The New Way Apps Dominate in 2026 — Transcript

Tim Gabe explores how apps in 2026 dominate by building personalized intelligence traps that lock users in beyond engagement.

Key Takeaways

Engagement alone is insufficient; apps must build personalized intelligence that compounds with use.
Personalized AI models create a form of lock-in that users cannot easily export or replicate elsewhere.
Designers should identify and optimize investment loops that improve the product specifically for each user.
Visible personalization and measurable performance improvements increase user retention and perceived value.
The intelligence trap is a powerful growth strategy that outperforms traditional engagement tactics.

Summary

Apps in 2026 win not just by engagement but by turning user data into personalized intelligence that cannot be exported.
Tim Gabe, a product designer with experience at Spotify and startups, explains this new design pattern called the intelligence trap.
Layer one is engagement with dopamine loops and habit-forming interactions; layer two is stored value evolving into personalized intelligence.
Midjourney uses AI to build a creative personalization profile that learns users' visual tastes, creating deep user investment.
Oura Ring builds intelligence lock-in around users' physiological data, offering personalized health advice that can't be transferred.
RAMP applies intelligence traps in corporate spend management by training AI agents on company transactions to increase retention.
Key strategies include mapping product investment loops, designing cumulative personalization, and measuring product improvement over time.
The intelligence trap makes switching apps feel like starting over with a stranger, increasing user retention and exit friction.
Regulations like the EU Data Act allow data export but cannot free the personalized AI models built on user data.
Tim offers free design strategy calls via ZipSap to help founders apply these principles to their own products.

Chapters

Full Transcript — Download SRT & Markdown

Speaker A

Every founder I talk to is obsessing over the same thing: engagement, dopamine loops, addictive micro interactions, hooking animations, and they should be. Those things matter a lot. But here's what's keeping me up at night. The apps dominating 2026 aren't

Speaker A

winning just because they're more engaging. They're winning because they've turned your usage into an intelligence that you literally cannot export. Your sleep patterns, your creative tastes, your company's financial logic, living inside products that get smarter about you every single

Speaker A

day. And no regulation, no competitor can give you that back. Now, before I show you how they're doing it, I'm Tim and I've worked as a product designer for over a decade, designing software for tech giants like Spotify and

Speaker A

countless multi-million dollar startups along the way. So, I obsess over these things all the freaking time. Now, in 2026, the best products don't just solve problems. They learn you. Over time, they build an understanding so deep that switching feels less like changing tools

Speaker A

and more like starting over with a stranger. That's a design pattern. And today, I'm breaking it down using three companies that have mastered it. Lots of founders are building engaging products now. Smooth onboarding, satisfying interactions, habit forming loops.

Speaker A

That's what I call layer one. And you absolutely need it in today's AI sloppied world. But engagement alone is like a treadmill. It keeps users running, but it doesn't stop them from jumping off and onto a new one. The apps

Speaker A

people genuinely can't quit have figured out a second layer where every session deposits something irreplaceable, something that compounds, something you can't take with you. In 2014, Nir Eyal introduced stored value in his book Hooked. Users deposit value into

Speaker A

products and that investment makes them stick around. Powerful idea, but in 2026, AI has supercharged it into something AI couldn't have predicted. I call this evolution the intelligence trap. Every session doesn't just store data anymore. It trains an intelligence,

Speaker A

one that can't be transferred, exported, or rebuilt overnight. This is more powerful than any growth hack you can implement. And by the end of this video, you'll understand exactly why. Before we dig into the three companies that have

Speaker A

mastered this, if you want help applying principles like this to your own products or website, we open up free design strategy calls monthly at ZipSap.

Speaker A

You can grab your spot in the link below. All right, let's start with the example that surprised me the most personally. Midjourney. $500 million in revenue in 2025, zero in VC funding, no traditional marketing budget to speak of, roughly 150 employees. That's over

Speaker A

$3 million in revenue per employee. Midjourney started as a Discord bot. No website, no native app, no polished onboarding flow. It grew entirely through word of mouth. And even today, competitors like Adobe Firefly or DALL·E have billion-dollar backing, massive

Speaker A

distribution, and years of brand recognition. By every conventional measure, Midjourney should not even be competing with these guys. So why is it?

Speaker A

Because after you go through their image pair ranking, Midjourney builds what they call a personalization profile. We're not talking about a settings page here.

Speaker A

We're talking about a trained submodel of your visual taste. It learns whether you prefer warm or cool tones, maximalist or minimal composition, photorealism or painterly style. In version 7, personalization is on by default, and over 70% of active users adopted it

Speaker A

within three months. Every image you generate, every variation you select, every mood board you curate trains this submodel further. Switch to DALL·E and you're starting with a stranger's eyes.

