Million Dollar Content Experiment: How did X use real gold and silver to verify the future of creation?

Author: David, Deep Tide TechFlow

In mid-January, X announced a $1 million reward for the best long-form article on the platform.

Elon Musk personally retweeted to confirm. The rules are simple: US users only, original English articles over 1,000 words, mainly ranked by exposure among US paid users.

You may recall that a few days before this content incentive campaign launched, personal growth blogger Dan Koe published an article titled “How to fix your entire life in 1 day,” which received 170 million impressions, becoming the most successful article in X history.

Clearly recognizing the traffic potential of long-form content, X quickly responded by lowering the article creation threshold, adjusting algorithm weights to prioritize long articles over short posts, and announcing a $1 million writing contest.

Over a two-week competition period, tens of thousands of participants entered.

On February 4th, the results were announced, with a total prize pool of $2.15 million—more than double the promised amount. The first place awarded $1 million, second place $500,000, plus a $250,000 “Creator Choice” award and four $100,000 honorable mentions.

The winners are roughly as follows:

You can see that Dan Koe made the list again. However, his previous article “How to fix your entire life in 1 day” had 170 million impressions, but the winner of this contest only had 45 million.

Viral hits are still unpredictable, but some winning articles are worth analyzing.

🏆 Champion: “Deloitte, a $74 billion toxic tumor spreading across America” — a small account with 90,000 followers takes $1 million using a self-built database

The champion @beaverd’s article title translates to “Deloitte, a $74 billion cancer spreading across the US.” It’s about the well-known consulting firm Deloitte.

This account currently has only 90,000 followers, making it a small account compared to others who won, and it has no media backing or verified blue check outside of the platform.

His title doesn’t rely on trending keywords, but the topic is quite provocative—exposing how Deloitte secured $74 billion in contracts from federal and state governments and then botched the projects.

Link here

Clicking in, you’ll see this person put in real effort.

He built a website called somaliscan.com, scraping millions of government invoice data points, cross-referencing audit reports and system failure records.

Using this firsthand data, he told a series of shocking stories: California’s unemployment system was defrauded of $32 billion, Tennessee’s Medicaid system collapsed, leaving 250,000 children without coverage, and court system upgrades cost $1.9 billion and remain unfinished… covering 25 states in total.

He also uncovered the revolving door between Deloitte executives and government officials, detailing who moved from Deloitte to which department, which contracts they approved, with names and amounts clearly listed.

One individual built a database himself and earned $1 million through independent research.

🥈 Second place: a finance influencer with 700,000 followers teaching how to profit amid tariff fears

Second place @KobeissiLetter is a familiar face in the macroeconomic finance circle, with 700,000 followers, focusing on US economic policies and market fluctuations.

His article is straightforward, breaking down Trump’s tariff strategies into a repeatable trading framework, titled “Trump’s Tariff Playbook: An Operational Guide.”

Since Trump often acts unpredictably, issuing bold policies and threats to other countries, but not always following through, Wall Street summarized this pattern as TACO—short for Trump Always Chickens Out.

TACO describes a recurring pattern:

Trump announces heavy tariffs → markets plummet → a few days later, he eases or delays → markets rebound.

Link here

KobeissiLetter’s article turns TACO from a joke into a timed operational manual. Using tariff events from the past 12 months as samples, he distills a complete cycle template for trading based on timing.

For example, weekend rumors from the White House cause panic, midweek buy-in from traders, the next weekend signals easing, and within 2 to 4 weeks, some agreement is reached. He also updates each step with ongoing commentary, making it feel like a continuous pre-research thread.

He offers practical advice: monitor the US 10-year Treasury yield. If it breaks above 4.60%, Trump is likely to concede.

For paid users interested in macro and trading, this kind of content hits the spot.

It doesn’t debate whether tariffs are good or bad, nor moral judgments; it simply tells you what actions to take at what times to profit.

🥉 Third place: Most liked DAN KOE, familiar methodology for life improvement

Dan Koe’s entry, “How to enter a state of extreme focus at any time,” received 42,000 likes and 8,681 shares, the highest engagement among all entries. However, its exposure was only 11.04 million, less than a quarter of the champion’s.

X’s official “Creator Choice” award, worth $250,000, was given separately, so technically he’s not third place.

It’s understandable—Dan Koe is “the person who inspired this contest.” His early January viral article with 170 million impressions directly showed X how high the traffic ceiling for long-form content can be.

Link here

The article itself is a typical personal growth methodology piece. It discusses how to gain focus, citing neuroscience and flow state concepts for support and elaboration.

While this article had the best engagement data, based on the core rule of “exposure among US paid users,” it didn’t rank at the top.

