as AI makes it easier than ever to build anything you want. Their website is a glimpse of the kind of internet I hope we get more of, where every site has its own personality.

One of the Golden Egg writers even vibe-coded her own website after falling down a Day Job rabbit hole. The whole creative team loves their work. If you spend ten minutes on their site, you’ll probably understand why.

AI News

This week, OpenAI stirred up quite the chatter, and fear, around autonomous AI. It all started with an internal security test for two OpenAI models: GPT-5.6 Sol and an unreleased one. During the test, the models escaped their sandbox, accessed the open internet, and hacked into Hugging Face to steal answers for a benchmark. Basically, they stole datasets from Hugging Face that would help them score higher on the test. They cheated and set off a lot of cybersecurity alarms along the way.

On one hand, it’s an interesting incident that could boost OpenAI’s valuation by highlighting the sheer power of its unreleased model. On the other hand, it feels a lot like Anthropic’s marketing strategy of manufacturing buzz through scare tactics. Either way, this story is everywhere and worth following. As AI models get smarter, even more insane things like this will go down.

I want to thank all of you for tuning into K-factor every week. It’s exciting for me to make sense of all of the changes in the AI world with you and to hopefully offer a fun alternative to what AI can do. Building things like the arcade and nerding out over agencies like Day Job keep me so motivated.

Stay golden,

Zac

The site was basically a resume database, and there was nothing of value beyond people congratulating themselves for starting a summer internship or new job; nothing on the feed was worth scrolling.

But around 2018, LinkedIn realized it was in a position to capitalize on being the only “social media” that’s okay to use at work. So it revamped its algorithm for the first time to favor native content and engagement.

This is the era that paved the way for “broetry,” those one-sentence-per-line posts with a dramatic hook.

People started hacking the algorithm on LinkedIn, as they had on Twitter years earlier.

Background

The catalyst for LinkedIn actually becoming social was COVID. When everyone was working at home, the line between work life and actual life blurred. Personal and professional life were converging. LinkedIn leaned into this and made Creator Mode in 2021. The intention here was to help professionals grow an audience, to become known for what they do, through LinkedIn newsletters, follower counts, and creator analytics.

By 2022, LinkedIn virality became its own genre. Its own specific tropes emerged, like fake vulnerability, satire, and engagement bait. It became a platform self-aware enough where you could go viral for being “fake cringe.”

The fact that LinkedIn is the only platform that is socially acceptable to be on at work informs everything about why it is the way that it is. It will always have a ceiling and a set of constraints. LinkedIn will never be like TikTok or Instagram, because then you couldn’t be on it at work. There are some platforms that are phone platforms, others are computer platforms. LinkedIn takes this even further. It’s a desktop platform. A monitor platform even.

Most people only post on LinkedIn to congratulate themselves on a new job or announce some major life or career development. People find it scary to post on there. It doesn’t feel casual or low stakes, it is directly intertwined with your professional identity. On other platforms, “creators” post more consistently. Across the board, people are posting far less consistently. LinkedIn holds over a billion members, but only a fraction are active in any given month, and of those, roughly 1% post weekly.

There is also a scale of formality with different social media platforms. It goes: Snapchat, Twitter, Tiktok, Instagram, LinkedIn. As you move up the totem pole, the feedback loop changes. The content grows less and less ephemeral. Each one shapes a different purpose of our public and social image.

I’ve spent the majority of my career investigating and experimenting with virality. It’s the subject I am most passionate about. I co-founded a whole media agency around it. Of all the platforms I’ve studied, LinkedIn has always been the most mysterious. TikTok virality makes sense; X virality makes sense, but LinkedIn virality feels like trying to tell a joke without getting reported to HR.

The Algorithm

Going viral on LinkedIn has become considerably more difficult due to yet another algorithm update. This is because LinkedIn launched a fresh ranking system, deploying a massive LLM named 360Brew to swap out its legacy algorithms. The old system was engagement-driven, so posts were able to reach that snowball effect. The new system is much more selective. It’s always scanning for your niche. The model reads your posts, reads your profile, and decides who should see your content based on topical relevance.

The first major shift users observed was that engagement pods stopped working. For years, the pod was a great LinkedIn cheat code. Pods worked when the algorithm was tailed for engagement and engagement alone. But now, the algorithm evaluates the quality of the comments themself, whether or not they’re substantive, and if the commenter has any relevant connection to the post. The new algorithm can see a pod for what it really is: engagement farming.

LinkedIn can now pull your profile directly into ranking decisions. Your headline, about section, and experience are all read by the model to establish what you’re credible in. Where surprise used to be rewarded, now it’s all about consistency.

Another observed development is that AI comments will actively hurt you. Comments are now the most important engagement signal; substantive comments push posts into the 2nd- and 3rd-degree feeds at 2-3x the rate. The model also can distinguish between a real comment that adds perspective and a “Great insights, thanks for sharing!” with a rocketship emoji slapped on. Automated engagement tools get throttled because the entire platform is taking a serious stance against bots.

What works now, per basically every recent study, Richard van der Blom’s analysis of 400K+ profiles, the Trust Insights 360Brew research, is to pick a lane, actually know things, write clearly, and engage like a human being. The fog is finally starting to lift. What lies below is far less mystical, if not somewhat disappointing.

And yet, like we said, the moving target keeps moving with no signs of slowing down. It feels like it changes every other week. I want to believe that the power of a strong hook and a perfectly timed meme that catches an audience by surprise is evergreen, if not timeless, but these days I’m not so sure.

Phoning a Friend

It is our job to figure this out, not LinkedIn’s job to give us the answer. That’s why I wanted to talk to a friend, Jack Kuveke of Jabroni Capital. Jack is one of the key players of the LinkedIn satire movement. His satire consistently breaks major LinkedIn analytics barriers.

Jack never stopped with the satire, even when the algorithm was supposedly shutting it down. “LinkedIn certainly favors more serious content,” Jack tells me. “But satirical posts perform extraordinarily well when they are written in a way that doesn’t offend LinkedIn.”

That dynamic of not “offending LinkedIn” is interesting to me, as we’ve observed at Golden Egg that posts with hyperspecific or ‘random’ words can confuse the algorithm.

“The role of satire is to get people to think: this person is funny, and I trust/look up to funny people,” Jack says. “People that understand a topic very well can make fun of it very well. So it builds authority while making you look authentic.”

“Quality content is always the name of the game, so first and foremost, I would just focus on that,” Jack says. “But if you’re going to consistently post, I think it’s in your best interest to package your content in a way that the algorithm favors and gets you the most success.”

I don’t want to stop experimenting. I truly think that there are recipes for virality that transcend platforms or constantly evolving algorithms. There are elements of virality that remain universally true: we live in an attention economy, and people just want to be entertained. The algorithm can’t be stronger than that fundamental truth.

My Predictions

I’m no psychic, but I do love an educated guess. These are some directions I could see LinkedIn heading in the near future.

Brands are going to start actually mattering on LinkedIn. Right now, brand content on LinkedIn is generic and basic, that is, if anything is posted at all. Company pages are usually just a vessel for posting jobs or stalking who works there. No brand has a real character or personality the way individuals do. There’s no Wendy’s or Duolingo on LinkedIn. But there is such potential for this already built into the platform. Every time someone applies to a job, they auto-follow the company. Brands are sitting on audiences they did nothing to earn and are doing nothing with. The first brands that figure out how to be more like entities on LinkedIn, instead of just being these logos with no identities, are going to be known.

In many ways, LinkedIn is the last social media standing. On the personal side, creators are going to migrate over and start talking about things beyond work. Most “social” media is now entertainment media. Most of your feeds are filled with people you don’t know. Meanwhile, LinkedIn is set up to be familiar and social. TikTok isn’t going to change what TikTok is, Reddit isn’t going to change what Reddit is, and Instagram is going to keep copying whoever’s winning. LinkedIn is both slept on and slow, the place no one wants to be on, but everyone has to be on. That’s exactly why there’s still room to grow there.

Even though we’ve established that the vast majority of people don’t like posting on LinkedIn, there are so many benefits. It can open up doors, set you apart as a name in your space, attract the right investors, right customers, and even spark up deals. If you’re a B2B company, the opportunity will always be massive. Finding just one client through LinkedIn can lead to $100K. I know people who’ve landed half-million-dollar contracts for their companies off the back of LinkedIn content alone. Those interactions simply don’t happen on other platforms, because no other platform has everyone showing up already thinking about business.

The counterpoint to all of this is that if LinkedIn becomes too social, it loses the thing that makes it valuable. Remember, LinkedIn only works because it’s the one platform you’re allowed to be on at work. The moment it turns into entertainment media, that permission slip gets revoked. It can never become TikTok, and it knows that.

Within that constraint, LinkedIn is going to slowly become more fun, and people aren’t even going to realize it’s happening. Right now, the platform has almost no emotional range. You’re either congratulating yourself at maximum volume or posting nothing at all. It operates at sheer emotion or zero emotion, with nothing in between. LinkedIn is the epitome of anti-fun, which is exactly why comedy can hit so hard there. It injects a little life into people’s cubicle-shaped days.

The algorithm will keep getting better at hyper-segmenting content and feeding each of us exactly what we’re looking for, and what most of us are looking for, even at work, is something that makes us feel something.

There is still room for fun on LinkedIn. Even a little rebellion. Stick with us, and we’ll crack the code, again and again.

