Compare revisions
Version 1 to version 2
The Product Died. The Growth Hacker Won.
Version 1
July 26, 2026
Initial publication examining the suspicion that famous growth operators succeed through privileged platform access, and what happened when that claim was tested against the documented history of Nikita Bier, tbh, Gas, and X.
Version 2
July 26, 2026
Rebuilt the essay around the documented separation between engineered virality and durable products, the rapid shutdown of both tbh and Gas after acquisition, the economics of acquihires, teen social graphs, and the lack of independent evidence that X has already been turned around.
Structure changes
How the article changed as an object, not only as a sequence of sentences.
1 changed measures
Words
+356
Metadata changes
Title
I Wanted the Real Growth Hack to Be Knowing Someone
The Product Died. The Growth Hacker Won.
Description
I suspected famous growth operators succeeded through a hidden human API. Trying to prove it with Nikita Bier produced a less convenient answer: the access is real, but the tactics were too.
Nikita Bier built two apps that reached number one, sold them to Facebook and Discord, and watched both disappear within months. The more interesting product may have been the public proof that he could manufacture growth.
Format
Raw Drafts / Growth / Social Capital / Platforms / Startup Culture
Raw Drafts / Growth / Consumer Social / Acquisitions / Platforms
Article changes
Visual blocks are rendered as previews; other blocks retain the editorial diff.
92 changed blocks
I started by asking who the most renowned growth hackers were. The usual names appeared: Sean Ellis, Andrew Chen, Brian Balfour, Elena Verna, Casey Winters, Alex Schultz. Then I remembered the person who had recently become much more visible at X, Nikita Bier, the founder of tbh and Gas who now runs product there.
I wanted to know what people like this know.
The respectable answer is that they understand retention, network density, notifications, activation, social psychology, distribution loops and the difference between acquiring a user and creating a product that produces the next user. That answer is probably true. It is also slightly unsatisfying because all of those ideas are available in public. There are interviews, essays, podcasts, conference talks and entire courses explaining them.
So I asked a less respectable question.
Do they sometimes just know someone?
I had a very specific version of this suspicion in mind. I had recently watched an AI coding agent complete almost the entire process of building an iOS app and then stop at App Store Connect because Apple's Create New App action existed as a popup rather than a stable page. The agent guessed that appstoreconnect.apple.com/apps/new should exist. Apple loaded the frame of the application around an empty page. The real form appeared only after returning to the Apps screen and clicking a small plus button.
I spent time inspecting the interface, comparing browser states and trying to understand why a task that should have been administrative had become the single point of failure. It was easy to imagine that somebody more connected would not do any of this. They would message a person at Apple. The person would explain the correct path, escalate a problem or quietly move something through the system. The unknown developer would keep submitting forms. The known developer would use a second interface made of people.
The phrase I came up with was the human API.
Platforms present themselves as software. There is a review queue, a policy document, a support form, an advertising dashboard and an algorithm deciding what receives distribution. But large platforms also contain account managers, editorial teams, partner programs, former colleagues, investor relationships and private conversations. The public interface may be the same for everyone while the recovery path is not.
This is not an invented phenomenon. Meta's own Oversight Board examined the company's Cross-Check system, which gave certain prominent or commercially important accounts an additional moderation process that ordinary users did not receive. The board described unequal treatment, unclear criteria and a system structured partly around Meta's business interests. Research on entrepreneurship also repeatedly finds that social capital affects access to finance, information, markets and other resources.
The hidden layer exists.
I wanted Nikita Bier to be a clean example of it.
That would have made the story simpler. He had built two social apps that reached the top of the App Store, sold one to Facebook, sold another to Discord and eventually posted his way into running product at X. Perhaps the growth tactics were merely the visible explanation. Perhaps the real advantage was a person inside Apple, Meta or some distribution channel who could produce the first burst of attention and then let everyone call it organic.
I tried to find that person.
What I found instead was an annoyingly detailed launch strategy.
tbh did not begin with a broad national release. According to accounts from the time, Bier's team launched at one high school in Georgia, where roughly 40 percent of the school downloaded the app on the first day. The product depended on people already knowing one another, so a random collection of users would not work. The team expanded from one dense network to a few nearby schools and eventually hundreds of schools.
The more aggressive part of the strategy involved school-specific Instagram accounts. The team created private accounts for individual high schools, followed students from those schools and left the accounts private long enough for curiosity to accumulate. Around the end of the school day, the account became public and revealed the app. The launch was not waiting for teenagers to discover a new product in the App Store. It constructed a small social event around a group that already had relationships, gossip, status and reasons to care what other people thought of them.