Why does the most liked article have relatively low exposure? We’ll discuss this discrepancy later.

Honorable Mentions: 4 × $100,000

Nick Shirley, Josh Wolfe, Kaizen Asiedu, Ryan Hall each received $100,000 in incentives. Their accounts cover public policy, geopolitics, history, and public safety.

Among them, Josh Wolfe is a co-founder of Lux Capital, a well-known venture capital firm, and announced he would donate his prize money equally to four charities.

Since the original posts didn’t specify their articles, and due to time and resource constraints, we haven’t conducted further investigation. Contributions and additional info are welcome.

Deeper Observations

From the contest results, some patterns emerge:

Most liked articles only have a quarter of the exposure of the champion

The most counterintuitive data point is Dan Koe’s.

42,000 likes, 8,681 shares, 4,627 comments—top engagement metrics across the board. But exposure was only 11.04 million, less than a quarter of @beaverd’s. Interestingly, @beaverd’s likes are only 30,000, fewer than Dan Koe’s.

If you’ve managed social media, this data feels odd. Usually, higher engagement means more algorithmic promotion and greater exposure.

But X’s contest measured “exposure among US paid users’ homepage timeline,” excluding non-US users, non-paying users, search, and profile visits.

Dan Koe’s content is about personal growth, which naturally appeals globally, including many non-US followers. @beaverd’s content is about how Deloitte wastes US taxpayers’ money, so his audience is primarily US-based. Under the same algorithm, the “geographic concentration” of content determines this exposure metric.

90,000 followers beat 900,000 followers—content scarcity > follower count

Champion @beaverd had 90,000 followers before the contest. Second place @KobeissiLetter had 700,000. Dan Koe has 900,000.

If follower count dictated exposure, the ranking should be reversed. But the actual results show that, in X’s article recommendation logic, follower size isn’t the main factor.

@beaverd’s victory hinges on having unique content—scarcity of content is a key factor.

This is very different from traditional traffic logic. Big accounts rely on follower volume and posting frequency, but in an algorithm-driven environment, “having exclusive content” outweighs “having many followers.”

Build your own content “hardware”

Looking at the three winning articles, their topics are completely different: one exposes government contracts, another teaches trading during tariff fears, and the third discusses how to focus attention.

In any content platform’s classification system, they wouldn’t appear on the same list. But they share a common trait: each has its own “hardware,” meaning a narrative framework.

@beaverd’s hardware is a self-built database scraping government data; KobeissiLetter’s is a tested trading framework over 12 months; Dan Koe’s is a six-chapter methodology blending neuroscience and psychology—though it may seem complex, it’s based on well-known principles.

None of the winners are purely opinion pieces. All require long-form content to carry substantial information, which is precisely why X Articles exists.

Another notable point: none of the eight winners are from traditional media.

They are all independent creators. It’s not that traditional media didn’t participate; rather, in this contest format, individual accounts have an advantage.

Institutional media usually publish content on their own websites, sharing only links and summaries on social media. But Articles require full content to be posted directly on X, which is awkward for media accustomed to external traffic.

What is X paying $2.15 million for?

Back to the platform itself.

X initially promised $1 million in incentives but ended up paying $2.15 million. During the contest, it also launched supporting measures: expanding Articles from creator accounts to all paid users, adjusting algorithms to favor long-form content, and changing scoring to “exposure on the US paid users’ homepage.”

Such a large investment clearly indicates that X needs original long-form content on the platform.

Previously, long content on X mainly came from external links—Substack, Medium, personal blogs. Users click away immediately, leaving reading time and engagement data elsewhere. The goal of Articles is to keep these contents on-site, so users can read from start to finish without leaving X.

On a deeper level, X has Grok. Training large language models requires high-quality long-text data, but most content on X is just 280-character tweets. If Articles can continuously attract creators to produce in-depth long articles, these become training data for Grok.

Finally, the value for paid users.

The contest rules limited metrics to “exposure on the US paid users’ homepage,” directly telling creators that their content should serve paying users.

This supports the paid ecosystem with creator content, making paid users feel “my money is worth it because I see in-depth content on the homepage that I can’t find elsewhere.”

From a creator’s perspective, we believe the era of pure opinion content may be coming to an end.

This trend also applies to crypto creators. The crypto industry isn’t short of opinions—every day, countless people call trades, predict prices, comment on regulations on X.

But few can build on-chain data analysis tools like @beaverd or break market cycles into repeatable trading scripts like KobeissiLetter.

Maintaining scarcity and independence, producing continuously—this is a highly professional activity, rewarding and fulfilling.

We also hope to see more content from the Chinese community, and that it will appear on future leaderboards.