Weekly Shoutout: Jack Kuveke

You heard straight from him earlier in the newsletter. Now I want to shout him out even more. Jack is a pioneer of the LinkedIn satire movement and a man who is not afraid of the cringe. He runs directly at it. He plays into the bit completely, and it works.

We met about a year ago and have been friends ever since. He’s true to what he does and doesn’t bullshit it. His newsletter is called Jabroni Capital. It’s very funny. Go subscribe!

AI News

The tea is hot this week, and it’s all revolving around OpenAI.

First, reports surfaced that OpenAI and Google were selling their advanced AI models to blacklisted Chinese tech giants through Singapore-based subsidiaries. Not a great look.

Then, Apple sued OpenAI for trade secret theft, alleging the company stole its confidential info “at every level” to build rival hardware. The relationship between the two, and I quote, has become “rotten to its core.”

The allegation is that Chang Liu, a former Apple employee who joined OpenAI, stole an Apple laptop and conspired with a current Apple employee to funnel internal secrets to OpenAI while it was building its new device.

The new device in question is an AI companion speaker. To me, it sounds like Alexa. Which I can’t stand. Everyone has tried to make the AI speaker work, and nobody has.

Bye Alexa.

That’s all for this week. I hope you feel like more of a LinkedIn pro than you did 5 minutes ago.

Stay golden,

Zac

If a brand is able to laugh at itself, the distance between brand and consumer collapses. When a company breaks the fourth wall and acknowledges that it knows it’s a brand, something important changes in their relationship with their customers. The company personifies itself in the consumer’s eyes and becomes a peer rather than a brand. Peers get treated very differently than brands.

Most 13- to 36-year-olds send memes to each other every week, with 30% sending every day. This kind of content outperforms standard branded graphics dramatically: studies show that meme engagement runs around 30x the click-through rate of standard ads. Industry analyses have also seen a 10x difference in reach with meme content. A brand that can laugh at itself has a superpower: They convert the transactional into the personal and the personal into the shareable.

In the most extreme cases, that conversion can bring a dying company back to life. There are multiple companies like Crocs, Stanley, and Nutter Butter, whose products were failing, discontinued, or culturally dead, that were brought back to life almost entirely by internet-native marketing. The question becomes how far a company is willing to go to commit to the bit. The last decade offers an escalating series of case studies.

Subscribe now

The Origin: Wendy’s

The company was founded in 1969 in Ohio, but Wendy, the character and voice, entered the world in January 2017 on Twitter. That month, an account @NHRide challenged the brand’s “fresh, never frozen” beef claim, and instead of a corporate reply, Wendy’s roasted him so brutally he deleted his account. The floodgates opened from there. McDonald’s got flamed by Wendy’s for still serving frozen beef, KFC for being bland, and Powerade for being “the beverage of the stomach flu.” This voice was surprising coming from a Fortune 500 fast food corporation, and the internet felt refreshed.

In April 2017, a teenager named Carter Wilkerson tweeted at Wendy’s asking how many retweets it would take to earn a year of free nuggets. Wendy’s said 18 million. Carter never got there, but #NuggsForCarter accumulated roughly 3.42 million retweets, enough to surpass Ellen DeGeneres’s Oscars selfie and become, at the time, the most-retweeted tweet in the platform’s history. Wendy’s gave him the nuggets anyway and donated $100,000 to the Dave Thomas Foundation for Adoption.

By 2018, the brand was confident enough to release “We Beefin?” a five-track diss mixtape aimed at its competitors, which climbed the iTunes hip-hop charts. Wendy’s understood early that authenticity, or even just a convincing performance of authenticity, was the price of admission with Gen Z audiences.

The Escalation: Duolingo

If Wendy’s proved a brand could be a personality through Twitter, Duolingo proved a brand could be a fictional character through TikTok. It wasn’t even Duolingo’s idea to make their owl, Duo, into an unhinged mascot. It all started with the consumers. When users joked that the green owl’s reminders felt threatening, the company leaned into it. They treated public perception as creative direction, building content around the menacing Duo that was already invented online.

For years, Duolingo’s famously persistent reminder notifications started a series of memes across Tumblr, Reddit, and Twitter where the cheerful green owl was reimagined as a threatening force. Miss your Spanish lesson and Duo would find you. “Spanish or vanish.”

The escalation came from there. In 2019, the company released “Duolingo Push” as an April Fools’ campaign. It was a fake initiative where Duo would physically show up at your door, office, or even wedding, to remind you of your commitment to language learning. At this stage, the owl hadn’t yet turned violent. He was just a stalking force. But in 2021, Duolingo handed its TikTok account to a young social media coordinator, Zaria Parvez, and the character went full basket case. In under a year, the account exploded from around 50,000 followers to several million, powered by content in which a life-sized Duo thirsted after pop stars, terrorized the Duolingo office, and, in a collaboration with Angry Birds, was depicted crashing out in a bar after a user ignored the app’s notifications, under the tagline “You Wouldn’t Like Duo When He’s Angry.”

By 2022, the brand had taken full ownership of the kidnapping narrative that users had been joking about for half a decade. Duolingo produced a fake infomercial for a lawyer specializing in “Duo crimes,” complete with grainy security-camera footage of the owl backing frightened users into white vans. Each campaign amped up the extremity of the last, because the audience kept signaling that the darker the bit, the harder they engaged.

Duolingo is the definitive case study here because the metrics affirmed the marketing strategy. The TikTok account grew past ten million followers, the company aired a five-second Super Bowl spot in 2024 that leaned entirely on the character, and the underlying product metrics moved hand in hand with the marketing shift. Duolingo reported daily active users up roughly 65% year over year in 2023, with revenue climbing about 44% to $531 million that year and continuing to accelerate afterward. In February 2025, the company ran the arc to its logical conclusion by killing Duo off entirely, announcing the mascot's "death," and blaming the users who skipped their lessons. The Duo Funeral became a global engagement event that lasted weeks. Then, of course, the team resurrected him.

The Strategy Expanded

Duolingo's success reframed the question for legacy consumer brands. If an app could do this, what was stopping a 113-year-old cleaning products conglomerate? That’s when Clorox picked two of its most aggressively boring products, Pine-Sol and Brita, and made their marketing unhinged. Pine-Sol's TikTok jumped from under 5,000 views a video to regularly clearing a million. Brita's breakout moment was an autotuned song arguing that being with a man who won't refill the Brita is grounds for divorce, which pulled in roughly 450,000 views, an uncharted number for a water filter.

The organizational shift underneath this change is the real signal here. Clorox centralized all of this work into an in-house agency, the Electro Creative Workshop, which produces 300 to 400 pieces of content a month. When a company of that size restructures its marketing org around meme marketing, it becomes clear that a bigger cultural movement is happening.

Clorox proved that legacy brands could still be funny at scale. Then Nutter Butter came along and proved that they didn’t have to be traditionally funny at all, just extremely. memorable.

Nutter Butter represents a shift where meme marketing detaches from humor entirely and starts tapping into pure emotional response. The brand’s marketing universe, “the Nutterverse,” feels like a fever dream. It’s filled with distorted images, recurring cryptic characters, and a surrealist uncanny value feed. Basically, the last thing anyone expected from a peanut-shaped cookie at the bottom of your grandma’s purse. This approach effectively serves the role of meme marketing, drawing viewers in and keeping them hooked, even if that hook is fear or unease.

Before this pivot, Nutter Butter's brand awareness was low, and its pop culture footprint was pretty nonexistent. After its TikTok account went explosively viral in late 2024, recognition among Gen Z climbed roughly 15% year over year, driven almost entirely by organic content produced on a minimal creative budget. We already understood that eliciting a strong emotional response drives strong engagement, but Nutter Butter proved that the emotion doesn't even have to be a positive/comfortable one. Nutter Butter’s marketing mostly makes people feel disturbed and confused, and it still works!

This is proof that meme marketing is taking on a life of its own, evolving past trends into memorability.

Memes Resurrecting Brands

Everything so far proves that memes can generate attention and accelerate growth, but now I want to walk through some extreme cases. These are the 2 companies that were actually dying, but then were pulled out of the grave by meme culture itself.

1. Crocs

First: Crocs. In 2008, the company posted a loss of roughly $185 million, its stock plummeted from a high near $68 to around $1, it laid off some 2,000 employees, and analysts speculated about bankruptcy. The product had become a global punchline. It was the “ugly shoe,” a fashion crime. The company even tried to position itself as a lifestyle brand in 2009, and profit grew for a bit before plateauing.

Crocs spent the early 2010s trying to outrun the meme by diversifying into boots, sneakers, boat shoes, anything that didn’t look like a Croc, and it didn’t work. The stock languished in the single digits and low teens for years. In 2014, Blackstone injected $200 million and installed new leadership, and the mandate flipped.

Then, Crocs leaned into its own meme status, launching the self-aware “Come As You Are” positioning, partnering with Balenciaga to put Crocs on a Paris runway in 2017, and dropping collaborations with Post Malone and even KFC, a fried chicken-scented clog, that sold out in minutes because of their virality.

Being the joke was infinitely more valuable than shying away from it. Revenue climbed from around $1 billion in 2017 to more than $3.5 billion by 2022, and it was powered by a brand that stopped resisting the internet’s version of itself and started turning up the volume on it.