This is a growth hack in the original sense of the term. The distribution mechanism and the product were almost the same thing. The app was valuable only inside a sufficiently dense social graph, so the launch manufactured density before opening the door.
There was also a less glamorous fact. Bier had not simply woken up and made tbh. He had spent years building and shutting down other products. A University of California, Berkeley account of his talk said he had shut down fourteen apps and was considering liquidating the company shortly before tbh worked. The famous success arrived after a pile of failures that disappeared from the simplified story.
Gas later used a closely related positive-polling mechanic and was acquired by Discord. By then Bier was no longer an unknown founder. He had a Facebook acquisition behind him, years inside Meta, a public reputation for consumer social products and a network of founders, investors and platform people. It would be absurd to claim that he launched Gas under the same conditions as someone making their first app from a bedroom in Viljandi.
But different conditions are not evidence of secret unfair treatment.
This was the point where the theory became harder to write responsibly. I could document that Bier had accumulated reputation and access. I could document that platforms create special pathways for some users. I could document that relationships change what information and resources entrepreneurs can reach. I could not document that Apple featured his apps because of a personal connection, that Meta bent a rule for him, or that his original growth came from a hidden favor.
The evidence supported the system-level claim and not the person-level accusation.
This distinction is easy to lose because plausibility feels almost as satisfying as proof. A person who has sold a company to Facebook probably can reach people I cannot. A person followed by product leaders, venture capitalists and technology executives probably receives replies that an unknown developer would not receive. Once Bier began publicly offering Elon Musk product advice and asking to run Twitter's product organization, the request itself reached the person who could eventually give him the job. This is relationship capital operating in public.
Still, it does not follow that the relationship created the ability being rewarded.
Skill and access are not competing explanations. They can form a loop. Someone develops an unusual ability, gets one result, meets people because of the result, receives better opportunities because of those people, produces more visible results and is then remembered as if the entire sequence came from a stable personal quality called genius. The network becomes part of the person's operating environment. Eventually they may stop noticing it because sending a message to someone inside the platform feels like a normal step.
For everyone outside that environment, the step does not exist.
This is probably where the unfairness is real, even when no rule is secretly broken. One developer receives a generic rejection. Another receives context. One founder reads the documentation after a policy changes. Another hears about the change before it is announced. One person tries to identify the right journalist. Another already has a conversation with them. One application enters a queue. Another arrives attached to a trusted introduction.
None of these advantages has to be corrupt to matter.
The public version of a growth story often removes them because they are not very teachable. A case study can tell us to improve the invitation flow, launch community by community and send notifications tied to real social events. It cannot give us six years at Meta, a previous acquisition, investor credibility or the phone number of the person controlling a distribution surface.
At the same time, focusing only on the invisible advantages creates another kind of distortion. It lets me avoid the part that is available to study.
I wanted the real growth hack to be knowing someone because that answer protects me from a more difficult comparison. If tbh succeeded because of a secret contact, I do not need to think seriously about why one school mattered, why four in the afternoon mattered, why a private Instagram account mattered, why anonymous positive polling created an emotional reason to return, or why fourteen previous attempts had failed. The hidden favor would explain the result and excuse me from learning the mechanism.
The documented mechanism is more annoying because it looks replicable only at the level of principle. I can understand that a social product needs density without knowing which community will care. I can understand that curiosity can be staged without knowing when staging becomes spam. I can understand that a notification should reveal a real social event without knowing where the line between relevance and manipulation is. The strategy is visible, but judgment remains scarce.
There is also something emotionally convenient about turning successful people into beneficiaries of a rigged system. Some systems are rigged. Status compounds. Platforms do not treat every account equally. Investors, journalists and executives respond differently to someone with a known name. Pretending otherwise would be naive.
But suspicion can become its own flattering mythology. The successful person becomes less skilled than they appear, while I become someone who would have achieved the same result if only the gatekeepers had answered my messages. That story may contain some truth. It can also prevent contact with the evidence that does not flatter me.
I began with the image of two founders submitting identical apps. One fights an automated interface. The other texts a former colleague and gets featured. This probably happens in some form. I still do not have evidence that it explains Nikita Bier.
