35 minutes: the industrial dose of captivity — and why deleting an app today is more than an act of freedom

35 minutes. That is the precise dose beyond which TikTok's own internal documents —exposed in 2024 by faulty redaction in a lawsuit filed in Kentucky— estimate that an average user crosses the threshold of compulsive use: 35 minutes, some 260 videos.

The same internal research linked that threshold to measurable losses in analytical thinking, memory formation, contextual reasoning, conversational depth and empathy. This is not moral panic, it is an industrial specification. Captivity, calibrated.

For much of the digital era we framed addiction as a personal failing: a matter of willpower against a supposedly neutral platform. The leaked documents settled the debate. The dose is designed. The threshold is known and the harms are quantified. The public screen-time control features, as some employees acknowledged internally, were meant more to improve public trust than to reduce actual use.

In other words, the platforms that have designed access to reality for many of us are industrially designed to manage users' attention funnel. But is that the end of the story? Now let's look at the top of that funnel.

FROM MANAGING ATTENTION TO BUILDING THE NARRATIVE

The current president of the United States posts several times a day on Truth Social —a platform he owns— and every post turns into a headline somewhere before the following morning. Sometimes even into a source of enrichment for his close circles (suspicions of insider trading).

The asymmetry is precisely the mechanism. He does not need to convince an editor at a media outlet. He needs to set a topic, to build the narrative. Reality is no longer reported. It is programmed.

The comparison is uncomfortable, but precise. TikTok designs the dose beyond which a teenager will keep scrolling. Truth Social designs the dose beyond which a media cycle will keep covering a pre-designed narrative.

Both operate with the same physics: low cost of emission, frictionless distribution, algorithmic amplification and an audience whose analytical reserves have already been eroded by the previous post, the previous video, the previous scroll. Truth and lies last only until the next narrative begins.

Truth and lies last only until the next narrative begins.

Even if it is possible to perceive the manipulation, we must understand that we are part of a hamster wheel designed to manage our attention and steer our attachments, affinities and aversions towards a pre-established narrative. The solution lies in educating and cultivating our critical thinking. Something that, these days, we can even do with the support of AI. We look for the antidote through deeper reasoning, the cross-checking of sources and the verification of facts.

To achieve it, we turn to Artificial Intelligence, to the solution we have on our phone. More than one in three internet users are already intensive AI users. We turn to the corresponding LLM. We ask it to summarize, weigh and refute the facts. We treat it as a cognitive prosthesis (digital companion) that will save us from the feeds. And we think we have neutralized the narrative. But the LLM has its own bias.

In March 2026, Angelo Romasanta, Llewellyn Thomas and Natalia Levina published a study in Harvard Business Review based on more than 15,000 strategic advice simulations run on the main frontier models. They found a systematic preference for whichever side of the dilemma sounded more “on trend” within the training corpus, regardless of the context. They called it Trendslop.

Even rich, industry-specific, context-loaded prompts only reduced the bias by around 11%. This matters for the framework I proposed in "Synthetic Feedback Loop: La nueva alcoba de IA o el nuevo filtro burbuja de la IA" —published on LinkedIn on April 7, 2026—. Trendslop is a new bias we must incorporate into the loop. And it may be the most relevant one of all, because it operates precisely when we believe we are escaping the other biases.

We know to be suspicious of the filter bubble. Trendslop disguises itself as analysis.

SOLVING THE EQUATION: AN ACT OF FREEDOM

Let's stack the three elements:

• TikTok's designed dose.

• Truth Social's pre-cooked headline.

• The LLM's trend-aligned, biased answer.

The diet is the same, only the cutlery changes. And that leads us, paradoxically, to the simplest act available:

Delete the app. #deleteSocialMedia

This is not a theoretical debate or a matter of digital hygiene. It is a technical urgency: deleting is the only action in the contemporary context that does not require a counter-platform (BlueSky vs Twitter/X) or a counter-LLM in order to exist (Multi-LLM Orchestration, or, Ensemble Prompting). Don't try to improve the data; just remove the surface where your information is turned into a weapon. Stop outsourcing your will through that algorithmic friction. Self-moderation is a fantasy of control against a system designed to subdue us. If the only real independence is refusal, then analysis may not be enough. The only honest answer is total disconnection. Delete the app.

If the only real independence is refusal, then analysis may not be enough. The only honest answer is total disconnection. Delete the app.

What I called "regurgitation bias or AI mad-cow-disease bias" in Synthetic Feedback Loop will go on operating without our individual participation, and Trendslop will go on shaping the strategic memos of organizations that perhaps never opened TikTok and into whose future narratives their future strategies have already been inoculated via the data corpus of the AI models. It is an unintended self-fulfilling prophecy. Trends will be overweighted because the models are trained on the pre-designed narratives that turn out to be viral.

Is everything lost, then?

I don't think so. Personal cognitive sovereignty has to start somewhere. And that somewhere is surprisingly and uncomfortably small: an icon held down, an option confirmed, a few days of abstinence, a capacity for boredom that slowly comes back. The precondition of all original thought that we used to take for granted. We have stepped out of the trap of the pre-built narrative.

35 minutes is the dose of captivity. And 3 seconds is the dose of freedom.

That is how long it takes to delete the app.

The asymmetry, for once, is on our side.

Note: [This article is tagged as "Human led". Human-led content. Machines conduct checks, highlight and correct errors, enhance output. Scheme based on the Dubai Future Foundation. "Human-Machine Collaboration (HMC) Icons." Dubai Future Foundation, dubaifuture.ae/hmc]

Bernardo Crespo is a seasoned digital transformation and data strategy leader with over 25 years of experience. He has held leadership positions in Fortune 500 companies and digital consultancies, and has founded and advised numerous startups and venture builders. Currently, as CEO of his own firm, Quantum Markethink, he provides strategic guidance to C-suite executives, helping them navigate the complexities of digital transformation.

Furthermore, he serves as an Academic Director at IE Executive Education, where he brings his expertise in emerging technologies, digital strategy, data strategy, and artificial intelligence to the classroom. Prior to these roles, he spearheaded digital transformation initiatives at Merkle Spain and led digital marketing at BBVA, where he notably pioneered the application of gamification in banking.

Bernardo is also the co-author, with Gam Dias, of "The Data Mindset Playbook: A Book about Data for People Who Don't Feel Like Reading about Data" (KDP, March 2023).

AI-augmented articles: Newsletter "My AI-ter Digital Ego"

Human craft: Newsletter "Digital Sustainability"