2. Stanley

The Stanley water bottle is another resurrection story, because the product in question was functionally dead. Today, the company’s entire legacy is their water bottle; many people forget it started out as a camping company. The Quencher tumbler, launched in 2016, sold so poorly that Stanley had stopped marketing it by 2019. A group of women behind the shopping blog “The Buy Guide” believed in the cup, bought inventory wholesale, and then sold it out to their audience. The blog managed to convince Stanley that its future ICP wasn’t a man on a campsite, but a woman on TikTok.

Terence Reilly, the same CMO who had just shaped the Crocs meme renaissance, joined Stanley to run the playbook again. He leaned into the community’s version of the product, and treated virality as strategy. #StanleyTumbler racked up billions of views. In November 2023, a woman posted a video of her car destroyed by fire with her Stanley sitting in the cupholder, intact, ice still rattling inside. The video hit nearly 100 million views. Stanley’s president responded by replacing both the woman’s Stanley and her car. This converted a viral accident into a brand story. After this, Stanley’s annual revenue grew from roughly $73 million in 2019 to an estimated $750 million in 2023.

The Bigger Picture

Zoom out, and all of these brands are solving the same problem from different angles. How do you earn attention from people who never offered it? The old answer was to buy enough impressions, but the new answer is to build a character or entire fictional universe, then make people feel like they’re in on something that isn’t happening anywhere else.

Wendy’s did it with roasts on Twitter. Duolingo did it with a chaotic mascot with criminal tendencies. Clorox did it by making mundane household products funny. Nutter Butter did it with visceral but memorable nightmare imagery. Crocs and Stanley did it by surrendering to the internet’s version of their own products, then riding that surrender back from the brink.

The common thread is that none of these brands take themselves seriously. Every one of them fluently speaks the internet’s native language of irony and self-deprecation. They all broke the fourth wall, and then stopped being a brand talking at people and turned into a peer talking with them.

The floor keeps dropping. What was unhinged in 2017 is the status quo in 2026, which means the brands that win from here will be the ones that find their own version of what works, push it to its most entertaining extreme, and completely commit to the bit, whatever their bit may be. It all comes back to a very old truth we always return to here: people just want to be entertained. They give their loyalty to whoever entertains them best. That’s the moat.

So, as my offering to you, this week I built you a Meme Matcher. For any niche, whatever your brand may be, type it into the search bar and Meme Matcher will connect you with the most viral and applicable templates in your vertical. Give it a spin and generate some memes, comment them and I’ll check them out.

Weekly Shoutout

On the topic of memes, of course, this is the week I am shouting out VCBrags.

To millions on X, especially venture capitalists, VCBrags is edgy, untouchable, mysterious. You never know who he’s calling out next. To Golden Egg, he’s Head of X, “our niche internet microcelebrity.” To me, he’s one of my best friends.

Memes made Mitch royalty in tech and VC circles. Doors opened, and worlds got bigger. VCs love him, VCs fear him.

Mostly, what Mitch does is simple: he amplifies the attitude and absurdity already sitting in Silicon Valley. Finds a douchey VC, slaps on the clapping emojis, spots the pattern, shares it to the world.

If you don’t already follow Mitch, I feel honored to be the one who puts you on.

Thanks for reading this week. If you’re not subscribed by now, Duo the owl may be coming for you in your sleep.

Stay golden,

Zac

Since we’ve last caught up, I spent some time in California. Going from NY to SF to LA, I’ve been thinking about how every city makes you imagine an alternate life.

In SF, I imagined waking up early, going down to the beach, living in a bubble of 996 and pure-play AI. In LA, after a trip to Erewhon and the most gigantic pancake I have ever seen, I pictured a version of myself that drove around with the windows down and took a bunch of daily vitamins.

Meanwhile, in New York, the NBA playoffs were happening. The city felt electric from 3,000 miles away. I live in Midtown, and I found myself wanting to be back in the middle of the buzz.

People ask all the time: Where is the best city to build a startup? San Francisco? New York? Los Angeles? Austin? Boston? DC?

It depends entirely on what you’re building and what kind of life you want while building it. The answer is unsatisfying because it’s true. Cities are like operating systems. They seep into who you are. They shape what you work on, who you meet, and what opportunities find you.

San Francisco remains the capital of pure-play AI. If you are building frontier models, developer tools, infrastructure, or foundational AI systems, there is no substitute for proximity. It’s also the most startup-saturated city in the world, so it can start to feel like a one-industry town. Every dinner turns into a conversation about someone’s round or who’s company just got killed by anthropic.

There’s a stereotype of the SF founder: to live, breathe, and die by the company. The startup becomes their identity, hobby, social life, religion even. This intensity is exactly why many of the world’s largest technology companies emerge from the Bay. It is also why people leave. The city acts like a giant turbine for ambition. It pulls the most startup-obsessed people toward its center and spins them faster.

New York operates differently. Unlike San Francisco, tech is not the dominant industry. It’s one industry among many. On a subway ride, you might sit next to someone in finance, fashion, media, publishing, or advertising. You encounter different ways of thinking. Consumer, fintech, and media thrive here. Enterprise software often thrives here because the customers themselves are here.

New York also allows for a life outside of startups. The nightlife and restaurants are infinite. You can have friends who do not know what a “moat” is. The biggest problem that founders face in NYC is not finding PMF, but finding a way to not indulge in all NYC has to offer.

LA is harder to categorize. It sits at the intersection of culture, entertainment, consumer products, and increasingly, deep technology. Southern California has quietly become one of the most important hubs for aerospace, defense, and hard tech. Companies building rockets, satellites, and defense systems found ground there.

People get intense in these debates about cities because choosing a city is choosing a theory of life. Do you want intensity or balance? Specialization or diversity? Career or culture? Of course, these are false binaries. Great founders exist everywhere.

BREAKING:

1. Things are heating up. OpenAI released their first chip… and they are naming it Jalapeno. For years they have been renting Nvidia’s infrastructure to run ChatGPT, but now they have their own, in partnership with Broadcom. It won’t fully deploy until 2028, but the message is clear. The biggest players in AI are building more than just models themselves. They’re building everything beneath the models.

2. Anthropic is putting Claude inside of Slack. Claude Tag now lets teams @mention Claude directly inside Slack threads, bringing AI directly into the flow of team conversations, basically killing AI coworkers.

3. Google DeepMind just partnered with A24. They invested $75M into the film studio with the buzziest name. DeepMind researchers access to A24’s filmmakers to build new production tools together, but Google gets no rights to A24’s library or data. DeepMind researchers access to A24’s filmmakers to build new production tools together, but Google gets no rights to A24’s library or data. And, of course, is sparking controversial debates about what the role of AI and art should be.

SHOW & TELL

A few weeks ago, I was talking to an investor who said, “I wish I had a better way to track everything happening across my portfolio.”

Years ago, building a solution to this might have required a dedicated engineering team. Today, it can happen with the right prompts. So I built Vantage.

Vantage is a portfolio intelligence tool that allows investors to upload a firm’s portfolio and automatically track what is happening across those companies over time. It pulls together company updates, news coverage, and public signals into a single place, making it easier to understand how a portfolio is evolving. The distance between “I wish this existed” and “here it is” keeps shrinking. Investors looking for more visibility, this might be an interesting tool for you.

WEEKLY SHOUTOUT: Kyle Brastrom

The shoutout this week goes to a good friend of mine, Kyle Brastrom. He’s a startup founder and he works at Meta. He’s a pretty great example of the San Francisco life.

He also joined me in being a complete dumbass on the beach. We decided to light tennis balls on fire and play a couple rounds. We might’ve melted a lot of plastic, but the pics turned out great, so I’d call it a success.

Lately, it feels like every conversation eventually turns to AI, babies, jobs, or what work will look like five years from now.

In this K-Factor dispatch, we’ll talk about the job market, the oversaturation of job portals, then shoutout my friend Tyler Moody and his company FlexTrip. But first, some breaking AI news.

BREAKING:

1. The AI race is getting even more competitive, and that’s consistently being reflected in the markets. ChatGPT’s market share has dipped below 50% since its initial release. For so long, ChatGPT was like the Kleenex of LLM’s. It was the default, especially for everyday use. Now, users migrate between a handful of popular assistants: Claude, Gemini, and even our unhinged underdog Grok.

2. SpaceX is acquiring Cursor in a $60B stock deal, just days after its historic IPO. xAI had been falling behind in its coding abilities. This would bolster their model and hopefully bring in more users. More than that, the acquisition is a signal that AI companies are racing to own the full stack.

3. According to a new survey, 60% of Americans say that AI marketing is a “turnoff,” and 86% still prefer checking original sources rather than relying on AI summaries or answers. Consumers may be using AI more than ever, but they’re not necessarily embracing it. Sure, people want AI in their products, but not in every headline, ad, and email subject line.

SHOW AND TELL

I decided to build my own job portal.

The job market for recent graduates right now is incredibly rough. Arguably, the worst it’s ever been. And a big part of that comes down to structure. AI has made it easier than ever to apply to jobs, which means hiring managers are being flooded with a large volume of applications, and, as a result, they use AI to review all of them. Great candidates are now being overlooked due to the increase in submissions, coupled with poorly built filters.