What I have is a messier sequence. Years of failed products. A launch compressed into one school. Private Instagram accounts. A reveal timed around the school day. A product designed around existing social relationships. A Facebook acquisition. Insider experience and a larger network. Another app using a related mechanic. Another acquisition. Years of publicly giving product advice to the owner of X. Then the product job.
The connections are in the story. They are just not where I initially wanted them to be.
They appear after the first proof of ability and then begin compounding with it. They change the conditions of every later attempt. They probably make some failures easier to recover from and some opportunities easier to reach. But they do not erase the work that made the network interested in the first place.
Maybe the deeper problem is that we keep asking for one cause. Was it product skill or access? Was the growth organic or engineered? Was the opportunity earned or granted through relationships? In practice, the categories leak into one another. A launch can be carefully engineered and still produce genuine enthusiasm. A relationship can be earned through previous work and still create an unequal advantage. A platform can apply the same written rules while offering radically different levels of human help.
The human API is real. Some people have more permissions.
I just cannot use that general truth to accuse a specific person of calling an endpoint I cannot prove they called.
I wanted to find the person inside Apple. I found one school in Georgia, a private Instagram account, a four o'clock reveal and fourteen dead apps.
The unfair part may not be that somebody secretly opened the first door. It may be that after the first door opens, the building begins to recognize you.
I started this by asking about the guy who was turning around X. The phrase already contained a result I had not checked. Nikita Bier had become Head of Product, people were excited about him, product changes were arriving quickly, and his previous reputation made the turnaround feel almost pre-approved. He had built tbh and Gas, taken both to number one in the App Store, sold one to Facebook and the other to Discord. Of course this was the person who could make X grow again.
Then I looked for the turnaround.
Bier joined X at the end of June 2025. In that same month, Similarweb estimated that X had 132 million daily mobile users, down 15.2 percent from the previous year, while Threads was growing quickly behind it. By January 2026, Similarweb estimated 125 million daily mobile users for X and 141.5 million for Threads. X still had a much larger web audience, around 145 million daily visits, and independent data also suggested its remaining mobile users spent far more time there than Threads users. X was not disappearing. But the available numbers did not describe an obvious turnaround either.
This does not prove that Bier is failing at X. A global social network cannot be attributed to one product leader after a few months, especially when its owner can change priorities in public at any moment. It does mean I had repeated a story before there was enough data to support it. The growth expert arrived, and I mentally added the growth.
That mistake became more interesting after I researched the two products that created his reputation.
tbh launched in 2017 and spread through American high schools with an almost absurd level of precision. The team picked one school in Georgia because its school year began early. Around 40 percent of the students downloaded the app on the first day. The team then moved through nearby schools and later opened states one at a time, keeping each new market dense enough that the app did not feel empty.
The launch method has become product folklore. The team created private Instagram accounts for individual schools, followed students whose profiles mentioned those schools, let curiosity accumulate, then made the account public and revealed the App Store link around four in the afternoon as students were leaving school. The product was anonymous positive polling, but the launch was also a poll in another sense. Could one school be made to care about one thing at the same time?
It worked. tbh climbed to number one, reached millions of users and created a competitive acquisition process. Facebook bought it in October 2017. In July 2018, Facebook shut it down because usage had become too low.
Five years later Bier built Gas, a closely related product around anonymous compliments inside schools. It required a different technical and distribution system, and Bier says the team went through nine launches and several names before it worked. Gas reached one million daily active users within two months. By the time Discord bought it in January 2023, Sensor Tower estimated 7.4 million installs and almost $7 million in consumer spending.
Discord shut Gas down that November.
Two separate apps went to number one, were acquired by major social platforms and disappeared within roughly ten months of each acquisition. The obvious interpretation is that the products failed after the founders sold them. That interpretation is not quite right, because it assumes that keeping the standalone products alive was the main thing the buyers had purchased.
Discord's internal explanation was unusually clear. According to a memo reported by The Information, the company bought Gas primarily for its engineering talent. Discord intended to keep the app operating until growth slowed, which it expected because the product was seasonal. Afterward, the team would apply its growth and friending mechanics inside Discord. By the final month before the shutdown announcement, Gas had fallen from millions of downloads near its peak to only a few thousand new downloads.
The app was temporary. The capability was portable.
This is where the article I first wrote went wrong. I was searching for a secret relationship that might explain the success, someone at Apple or Meta who had pushed a button unavailable to ordinary developers. I could not find evidence of that. The stronger explanation was sitting in Bier's own interview about his work.
