Dmitry Rudenko made an AI-generated Reel about a chef in Sicily. He thought it was charming. Almost nobody watched it.
The day after, he made a Reel of a couple sitting on a couch, borrowed loosely from another creator’s idea. He didn’t think much of it going in. It became the video that changed how he makes everything else.
That gap, between what Dmitry thought would work and what actually did, is the real subject of the case he brought to ContentGrip. Dmitry runs Rubicone, a digital agency in Siberia, and treats his personal Instagram account like a lab. He posts constantly, watches what survives, and throws out the rest.
By day 18 of the account’s life, one of his Insights screens showed 165 million views and 98% of them from people who didn’t already follow him. That’s the headline. The more useful part is everything he had to unlearn to get there.
Table of contents
- Repeat the reason, not the video
- When taste loses to distribution
- Controlled variation beats creator boredom
- The workflow: keep, kill or multiply
- Chess and the habit of losing
Repeat the reason, not the video
The couch Reel worked because you understood it in under a second. Two people, a relationship, an emotional shift. No dialogue needed, no setup required.
Dmitry’s first move after it took off was the wrong one. He wanted to invent something completely different, because repeating himself felt lazy. The data disagreed. Variation after variation of the same underlying idea kept finding new viewers, on its own, without him doing anything clever.
So he started separating two things he’d been treating as one. The surface of a video, characters, setting, staging, could change completely. What had to stay the same was smaller: one relationship a viewer could read instantly, one clear feeling, almost no cognitive load.
“I stopped trying to repeat the video and started trying to repeat the reason the video worked,” he said.
Looking back at what actually traveled, he sees a pattern he now designs around. In materials he later shared about the account, he put it simply: the strongest films weren’t technical showcases. They were emotionally legible stories about parents, childhood, relationships, aging, memory.
When taste loses to distribution
The chef in Sicily wasn’t the only miss. Dmitry made a tribute to Sergei Yesenin’s poetry that he genuinely liked. It’s publicly visible now at 13 likes and a single comment, about as flat as engagement gets.
He also made a character-format Reel, a baby reaching for chili peppers, that he expected to clear 10 million views. Public engagement on that one sits at just over 3,100 likes and 22 comments; by his own account, it landed around 185,000 views.
That forced a split he hadn’t wanted to make: what he personally thinks is good work, versus what makes a stranger stop scrolling and send it to someone else. Those turned out to be different questions with different answers.
“I just no longer trust my taste to predict the distribution,” he said.
He still relies on creative judgment to make anything at all. He just no longer lets it decide what gets a second version. That job now belongs to reach beyond his follower base, share and send behavior, retention, and whether new audiences keep discovering something on their own.
Controlled variation beats creator boredom
One sequence makes the point cleanly. A Reel hit roughly 3.8 million views. A later variation of the same idea hit 13 million. A third version hit around 26 million.
Those specific view counts are Dmitry’s own reporting. The pattern he describes runs against a common assumption: that an audience gets tired of an idea once they’ve seen it.
Usually, he says, it’s the creator who’s tired, not the audience. He might watch a concept 50 times while generating images, fixing a character, cutting the edit. A stranger scrolling past sees it once, for the first time, with none of that fatigue attached.
“I think of it as controlled variation,” he said. Keep whatever’s actually doing the emotional work. Change everything else. Stop only when the data says the idea itself is spent, not when you personally get bored of looking at it.
The workflow: keep, kill or multiply
There’s no content calendar behind any of this. Some days Dmitry posts five Reels. Some days ten. Some days none.
On the day he wrote his answers, he’d made five versions of the same premise that morning. None had crossed a million views yet, but one was already pulling ahead of the other four.
He describes the loop as: idea, variations, publish, observe, then keep, kill, or multiply.
Multiply is the part that took him longest to learn. His old instinct, once something worked, was to congratulate himself and move to something new. Now the question is different: what exactly worked, and what happens if he changes everything except that one thing.
He’s tested posting times too, mornings, afternoons, 2am, 5am, and found none of it mattered nearly as much as the content itself. Some nights show stronger accumulation than others. He’s not willing to claim there’s a magic hour hiding in the data.
Chess and the habit of losing
The habit underneath all of this predates the Reels account by years. Dmitry plays rapid chess on Chess.com most days, two to four games, and jokes that his rating has stayed modest the entire time.
What chess actually taught him wasn’t strategy. It was what happens after a move that looked brilliant and turns out to be terrible ten seconds later. The game ends. You start another one.
“That ability to lose repeatedly without interpreting every loss as a reason to stop has probably been surprisingly useful when dealing with content algorithms too,” he said.
The 165-million-view number isn’t proof of a formula anyone can copy. What Dmitry’s account actually documents is a discipline: treat your own taste as a guess, let the audience tell you when the guess is wrong, and don’t walk away from something that’s still working just because you’re the one who’s seen it too many times.