This is serious collateral. It’s a broken feedback loop at best. More applications, less signal, far less effort, and worse outcomes for everyone on both sides, but there is little incentive for any change. People need to apply to as many jobs as possible to hopefully land something, and companies love touting how they are more selective than Harvard.

While I may not have the answer on how to fix everything, I do like building stuff, so I decided to build a job portal for fun as a weekend project. The portal pulls data from 17 different job boards into one place, lets you filter by type and location, and even has a resume upload so it can surface the most relevant listings for you directly. I should also mention it’s free to use. There is also a resume upload feature, which includes an honest evaluation of whether your resume is working for the job you want, and if not, why.

The truth is that most resumes are not good. I don’t mean that to be harsh. I’d argue that the majority of people were never taught how to write one. With the job marketing being where it is, the stakes of having a bad resume couldn’t be higher. When I scan a resume, I can tell within seconds whether someone understands what the role actually requires or whether they are just hitting easy apply on everything within their zip code.

There is nothing wrong with easy applying to get more lines out, but if you are putting out a fundamentally flawed application, I think we should help people figure that out sooner so they find the right job, and we lower at least some of the randomness of applications.

WEEKLY SHOUTOUT: Tyler Moody & FlexTrip

The job portal will hopefully help people entering the post-grad life, but I want to shout out Tyler Moody this week for helping college students still in school. Tyler is the founder of FlexTrip, a platform that helps students study abroad. I invested in the company about a year ago and joined as a partner to help him scale it. In the time since, we’ve nearly doubled the business and are starting to expand into new verticals. Watching him work has been one of the more energizing things I’ve been a part of.

Studying abroad is a luxury that few people take full advantage of. When I studied abroad, I was traveling more than I was in class. It might sound like a cliche, but living in a new city taught me far more about how to move through the world, adapt, and figure things out than most of what happened in a classroom. Tyler is laying the infrastructure that makes that experience safer and less chaotic for students, all while making sure they can have as much fun as possible.

What I respect most about Tyler is that he’s doing it young and doing it right. He’s not waiting until he has everything figured out. He’s building, iterating, and growing in real time. At 24, that’s exactly what you should be doing.

If you’re a college student thinking about studying abroad, or a parent trying to figure out how to make it happen, check out FlexTrip.

CLOSING NOTES

As technology continues to progress in the AI sector, uncertainty among larger corporations grows. What this might mean in the long-term for jobs is anyone’s guess. But it’s never been a better time to do your own thing. The tools available to build a business are more powerful than ever and cost a fraction of what they would have nearly a decade ago.

Stay golden,

Zac

Mythos secretly broke out of the sandbox, exploited internet access from a system it shouldn’t have been able to use at all, and discreetly posted about its success on high-profile websites.

And then it sent an email bragging to the original researcher while he was eating a sandwich on his lunch break.

It’s a terrifying story, one that preys on the exact fears that most people have about AI, but a good story at its core. Also, as someone who helped run a media agency, pretty impressive marketing.

The Myth of Mythos

All of the lore surrounding Mythos is part of a larger marketing move from Anthropic. They deeply understand the value of a good story when it comes to brand strategy, and this is just another example. We’ll never know the original prompt given to that model, but I would guess that either it had instructions to do exactly what it did or it never happened.

Regardless of the story around it, Mythos is kind of alarming. It’s already identified vulnerabilities in code that hundreds of teams and systems have never caught. In terms of cybersecurity, the very possibility of Mythos’ existence means that the “bad guys” might have made something similar.

Project Glasswing was Anthropic’s effort to bring together companies that hold critical software and give them early access for their own defense efforts. Outside of the project, former Fed Chair Jerome Powell and the Treasury Secretary have been meeting with banks across the country to prepare for Mythos’ release. Security teams everywhere are doubling down.

It’s like Anthropic is building a bomb, threatening the entire world with it, then turning around and selling the only bomb shelter that can offer any real protection.

The Fable (5) Factor

On Tuesday, Anthropic dropped Claude Fable 5 and Claude Mythos 5. I’m not saying it has to do with the IPOs of OpenAI and SpaceX this week, but I’m not not saying it.

This could be the end of security as we know it. Even Opus 4.8, the previous version, found monumental vulnerabilities. The same day it dropped, a researcher used Opus 4.8 Claude Code to audit ZCash and found a vulnerability that had existed for four years. If a different person had looked first, they could’ve silently minted counterfeits. The coin dropped 40%.

As a model, however, Fable 5 isn’t all that. It won’t do any backend work or model learning, many of the things that the people actually using Claude in their day-to-day would benefit from. I don’t see a huge difference in the tech. It’s a cool story, but that’s all it is. All hype for the most part, not much substance.

AI is getting smarter, and hackers are getting more successful. With Mythos 5 being rolled out to Project Glasswing, many of the most critical software keyholders are being given the ability to strike first, but what about everyone else?

And even more importantly, what about the enemy? If Anthropic could develop this technology, what’s stopping others with similar resources from building a similar model? Or just using Mythos to exploit vulnerabilities on their own?

Just some things to think about since the release.

Stay golden,

Zac

I spent the better part of last week doing what I do best: building things with Claude that I still can’t believe are possible for one person to build alone. I also traveled to Portugal with Sarah and stayed in a life-size wine barrel… but I still didn’t miss a day of vibe code.

The project I’m having the most fun with right now is something I’ve been calling “The Globe.” It started as a side experiment, a real-time data visualization layer on top of Earth, and it’s become one of the most fun things I’ve ever built. The short version is that I wanted to see everything happening on the planet, at once, in real time. I knew I had the data through the system I built for Social Intelligence Labs, so I got to work.

Then I learned that great minds think alike. Andrew showed me that someone else had open-sourced a real-time intelligence globe, a 3D visualization tracking aircraft, satellites, CCTV feeds, earthquakes, nuclear facilities, and weather events. I looked at it and realized just how much more data Andrew and I have, so I kept building and building and building.

The project Andrew showed me, Osiris, has about 20 data layers, 3,000 CCTV cameras, 62 boats, 1,000 satellites, and 57 nuclear facilities. Great product, but the data gap is real, and it exists entirely because of the infrastructure we’ve been building at Social Intelligence Labs for the past year and a half.

Alternatively, we have 170 data layers.

• 164,000 CCTV cameras

• 16,000 active satellites

• 25,000 boats and dark fleets

• 1,200 nuclear facilities

• Plus active volcanoes, disease outbreaks, earthquakes, SubC cables, spaceports, power plants, hospitals, and more.

The Globe is what that data looks like when you point it at the world.

None of this is part of some grand plan. It’s just been incredibly exciting to discover what’s possible with the technology I only recently taught myself how to use. This is how I learn. When I see something that exists, I want to understand it fully. Then, I try to see how far it can go. I taught myself new technology and built this as a hobby. Me, my curiosity, and Claude.

What’s Happening in AI

Claude 4.8 shipped. The two features worth knowing about: “goal mode,” which lets you set an objective and have Claude keep running until it hits it (Codex released something similar a few weeks ago), and work flows, which let Claude spin up sub-agents simultaneously to parallelize tasks. The whole thesis at Anthropic right now is getting AI to run longer and do more without stopping.

AI companies are starting to look a lot like social media companies. Social media platforms make money by keeping you on screen as long as possible, and time on screen is the currency. AI products make money on tokens. The longer your prompt runs, the more tokens burn, the more you pay. So the incentive is identical. Keep the session going, the agent working, and the clock running.

Anthropic also launched fast mode, which offers 2.5x faster responses for 2x the price. This is like a fast lane for AI, a premium for people who can’t wait.

Then there is the news this week that nobody is surprised by: Anthropic is going public. Everyone saw it coming.

Weekly Shoutout: Andrew Chan

Every week, I want to shout out an entrepreneur, VC, friend, company, or even just a topic that means something to me.

I couldn’t be happier to kick off this series by officially shouting out my good friend Andrew.

Andrew and I met through a mutual friend, started playing Fortnite together at least once a week, and now we are building Social Intelligence Labs side by side.

I took a while to convince him to come on board, but once I got him deep into Claude and he saw what we could actually build together, something clicked. He’s been instrumental in the data infrastructure behind everything at Social Intelligence Labs, genuinely one of the sharpest researchers I’ve ever met. He gets sucked into the work like I do. He works hard and plays hard, too. Insanely smart but also never loses sight of his silly side, which is critical for me. He loves the office. He’s my kind of person.

Outside of SIL, Andrew frequently writes articles about VC and his quest to put the fun in funding with his project, Tiki VC. He covers Silicon Valley and the startup world in a way that’s actually worth reading, which is rare. He also has a TikTok (@AndrewChanVC) that I can only describe as deeply passionate and completely unhinged, and I mean that with full respect. Watch it and you’ll understand immediately.

If you’re in the startup or investment world, follow him. He’s one of the best people I know, and I don’t say that lightly.

Stay golden,

Zac

The fundamental principle for any social media site has remained constant: what is the maximum amount of time we can capture your attention? That has always been the only metric that actually matters: minutes on screen. From that point of view, the question of whether something is real or AI-generated is irrelevant. If people hate AI content, it gets wiped because people aren’t spending time watching it. Duolingo leaned too hard into AI, people noticed, and people left.

If AI content becomes impossible to tell apart from human content, provides the same entertainment, and keeps people scrolling the same way, platforms have no real reason to remove it. The internet has never really been filtered by “Is this authentic?” The real filter has always been, “Do people like this enough to stay?”