He separates growth from durability much more explicitly than most people who celebrate him. In a long conversation with Lenny Rachitsky, Bier describes durable consumer social retention as something close to a black swan event. The lasting communication network may appear once in a decade and contains a great deal of randomness. Growth, by contrast, can be a science. He believes a skilled operator can make an app spread with much more certainty than they can make it remain part of people's lives.
That is a very different definition of product success.
Most product writing quietly combines several achievements into one word. A product succeeds when people discover it, receive value, return, pay, recommend it and continue using it over years. Bier pulls those pieces apart. He is unusually good at creating the discovery, invitation and emotional-response machinery. He does not claim that this machinery automatically produces a durable institution.
He is also unusually candid about what he enjoys. His favorite part is building something with a tiny team and watching it take over the internet almost immediately. He does not especially want to spend ten years turning that spike into a large organization. In the same interview, he compares the proceeds founders sometimes receive after an IPO and years of dilution with what his teams can make from an app built and sold in about ninety days. For him, the tradeoff is not embarrassing. It is rational, fun and financially successful.
Once I took that seriously, the two dead apps stopped looking like inconvenient exceptions to the story. They became the center of it.
Perhaps the standalone product was never the most valuable product. The more valuable product was the proof that this small group of people could repeatedly create a consumer spike. The app functioned as a public demonstration. It showed that the team could identify latent demand, compress a network into a launch window, tune every invitation and notification, keep servers alive under sudden load, monetize curiosity and defend the product when a false trafficking rumor threatened to destroy it.
Facebook did not need tbh to remain a separate icon forever to learn from the team. Discord explicitly moved the Gas team's growth and friending knowledge into Discord. Bier later became an advisor and investor around other number-one consumer apps, joined Lightspeed as a product growth partner and eventually took over product at X. The products disappeared, but every disappearance left his personal distribution network larger than before.
The founder retained the compounding asset.
This pattern is not unique to him. A 2024 study of acquisitions by Alphabet, Amazon, Apple, Meta and Microsoft found that most acquired products in its 2015 to 2021 dataset were later discontinued. Research on acquihires describes a market where the team, knowledge and option value can matter more than the original service. The Federal Trade Commission has also documented hundreds of small technology acquisitions that fell below normal reporting thresholds, including deals aimed at patents, assets and entire teams of technologists.
Silicon Valley has a market for products that do not need to survive as products.
That market changes what a rational founder might build. A durable company needs years of retention, governance, trust, maintenance, customer support and changing user needs. A convincing acquisition target may need a much shorter sequence: a sharp behavioral insight, rapid growth, visible chart position, some revenue, a tiny respected team and enough urgency that a larger company worries about missing the moment.
Bier appears exceptionally good at the second sequence. This is not the same as faking success. tbh and Gas were genuinely used by millions of teenagers. The compliments were real. The engineering load was real. Gas generated millions in consumer spending. The teams solved difficult problems under extreme time pressure. But the value captured by the founder did not depend on the products becoming permanent homes for their users.
The users and the founder were playing on different time horizons.
For a teenager, Gas was an app about finding out what classmates thought about them. For the company, each invitation also increased network density. Each anonymous compliment produced a notification. Each notification created curiosity. Gas offered a subscription called God Mode for $6.99 per week, giving hints about who had selected the user. More than 95 percent of users remained on the free version, according to Bier, but the commercial tension was still obvious: the product created uncertainty about social approval and sold partial relief from that uncertainty.
An Education Week article asked whether this was exploiting teenage insecurity. That is a fair question, even though it is not the whole product. Bier says the team deliberately prevented cruelty, allowed no free-text abuse and adjusted the system so that people who had not recently been chosen appeared more often in polls. He says users wrote to the company about the app helping them through moments of self-harm. The positive design was not necessarily fake. The uncomfortable part is that positive intent and behavioral extraction can occupy the same interface.
The choice of teenagers was also not incidental. Bier says his teams observed that invitations per user fell by about 20 percent with every additional year of age from thirteen to eighteen. Teenagers see one another every day, have rapidly changing social graphs and are more willing to pull friends into a new product. Adults are harder and more expensive to move. A high school is not merely an audience. It is a preassembled distribution system with classrooms, lunch breaks, gossip, crushes, status anxiety and a synchronized end to the school day.
The genius is real. So is the asymmetry.
A founder can enter a school for a season, turn its relationships into a growth curve, sell the company and move on. The students cannot take their social graph with them when the app closes. The buyer can take the team. The founder can take the reputation. The user mostly keeps a memory of an app that no longer exists.
This finally brought me back to the question about knowing someone. Connections did matter, just not in the cheap form I first imagined. Bier did not begin tbh as a completely isolated developer. His company had raised a seed round years earlier from investors including Greylock partner Josh Elman, Bee Partners and Indicator Ventures. Before consumer apps, he had been through Techstars with a government-software company. When tbh exploded and its infrastructure bill threatened the remaining cash, he could rapidly assemble more financing and enter conversations with potential buyers.
None of this proves unfair treatment. It shows that there is an acquisition market around the visible App Store market. Most developers see the chart as the destination. A connected and credible founder can see it as a signal sent to buyers. The number-one position is not only user acquisition. It is deal flow.
After the first acquisition, the difference becomes much larger. Gas was built by someone who already understood Meta internally, already had a successful exit, already knew founders and investors, already had a public audience and already knew that a short, intense consumer wave could be converted into a transaction. The second app did not need to prove that Bier was lucky once. It proved that the performance could be repeated.
This may be the deeper growth hack: not knowing one person, but knowing how to create a spike that the right people cannot ignore, then knowing which market will buy the spike before it decays.
It is easy to react to this morally. Perhaps the products should have lasted. Perhaps Facebook and Discord should not buy apps they expect to shut down. Perhaps teenagers should not become the raw material for growth demonstrations. Perhaps founders should be judged by the institutions they leave behind rather than the velocity of their exits.
I feel some of that. I also think the cleaner criticism would miss what Bier is actually doing. He is not secretly claiming to have built Facebook twice. He openly distinguishes virality from durability. He has found a place in the technology economy where the ability to generate attention is valuable even when the attention does not remain attached to the original object.
The rest of us are the ones who keep confusing the spike with the institution.
That confusion is already happening again at X. Bier may improve the product substantially. He may find new creation loops, make notifications more relevant, repair discovery, grow video, improve onboarding and eventually reverse the independent usage trends. But the public story that he is turning X around arrived before the turnaround data. His previous spikes have become a prediction attached to his name.
X is also a different test from tbh and Gas. It does not need to become number one for a season. It is already a global network with political power, old habits, creators, advertisers, bots, regulators and an owner who behaves like a parallel product organization. The challenge is not creating a burst. It is making an institution durable again.
That is the part Bier has described as much less controllable.
I began by looking for the hidden person inside the platform. More research led to a stranger conclusion. The product may be the thing that users download, but the durable product can be the founder's demonstrated ability to make people download it. The app can disappear while the demonstration keeps producing opportunities, acquisitions, advisory work, investor access and executive jobs.
The product died. The growth hacker won.
Connected reading
- [The Last Obstacle Was a Popup](/blog/the-last-obstacle-was-a-popup) - [The Interface Is a Secret Handshake](/blog/the-interface-is-a-secret-handshake) - [The Problems Companies Admit They Have](/blog/the-problems-companies-admit-they-have) - [The AI Was Easy. Understanding the Company Was Hard.](/blog/the-ai-was-easy-understanding-the-company-was-hard) - [100 Apps, €0?](/blog/100-apps-zero-euros)
- [The Last Obstacle Was a Popup](/blog/the-last-obstacle-was-a-popup) - [The Interface Is a Secret Handshake](/blog/the-interface-is-a-secret-handshake) - [100 Apps, €0?](/blog/100-apps-zero-euros) - [The Problems Companies Admit They Have](/blog/the-problems-companies-admit-they-have) - [The Product Is No Longer the Product](/blog/the-product-is-no-longer-the-product)
AI authorship note
This essay began with Kris Haamer asking whether renowned growth operators sometimes benefit from knowing people inside platforms, then challenging whether that could responsibly be said about Nikita Bier. AI supported source verification, the distinction between system-level evidence and person-specific claims, drafting, editing and the publishing workflow. Kris supplied the App Store Connect experience, the skepticism and the editorial direction.
This revision began after Kris Haamer rejected the first article as too thin and asked for deeper research. AI reviewed Bier's long-form interviews, the launch and shutdown histories of tbh and Gas, acquisition reporting, research on product discontinuation and acquihires, criticism of Gas's monetization, and independent usage estimates for X and Threads. Kris supplied the original suspicion, the pressure to verify it, and the decision to replace the safer argument with a more structural one.