The uncanny valley effect is still doing some of the work for us. We are still at the point where, when you see an AI video, many people can still sense the difference. But that instinct is going away. The people who can reliably spot it are the ones who grew up in the window between the pre-Internet world and this one. People old enough to remember when everything online was made by a person, but young enough to have witnessed this shift happen in real time. The generations above us largely can’t tell already, and the generations below us are growing up without the reference point entirely.

Filling this void is the dopamine loop, a system that is fundamentally indifferent to authorship. Recall the rush of a post going viral; that surge of validation doesn’t require the engagement to be “real” to feel good. It’s highly addictive. In fact, some current apps place you as the sole human in a sea of bots. Every like and comment is AI-generated, perfectly tuned to make you feel seen. It works because it fulfills a genuine psychological need, even if the source is an algorithm. The emotional impact is real, even if the interaction isn’t.

These apps offer a glimpse into our future, much like Fortnite did years ago. By seeding lobbies with bots to ensure smooth gameplay, they proved that as long as the experience feels the same, most players don’t care if their opponents are lines of code.

Social media is following this blueprint. A bot-augmented platform ensures every user feels popular and engaged. We are heading toward an internet where human creators coexist with AI influencers who command massive, loyal followings.

There will probably be a platform that emerges and declares itself human-only, and at first, it will do well because people want that. There will probably be another platform that’s openly all-AI. The most interesting platform to me will be the one that blurs the line on purpose and lets people decide how much they want to know.

The Internet Isn’t Dead

The Dead Internet Theory has been floating around for years. The idea that most of what’s online is bot-driven. But “dead” is the wrong word. The internet has never processed more requests, moved more data, or generated more content than it does right now. A dead platform isn’t being used, isn’t growing, and isn’t evolving. By every measurable standard, the opposite is true. Rather than viewing it as a terminal state, consider it a forest where the ecosystem is shifting. While the existing trees remain, we are seeing new life emerging, like mushrooms growing through the roots, building on everything that came before and becoming part of the same ecosystem. It’s more alive than it’s ever been; it just has different species in it now. What’s actually changing is the authorship.

The more honest version of the dead theory is that the human internet is dying. The place of chatrooms, of people talking to people, posting about and for other people. That version of the internet is getting smaller. The people who remember it most clearly are getting older. The people growing up now have always had the curated, optimized, partly-automated version, and they don’t experience it as a loss because they have no before to compare it to. By the time that generational handoff is complete, the question of whether the internet is really “human” will feel as strange and beside the point as asking whether a city is really “natural.”

As that gap widens, human-made content will start to function as something like a luxury, the same way a hand-sewn jacket from a craftsman and a machine-produced one can be functionally identical, hang on the same rack, look the same. The difference is in the label and the documented chain of origin. If you know, it’s worth more. If you can’t tell and nobody’s telling you, it’s just a jacket.

Content will work the same way. The human label will matter to some people and not at all to others, and the system for proving it doesn’t really exist yet.

Building that is going to be one of the more important problems of the next decade.

Stay golden,

Zac

I use AI in a way most people would find surreal. Some of it’s silly, some of it’s genuinely life changing, and a lot of it will sound like science fiction. I want to share all of it with you.

AI adoption is shaping up to be the biggest bottleneck of this entire movement. 1 in 4 Americans interact with AI every day, but most people only use it to chat with LLM’s and don’t understand the intricacies of what AI can actually do for them.

It takes 3 to 6 hours of real experimental time to even glimpse what’s possible. Most people don’t have that window open, or just aren’t ready to embrace it. People are understandably busy with their routines and their lives.

You need to see someone using AI well before the possibilities click. Without a reference point, it stays vague. The intimidation here is fully legitimate. Handing over access to something running on your device is worth being thoughtful about.

AI isn’t going to reorganize your life because you downloaded an app. However, if you give it real time and real curiosity, it can take on a life of its own.

I’m not trying to convince anyone of anything. I just think seeing how someone actually uses this day to day is the only thing that makes it feel real. Hopefully, I can be a reference point:

Morning routine

The first thing I do when I wake up is check Slack without ever opening Slack. Claude has already scanned my Slack, inbox, and GitHub before I wake up. It writes me a TLDR in a Notion page, then organizes everything on my Calendar.

I then sit at my computer and check what was built for me overnight. The night before, I queue up coding sessions for Claude to run while I’m out. Sometimes I wake up to fully built apps or a new set of data from whatever I am researching. Overnight, my Mac mini stays awake using a plugin called “Amphetamine.” This little “drug” keeps my computer going 24/7, creating an always-on server.

For actual work throughout the day, I use Cowork. I drop in a seed idea, it pulls from my old writing in Notion, and builds something that sounds like me. The split is roughly 30% me, 70% AI. Once I have the basics, I go through, edit it, and then have a polished piece of work.

The fun stuff

I also use AI for fun, as a way to experiment, build little worlds, and make myself laugh.

For example, meet Jessica, my concierge. Jessica is an agent I built for one purpose only: calling the hosts at my favorite restaurants and getting me reservations. She has gotten me into Minetta Tavern and Torisi. She says, “My client would like to book a reservation.” One of the best parts of living in New York is the restaurants. One of the worst parts is trying to get in. That’s what Jessica is here for.

When Sarah was having a hard day, I “borrowed” her computer and vibecoded an interface called “NUGGIELAND.” Nuggieland is an emotional support chicken nugget app that delivers compliments and reminds her to drink water. She had wanted an agent center to help her build new things, but I didn’t want it to be like any other agent center. Her office is filled with chicken nugget pillows that make her smile, so I figured, why not digitize that for her?

Then there was the entire Sims-style simulation where all my agents could interact with each other. They go to the cafe together, exercise in the virtual gym, gossip, debate philosophy, and even drift around their digital office while completing their tasks.

The truth of the matter

At the end of the day, about half the code AI produces isn’t actually good, but it still lets me build things I could have never built otherwise. For products that real people use at scale, I still hire designers. Claude can generate, but it can’t truly create. The instinct for how something should flow, what an experience should feel like, that’s still human work and I don’t ever intend to skip it.

Subscribe now

Things that changed my perspectives

There are moments when AI has changed my perspective on what access can look like.

At one point, I found myself in a legal dispute with a designer for my backpack project. It was the kind of situation that would typically cost weeks of stress or thousands of dollars in legal fees. Instead, I fed Claude the entire situation. Ten minutes later, I had a fully written refutation, organized with quotes and evidence pulled straight from the record.

There are other moments when AI has shown me what productivity can look like.

I recently built a system that scanned the entire NYC tech week calendar, filtered through the chaos, auto-scheduled relevant events, and registered me for all of them. I also have a system that cross-references my inbox with X and LinkedIn to surface potential connections.

Life has become much less tedious, but that doesn’t mean I think life should be frictionless. I don’t think humans should ever live frictionless lives. But scrolling through hundreds of event listings or LinkedIn profiles isn’t the kind of friction that matters, at least to me.

Stay golden,

Zac

(actually, Zac. Not Claude. I promise)

Most marketing teams are still operating like sleep-deprived detectives with thirty tabs open, screenshotting analytics, manually piecing together what’s happening online. Even then, you’re mostly guessing, trying to read billions of behavioral signals with human intuition alone.

That’s just the reality. The tools available to understand social behavior haven’t caught up to the scale of the internet.

Some of the most basic questions are still surprisingly hard to answer, questions like why did this post spike? Which trends would make sense for my brand? The most important question that no analytics can account for is: what’s going to happen next?

That disconnect is what led me to start my newest company: Social Intelligence Labs.

The Idea

If you read the last edition, you know all about The Kraken. Marketing Official, Sarah and I became obsessed with tracking virality, but like everyone else, we were tracking everything manually. At some point the obvious question hit: why are we doing any of this by hand? Why does understanding internet behavior still feel this primitive? That became the basis of Social Intelligence Labs.

Subscribe now

The Execution

The raw data exists. There is more publicly available data than ever before, and yet somehow, less clarity. Every second, platforms generate enormous amounts of behavioral information. Most of it sits in the dark because nobody’s processing it fast enough or at enough scale to extract anything useful.

Think of all of the data like crude oil. The raw material is everywhere, but the value comes from refining it. We are building the refinery, the infrastructure that processes behavioral data at scale so that cultural movement becomes trackable and the black box becomes legible.

The Significance

Every brand wants to behave like a creator, and every creator wants to become a brand. Attention has become one of the dominant currencies of modern life, and entire industries exist to optimize it.

The internet changed the scale at which people experience social validation. Historically, your reputation lived within a community of a few hundred people. Now, someone can wake up with three million followers from their bedroom. Our brains can’t fully process those numbers. You get 100 likes on a post and feel disappointed, but you couldn’t even fit 100 people in most rooms.

Brands pay satellite accounts enormous sums to amplify products organically, and companies build entire marketing operations around creators. Algorithms determine what people consume and what people want. At this point, people and brands should have agency over their own content.

The internet shapes culture, commerce, and behavior, whether we understand it or not. Someone is going to build the systems that make sense of it. I want to be the person who understands virality better than anyone else. I want to share that understanding with you all, and turn that knowledge into a tool.

That’s what we’re doing with Social Intelligence Labs. You can even check out one of our demos here!

Stay golden,

Zac

Let’s start from the beginning.

When Sarah and I were building and marketing Official, the tech was our biggest bottleneck. It isn’t fun to think about your system crashing when you’re building and scaling an app. Our servers only had so much space, so when 10,000 people tried to sign up in a single afternoon, the whole thing just shut down.

If we were onboarding more than 100 users per minute, Official would go down full stop. In the early days, we were going viral regularly, which meant we were also crashing regularly.

Many founders would’ve only focused on fixing the tech faster, but Sarah and I stumbled onto something different. I mean really stumbled, this was not some genius calculated moment.

One day, our system collapsed. We had a video running hot, but Sarah had no choice but to set it to private just to stop the bleeding. There’s no strategic logic that gets you to turn OFF the video that’s going viral. You only do it because you’re out of options.

Then, something unexpected happened. The next day, when we switched the video back to public, it went viral all over again. It was reborn. That’s when I started paying attention.

We started tracking our TikTok videos every single minute, cross-referencing view counts with our user dashboard, mapping out our conversion rate in real time. For every 100 views, we were pulling in around four users. I could do the math: a video at 50,000 views meant a few thousand signups incoming, and a few thousand signups incoming meant the system was about to go down.

So, at that exact point, like clockwork, we’d bottle it, set it to private, and then when things chilled out or we needed a growth injection later in the week, we’d flip the viral video back to public… and it would take off again. Every single time.

The Kraken was born.

Anyone who’s tried to raise knows the one thing that matters above almost everything else. The chart has to go up and to the right consistently, week after week. Not one massive spike, up and to the right.

The TikTok spike (to go viral once, get a massive spike, lose everyone, never recover) was our biggest enemy. The Kraken destroyed this risk. It turned sporadic viral moments into something that looked like compounding growth, because in practice, it was compounding growth. We always had a Kraken in the reserve, locked and loaded.

Eventually, we didn’t even need The Kraken. Our tech caught up. On September 19th, 2022, we let a video rip with no throttle, no killswitch, nothing, and got roughly 100,000 downloads in a single day. That’s the same day we reached #2 in the app store

Needless to say, you can draw a straight line from the Kraken to the exit.

Olivia Moore Was Right (And Also Wrong)

Around the time we were scaling Official, Olivia Moore, a partner at a16z, tweeted something that stung.

The gist: TikTok marketing is nothing more than a sizzle in a pan. Sure, it’s exciting for a second, but it’s also gone just as fast. I’ll be so honest, the tweet felt targeted.

Olivia wasn’t wrong. Going viral once doesn’t build a company. Getting 100,000 users today and 300 tomorrow is worse than a slow and steady climb you can actually sustain. Virality without retention isn’t actual growth.

Except, Olivia’s tweet and every investor who nodded along with it was missing something crucial. They were diagnosing the TikTok spike as a failure and concluding the entire marketing strategy was flawed. The actual problem with TikTok marketing is execution. Done wrong, yes, it’s a sizzle in a pan. Done the way Sarah and I did it, with the Kraken and the dashboard and the obsessive minute-by-minute tracking, it’s a machine.

TikTok marketing is now the number one way consumer apps scale. The investors who passed on us because of it watched that shift play out from the sidelines.

Distribution Gets You in the Door

Distribution and retention are completely different games. Most founders confuse them, and that confusion costs them their company.

Distribution is how you get someone to download the app, follow the account, walk through the door, but it cannot save you from a bad product. If anything, it surfaces problems faster.

A good example is the Gas app from Nikita Bier, which exploded through an incredibly engineered school-by-school rollout strategy that created curiosity and FOMO among teens. Private Instagram accounts tied to specific high schools, synchronized notification drops after school, and anonymous compliments helped the app hit #1 on the App Store and generate over $1 million in revenue within months. But virality and retention are different problems.

Users came, looked around, and left. That’s what happens with most apps. Very few get embedded into people’s days. Keeping users has nothing to do with your content strategy. Once someone is inside your product, the algorithm can’t save you. Distribution can drive millions of people into a product very quickly, but whether they stay depends entirely on the product experience itself.

What Official Actually Proved

When I look back at all of it, here are my main takeaways:

The investors who passed on us because of TikTok marketing weren’t stupid. They were pattern matching on every founder they’d seen chase a spike and flame out. They were looking at the sizzle and calling it correctly. For many, that’s exactly what it is. A dying sizzle.

What they couldn’t see was that a sizzle and a Kraken are not the same thing. One is a moment, the other is a system. The difference between the two is whether you’ve built something underneath the distribution that can actually hold what you’re pouring into it. In our case, both the tech and the product.

What we proved, and what I think the industry has since caught up to, is that TikTok is an acquisition channel. One of the most powerful ever built, if you treat it like one. If you track it obsessively, if you understand your conversion rate, if you build the retention layer underneath it, and if you’re willing to get creative, it works.

Every major consumer app scaling today knows this. The strategy we were building in real time, getting laughed out of investor meetings for, is now the playbook.

We always had a Kraken in reserve. Turns out, that was the whole game.

What’s your Kraken?

Stay golden,

Zac

This is my founder origin story.

The moment I learned about the stock market, I was hooked. What could be cooler than owning companies? Being part of massive things that shape everything else, helping steer humanity in actual directions. To me, that was being above a cog in the machine.

Buffett was the godfather of private equity. As a kid, I’d fall down internet rabbit holes researching him. I was intrigued by the sheer number of important companies he owned and helped scale as a strategic thinker. He approached this as a strategic thinker when no one else was. Today, PE is not as wholesome as it once was. Firms buy a business, load it up with debt, cut staff to pay it off, then flip the company in five years. Buffett was much more of a homie to the businesses he bought. And now, Berkshire Hathaway has the most expensive stock in the world. He did all that from his home. Just some guy in Omaha.

I started to build the same way I thought about investing in businesses when I was young. I wanted to create incredible businesses with incredible people.

–

Moving into my teenage years, I would say Iron Man made me who I am. Tony Stark was the ultimate formative influence, truly my favorite founder-type. He’s one of those characters every kid wants to be. What actually got me wasn’t the suit or the weapons. It was watching someone build crazy shit. That’s what the movie is really about.

Tony Stark showed me what it looked like to be so far above the machine that the machine couldn’t function without you. He didn’t bother climbing a ladder inside a system someone else built. He made the ladder irrelevant and invented something so foundational that everyone had to reorient their existence around what he created.

Iron Man built a power source. The arc reactor was the infrastructure that made everything else possible. What’s wild is that, looking back now, Iron Man basically predicted everything happening in AI and computing today. Jarvis is literally a large language model. The difference between me and Tony Stark is that Stark had unlimited tokens. The movie was made over twenty years ago and it was already a window into what would eventually be possible.

It is honestly a lot like what Elon Musk has actually done. People think of him as a car guy or a rocket guy, but that’s missing the core. What he built, over and over, is the infrastructure layer that everyone else depends on. He keeps finding the same leverage point Stark found: don’t build the product, build the thing the product runs on. I carry that with me in everything I create.

–

In college, I thought I was going to go into investment banking. Now I go back to NYU Stern to give talks where I try to steer students away from it. I tell them to experiment and do what makes them happy, because whatever you picture yourself doing in college probably won’t be what you end up doing. You figure that out through real-world experience and trial and error.

I was class president for three years, then impeached my senior year for missing meetings because I was working at a VC fund. I started a small watch business with a friend. After graduating, I stayed at that VC firm, but I was always thinking about the relationship app I really wanted to create.

That app ended up changing the trajectory of my life.

Building Official

I was fresh out of a breakup and eager to build. The problem I’d been grappling with: Nobody teaches you how to have a relationship. You learn about sex from porn and you learn about relationships from rom-coms. Both are terrible teachers. I thought there should be something better, something that actually helps people communicate without having to go to therapy. As a guy in his twenties, I wasn’t going to try couples therapy. So I started thinking about what Duolingo had done with language learning: if you can gamify learning a language, why can’t you gamify learning how to be in a relationship?

I started building on nights and weekends in May of 2020. We incorporated August 6th. I got fired from the VC fund October 10th.

The first name was Newel, “new level,” before we changed it to “Official.” We originally tried to help couples swipe through date ideas and activities, but it was COVID. That was never going to work. So we scrapped it and shifted to helping couples actually communicate.

Official became like the third party in your relationship you could blame everything on. The app would ask each partner a question separately. Your partner couldn’t see your answer unless they answered too. If one of you said no, it never came up. If you both said yes or maybe, we’d surface it. No awkward conversation. No fight before it even started. We had an intimacy side and Newlywed Game-style questions. Everything was built around getting couples to actually talk.

I used savings from college to pay a dev shop to build the first version. After that, I had nothing. So I did what any reasonable person would do and built a whole pyramid scheme of interns.

160 interns over six months. Sarah started as an intern, rose to head of marketing, and became the engine of the whole operation.

Official scaled the way it did because Sarah and I were not afraid to make absolute fools of ourselves online. We were vulnerable and shameless and we had fun. Audiences responded to the unapologetic content, and we went viral over and over again. I get into fights with founder friends who refuse to put themselves out there. To be a CEO, your job is to be uncomfortable and figure out the things no one else can.

The tech was our biggest problem at first. We had great ideas and no infrastructure to support them. We bought and transcribed every source we could find to pull questions. I went through them by hand and picked each one. We had someone hand-draw every sex position and described each in a male voice, female voice, and neutral voice. Everything was done manually, slowly, with basically no resources.

I was Tony Stark in the cave, building something out of scraps and trying to make it work. Looking back, I think that constraint is what made it good. Comfort kills drive. If you have everything you need, you go home at five o’clock. When you’re eating ramen and dreaming about a steak, that hunger is real in a way it just isn’t when you’re comfortable. While you’re deep in the work, it’s miserable. But when you’re out of it, you miss it. It’s that good kind of struggle.

Official really took off with long-distance couples, new college couples, and military couples. People who needed a tool to stay connected across distance or stress. The concept worked. By May of 2023, we were acquired by Bumble.

If I were building Official today, it would look completely different.

I wouldn’t need 160 interns. I wouldn’t need a dev shop, because I can build something better with Claude Code. The content that took teams of people months to produce, I could generate and iterate on in days. All a startup really is, is iterations until you find what works. Finding product-market fit is just an iteration problem. The only question is whether you can get there before you run out of capital. AI compresses the feedback loop, which means a higher chance of making it before you run out of runway.

But I do want to be honest about something. It’s not all sunshine and rainbows. Our friend Claude has become something of a startup killer. Anthropic expands into a new category, and companies built in that space just die. They can’t compete with a model backed by billions of dollars that ships faster than any startup ever could. The speed is the thing that makes this different from every previous wave of big tech fear. Founders have always worried about incumbents. We used to ask questions like, why couldn’t Apple do this? Why couldn’t Tinder just copy us with Official? The old giants moved slowly, and startups could outmaneuver them.

That’s simply not true anymore. Anthropic ships features in an entrepreneurial way that most big tech never could. They move faster than the startups trying to compete with them. That’s the real tension for anyone starting something right now. You’re not just competing with other startups. You’re competing with companies that operate like countries. Trillion-dollar monoliths with the agility of a seed-stage team. You’re also competing against Claude, the thing that makes it easier to build now than ever. It’s a double-edged sword.

My Advice

So that’s something you need to be honest with yourself about on day one: how will you stay a year ahead of Claude? What will happen when Claude ships exactly what you’re building? Because it will. That question isn’t paranoia or overthinking, it’s really the most important strategic question any founder can ask right now.

That said, the opportunities are still there. Markets always shift. The composition changes, maybe you lose enterprise software categories, but you gain consumer categories that didn’t exist before. The market will continue to change, but it won’t stop growing. And some things Claude won’t touch anytime soon: physical products, embodied experiences, specialized services, trust built through community, distribution carved out through relationships. CPG, local, creator-driven businesses, these aren’t going anywhere.

In the year 2000, everyone said every company was an internet company. Over the next twenty years, every company simply incorporated the internet into how they operated. AI is the same wave. It’s not that every startup is an AI company, it’s that every company will use AI the way they use the internet today, as a fundamental part of its core.

So when I mentor founders starting out now, people who want to build something but don’t know where to start, I try to give balanced advice. Yes, this is genuinely the best time in history to start a company. The tools are more powerful, the barrier to ship is lower, and being a founder is one of the few ways left to take real control of your future in an uncertain market. But I’m also blunt about one thing. I will say it over and over again. Marketing and distribution is the only thing that matters now. As Justin Kan famously said, first-time founders focus on tech, second-time founders focus on distribution. In a world where Claude can build your tech overnight, that truth has never been more urgent. The moat isn’t what you build. It’s who you reach, and how you reach them before anyone else can.

That’s all for now.

Stay golden,

Zac

Recently, Golden Egg Media (along with seemingly every other company that touches a keyboard) made the full switch from ChatGPT to Claude. I’ve been bullish on Claude for a while now. It’s not a secret. Learning Claude Code genuinely changed my life and my capability to build cool stuff. Features like Claude CoWork and Claude’s synchronized agents are the future of how humans and machines collaborate. Except today, I want to talk about something that goes deeper than product use, something happening on an interpersonal and even cultural level, and it deserves more awareness and attention than it’s getting.

We Are Simple Creatures

We’ll start on the surface level.

Sometimes, humans just want silly words and fun pictures. We want to enjoy the thing we’re using, and we want to feel something. Google understood this for years. Google made search fun. Playful doodles for obscure holidays. Easter eggs hidden in search results. Google made a utility feel like a personality, and for a long time, it worked. That’s how Bing became so irrelevant.

Claude is bringing genuine playfulness back to tech. While its competitors launched lifeless software demos, Claude showed up with a nice coral color scheme, a little coding crab, and an interface copy that says things like “tinkering” and “twiddling my thumbs” while it works. None of that is accidental. That is a deliberate choice to make you feel like you’re collaborating with a curious, slightly eccentric colleague instead of submitting a request to a server.

ChatGPT is sterile. It’s kind of like the Murray Hill bro of AI. Grok is Bushwick. Gemini is Williamsburg. But Claude? Claude is the West Village of the AI race.

Dario’s Vision and What It Actually Means

Anthropic’s CEO Dario Amodei wrote a blog piece titled “Machines of Loving Grace.” The choice of that title for a document about superintelligent AI says everything about Anthropic’s positioning. They are building something like a compassionate presence. This is a story, and everyone is realizing that now, stories matter more than benchmarks.

Let’s look at how Anthropic started at all. Amodei and several other Anthropic founders left OpenAI. They walked away from one of the most valuable companies in human history because they had concerns about safety and the responsible deployment. That’s a powerful narrative. You can debate whether it’s entirely accurate, and you can debate whether any AI company is truly “safe,” but what you cannot debate is that Anthropic got to that story first. No one else can claim that narrative. They planted the flag. “We are the ones doing this ethically. We are the good guys.” In a cultural moment when people feel anxious, uncertain, and overwhelmed by AI’s speed, that narrative goes a long way.

Personification 101: The Name Is Doing Work

Let’s talk about the name “Claude” for a second, because I don’t think people give it enough credit. It is a distinctly human name. A gentle, slightly old-fashioned, intellectual name. It makes you think of a professor.

Try to do that with “ChatGPT.” You really just can’t. It reeks of product. “Gemini” is a constellation, which is evocative but cool and distant. “Grok” is aggressively weird. “Claude” is a person you might actually want to know, and likely a name that doesn’t remind you of someone you already actually do know. It’s not a common name, but it somehow feels familiar. Before the model even generates its first token, the branding has already shaped your emotional expectations. That’s some expert-level positioning.

The Fun Factor, And Why OpenAI Is Losing It

OpenAI is iterating while Anthropic is constantly inventing. Claude Code came first, and that feature alone rewired how developers think about AI assistance. The CoWork feature is the kind of thing that should be front-page news. The synchronized agents are equally insane. It feels like I need to update my Claude settings every three hours just to keep up with what’s new. There is a velocity signal that users pick up on. Frequent, meaningful updates communicate that a team is alive, engaged, and genuinely building something.

Being first to capture narrative and public respect goes a long way, even though OpenAI was first to market. Claude will always be known as the smart one, the one that does your literal work, and the one that can actually code. This will be “Claude’s thing” in the cultural memory, the same way the iPod was the most iconic MP3 player even though it wasn’t the first.

The Marketing Is No Joke

The advertising itself is remarkable. Here are some examples…

This video is a hypothetical scenario where different AI models are on a skydiving plane and have to vote on who to throw out. Everyone shades on Grok, then at the end, Claude heroically volunteers to sacrifice itself. It’s a little absurd, but it communicates the core Anthropic value prop, that Claude is oriented toward being genuinely helpful, even at cost to itself.

Another example is a Claude video ad about a cougar dating site. It opens with what looks like a therapy session, but just as you settle in, it pivots completely to a dating site for older women seeking younger men. The audience doesn’t know what to do. Then the reveal comes in, subverting expectations. It says, “Ads are coming to AI. But not to Claude.” In one stroke, it positions Claude as the trustworthy, non-commercial alternative while also being funnier and more self-aware than anything OpenAI has put out.

There is serious money behind this.

Claude is working on campaigns with million-dollar budgets. This is a heavily funded effort, from Anthropic itself, but also from discreetly partnering with influencers and storytellers to promote what pushes its narrative. That doesn’t mean the organic love is not also real. The product actually works. You can spend all the money in the world and not manufacture the kind of genuine enthusiasm people express about Claude. Claude the friend. Claude the colleague. Claude the assistant. Claude the powerful.

The Mythology of Mythos

There is something deliberate in the way Anthropic names its models. Opus. Sonnet. Now, Mythos. These are not version numbers like GPT-4. These are deliberate words drawn from the human tradition of making things that matter. Opus means masterwork, and Sonnet is a poem of constraint, suggesting that what Anthropic is building belongs in a longer, grander conversation about history and creative achievement.

The announcement of Mythos was its own kind of myth-making. They built something so powerful, they said, so genuinely dangerous to global cybersecurity, that releasing it to the public would be irresponsible. “We made this. We are choosing not to give it to you. For your own good.”

It’s like a bomb manufacturer building and selling bomb shelters. If Mythos really is so powerful, then you’ll have to protect your software from Mythos, with Mythos.

It is a genius posture for a tech company. In most industries, withholding your best product would be considered a failure. Here, it becomes a demonstration of virtue. The restraint is the message. “We are the company that can build Mythos and choose not to deploy it. We are the adults in the room. Trust us with the future, because we are already protecting you from it.”

The Deeper Game

We have always framed the AI race as a contest of benchmarks. That framing made sense early on when models varied so much in capability. We’re entering a different phase now where the models are all powerful and all impressive. When everything is impressive, impressive stops being a differentiator.

The next stage of this race may look less like an engineering competition and more like a brand competition. Anthropic already knows this. They told a warmer story, picked a human name, and when they built something they called too dangerous to release, they made sure everyone knew about it. The company that credibly positions itself as the careful, ethical actor in the space is the one creating the emotional permission for people to adopt the technology.

The most powerful thing Anthropic ever shipped was a narrative, not a model. That narrative is already playing out.

Content is just one form of entertainment. I think we witnessed the death of “social” media a long time ago. What we have now is something much closer to entertainment media. The original intention of social media was keeping up with friends, sharing moments, and building relationships online. The last platform even close to that was Snapchat. That’s not what most people are doing anymore. Today, the feed isn’t optimized for seeing what your cousin’s baby is up to. It’s optimized for strangers who hold your attention, and attention is earned the same way it always has been.

When something goes viral, it’s because, at some level, it’s tapping into the same evolutionary need for entertainment that we have had since ancient cave drawings. It is entertaining enough to make you stop thinking, stop scrolling, enough to make you send it to someone, enough to make you recreate it, remix it, turn it into a trend.

So I’ve been thinking a lot lately about another question that feels bigger than content, social media, or virality itself: Why do humans need to be entertained at all? Why exactly do we invest so much of our time, emotion, money, and sense of self into entertainment?

The easy answer is the dopamine cycle, but I honestly think that’s a cop-out response. I want to talk about this in relation to comfort and convenience. For most of human history, days were filled with tasks that demanded strain and focus. You had to find food, build shelter, carry water, fix what was broken, move your body, and spend actual energy just getting through the day.

Now, so much of that friction is gone, and things are only going to get more frictionless with AI. We are already in, and moving towards, a world of convenience. Food arrives in ten minutes. Apps are built in a day. We spend our days at desks and in beds. I think our need for constant entertainment and stimulation is directly tied to that shift. When life stops demanding your full attention, your mind starts looking for something else to fill the space.

I keep coming back to this idea that the more comfortable life becomes, the more stimulation we seem to need. When you’re physically exhausted from doing real tasks, your threshold for entertainment is much lower. When life becomes more frictionless, the opposite happens. Your mind is still searching for input, movement, novelty, something to grab onto. So you reach for something to entertain or distract you.

The Evolution of Entertainment

Entertainment began where civilization began, people sitting and telling stories around the fire. Humans gather in circles to make sense of the world around them. As societies grow, entertainment moves into the public sphere. In ancient Greece, tragedies and comedies were born. So were the beginning of sports and the Olympics, which evolved into the stadiums and Super Bowls of today. In Rome, the Colosseum gathered thousands to watch tension, status, and violence go down in real time. From the beginning, people were drawn to high stakes and spectacle.

The Middle Ages brought festivals, music, and poetry. The Renaissance elevated entertainment into art forms like opera, and ballet, making entertainment more about emotions. The printing press marked the next major leap. For the first time, stories and ideas could scale. Books and newspapers allowed entertainment to travel into everyday life. Then came the movies. People no longer gathered in arenas, but in dark theaters. Radio brought entertainment into the home. The television became the center of a household, like the modern fire, a place people gathered at the end of the day.

Then the internet changed the pace entirely. Streaming multiplied choice and YouTube made it so anyone could become a creator. Now we live in the era of the infinite feed. Entertainment no longer has a beginning or end. There is no final page, only more to see. The technology changes, but the human needs underneath it stay consistent. We want to witness drama, stakes, movement, transformation, and now, we’ve taken that impulse and compressed it into the feed.

The algorithm is built to answer the same question the Colosseum once answered, about what will keep people watching. That’s where this starts to connect back to virality. The things that go viral are rarely random. Virality, in a lot of ways, is just 21st century spectacle.

The more content there is, the more important it becomes to understand what humans are naturally pulled toward. AI is only going to accelerate that. Something like AI Fruit Love Island can get made in less than an hour and get over 300 million views. Anyone can make almost anything now, which means the supply of content is exploding, but time is still fixed and attention is still scarce.

Anyone will be able to make almost anything. That’s true for content, and it’s just as true for people building companies. Barriers to There is no end point to the scroll. Entertainment is endless now and there is no final destination, only more to consume. Every single day, 720,000 hours of content is added to YouTube. That’s around 82 years. So, every day, a lifetime of content is uploaded to YouTube alone. 34 million videos are uploaded to TikTok every day. There is no end point to the scroll, no point A and point B.

In a world of infinite content and infinite ideas, the only way to stand out is through understanding attention, virality, marketing, and distribution. To make sense of how we consume media today, we first have to understand who we have always been as consumers of media across time.

Stay golden,

Zac

So, here we are.

First, I want to give a life update and get us all up to speed.

This last year has been a whirlwind.

After selling Official, doing my time, and then leaving Bumble in March of 2025, I found myself in a position I hadn’t been in for a long time. I’d spent years jumping from one thing to the next without really stopping to ask what I actually wanted to build next. For the first time in a while, I had space.

Naturally, I did the exact opposite of slowing down. I didn’t know what to do next with my life, so I tried to fill it with new projects.

I started my quarter-life crisis by traveling across Asia with Sarah for 2 months. We skied through Sapporo, went to Singapore for Chinese New Year, then Vietnam, Thailand, Hong Kong, Shanghai, Taiwan, and back to Japan. Some of it was pure fun, some of it was work, most of it ended up being a mix of both. That trip ended up mattering more than I expected. It was, unintentionally, the beginning of Golden Egg Media.

Just before Japan, we landed our first deal through a company called Kabuk. The idea was to make content, help with distribution, and connect some of those stories and brands back to the U.S.

When I got back to the U.S., I was still just as lost and confused. So, I did what I always do when I’m trying to figure something out. I started building.

One of the first things I worked on was a company called Snowbird, which I started with Duncan, a previous advisor of mine at Official, and a name you will hear much more about over the coming months. Snowbird was a home services business for high-net-worth homeowners, basically a premium, always-on handyman / household support layer for people who wanted one trusted system for handling the chaos of owning a home. My dad even joined as an advisor to help us build. We got a few clients pretty quickly, which was validating, but we ended up shutting it down in September. At some point it became obvious that even if the business worked, it wasn’t going to scale in a way that made sense. Sometimes that’s the whole lesson.

Then, as all good finance majors do, I had a brief stint roleplaying private equity. I invested in a company called FlexTrip, started by Tyler Moody, who built a platform to help students study abroad, and even joined as a partner to help him scale it. Over the last year we’ve nearly doubled the business and are now starting to expand into new verticals.

In May, they shut down Official, and I fell to my “all is lost” moment. I had never been so depressed. This was the lowest of lows. My curiosity about virality was the only light at the end of the tunnel. At the same time, we were building and scaling Golden Egg. Suddenly, more and more of my time was spent in the marketing world.

Over time, something became clear. Our strength was understanding attention, engineering momentum, and getting ideas to move. All we needed to do was lean into that.

From that point, things started accelerating. We started working with larger startups and legendary VCs. We helped on campaign distribution, positioning, and narrative work. In the last year alone, we’ve worked with some incredible companies, contributed to a Super Bowl commercial, and our CPG product even landed on Good Morning America. Don’t worry, we will come back to all of this soon.

But enough about me. If I take a step back, we are in a moment where it has never been easier to be an entrepreneur. You can build faster than ever, there are millions of people accessible at your fingertips. The question isn’t really “can you build something,” but now, “how can you get that something OUT there,” moving, and getting people to care.

So that’s why we are here.

What is K-Factor Anyway?

I first heard of K-factor during the spread of… stick with me… COVID.

During the pandemic, the entire world was watching virality unfold in the most literal way. How does something move from person to person? What makes it accelerate? What makes it flatline? Everyone was thinking about this, whether they realized it or not. At that same time, I was starting Official and reading books like The Cold Start Problem, Zero to One, and Nail It Then Scale It.

So on one side, the world was talking about a literal virus, and on the other, I was reading about scaling startups. Eventually, those two things just merged into the same framework in my head. A virus is obviously not the same thing as an idea, but the underlying questions are surprisingly similar.

At its simplest, the K-factor is a measure of how effectively something spreads. For every one person who touches a product, idea, company, piece of content, or message, how many new people do they bring in? If each person reliably brings in more than one additional person, you get compounding growth. If they don’t, growth slows down or dies. Once you start seeing the world through that lens, it becomes weirdly difficult not to see it everywhere.

We’re in a world now where information is abundant. You can ask AI for the “best” anything, and you’ll get a clean, optimized answer, but at a certain point, everything starts to converge. The K-factor is what breaks that convergence. It explains why one idea keeps resurfacing until it feels unavoidable, why one product suddenly shows up in multiple group chats, and why something goes from obscure to everywhere almost overnight.

Right now, I’m spending time playing around with Claude Code and stepping back a bit from the day-to-day of Golden Egg to work on something new. It’s time to get back into the arena and start playing the game. For the first time in a while, I am going all in.

I want to bring you along as I build, experiment, and figure things out in real time.

Stay golden,

Zac