Speaker A

No profile, no mood boards, no creative history. Your creative fingerprint lives on Midjourney's servers. And the more you use it, the more it understands what beautiful means to you. Specifically, Norton, Moan, and Arieli proved this with their IKEA effect study. People assign

Speaker A

63% more value to things they helped create. Now, here's how to apply this to your product. First, map your product's investment loop. For every core action a user takes, ask, does this make the product better for them specifically, or

Speaker A

just better in general? If it's just general, you're building features, not stored value. Second, design for cumulative personalization. Midjourney's aesthetic profile gets better with every ranking. So, what's your product's equivalent of their ranking system? Find that compounding action and make it

Speaker A

frictionless. Third, measure the compounding. Track how much better your product performs for a six-month user versus a 10-day user. If there is no measurable difference, the intelligence trap isn't tightening. Can your product show that a long-term user gets

Speaker A

measurably better results? That's the question you want to ask. So, Midjourney built an intelligence around your creative eye. Now, the next example might seem different because it's a physical product, not a software app.

Speaker A

But stick with me because the way they built intelligence lock-in around your body contains maybe the most transferable lesson in this entire video. I'm talking about Oura Ring.

Speaker A

Millions of rings sold on track for a billion dollars in revenue. And the US Department of Defense just gave them a $97 million contract. That's a health wearable, not a defense contractor.

Speaker A

After three years of wearing one, Oura has 1,095 nights of your sleep architecture data, heart rate variability trends, respiratory rate patterns, temperature baselines, and their AI advisor uses all of it to give you personalized health recommendations, not generic sleep more advice, advice

Speaker A

calibrated to your specific physiology. The problem isn't whether Oura is a good product. It clearly is. The problem is what happens when you try to leave. The EU Data Act mandates that Oura lets you export your data. So technically you can

Speaker A

leave, download your sleep CSVs, but that CSV doesn't contain the AI model that learned your body's rhythms over three years. You get the numbers, you don't get the intelligence. The regulation freed the container, but it couldn't free the intelligence. We've gone from

Speaker A

data lock-in to intelligence lock-in where a model trained on you lives inside the product. Oura maintains retention rates in the high 80s even behind a $5.99 monthly subscription on data your own body generated. So ask yourself, does my product get measurably smarter about

Speaker A

this specific user over time? If session 1000 is functionally identical to session 10, you've built a great engagement product, but you haven't added the investment layer that makes the exit door heavier with every session. So do this. First, identify

Speaker A

what data your product is collecting that could compound into personalized intelligence. Not just storing it, but learning from it. Second, make the intelligence visible. Oura shows your body has been tracked for 1,095 nights.

Speaker A

When users can see their investment growing, retention feels like partnership, not like being trapped. And third, ask yourself the exit question.

Speaker A

If a user left today and went to a competitor, what would they have to rebuild from scratch? If the answer is nothing much, you don't have an intelligence layer yet. Now, Midjourney built an intelligence around your creative eye. Oura built one around your

Speaker A

body. And if you're looking at your own product wondering how to build this kind of investment layer, at ZipSap, we run a select few free strategy calls with companies every month. Links down below if you want to grab a spot before they

Speaker A

fill up. So, we've covered creative intelligence and body intelligence, but the final example shows this principle working in a place you might not expect, corporate spending. RAMP is a spend management platform used by over 50,000 companies, including Shopify, Figma, and

Speaker A

Notion. Their AI agents made millions of autonomous decisions on company spend in a single month. Every automation rule your team confi

Speaker A

years to accumulate. This might sound like a B2B enterprise story that doesn't really apply to you, but the principle is universal. Every product accumulates decision intelligence. For RAMP, it's spending rules. For a project management tool, it's workflow patterns. For a CRM,

Speaker A

it's sales playbooks. The question is whether your product is actively learning from those decisions or just storing them. Ram's biggest competitor, Brex, was acquired by Capital 1 in January 2026 for less than half its peak valuation. Both companies had great

Speaker A

product design and strong engagement, but RAMP layered investment loops on top intelligence that compounds with every transaction. Great designs got users in the door for both, but only RAMP made the exit door heavier. Here's how to apply it in practice. First, identify

Speaker A

the decisions your users make repeatedly and ask whether your product is a learning from them to make the next one easier. If every decision starts from zero, you're leaving stored value on the table. Second, count your integration roots. Every external system that

Speaker A

depends on your product is another reason switching is painful. Ramps itself into accounting, HR, Slack, everything. What systems does your product plug into? Third, automate the deposit. The reason RAM's intelligence trap is so powerful is that users don't

Speaker A

have to consciously invest. Every transaction automatically trains the AI. Find ways to make your users deposit value without even thinking about it.

Speaker A

Now, let's recap. Three companies, three completely different categories. One, identical mechanism. The first lesson, build personalized intelligence, not just features. Mid Journey didn't outdesign Adobe. It built an aesthetic profile that gets smarter with every image. The second lesson, engagement

Speaker A

gets users in the door, but investment keeps them there long term. Aura proves that users stay when the intelligence feels irreplaceable. And the third lesson, the question isn't whether users can export their data. It's whether they can export the intelligence your product

Speaker A

built from it. That gap is where the intelligence trap lives. Now, if you found this breakdown helpful and you're wondering how this might apply to your product or your business, for the last time, we do open up a couple free design

Speaker A

strategy calls each month. Just check the link down below. Also, if you like this video, you'll probably love this one where I break down the emotional design layer, aka the first half of this equation. It's here somewhere. Anyways,

Speaker A

until the next one, have a great

Topics:product designuser engagementintelligence trappersonalizationAI personalizationuser retentionMidjourneyOura RingRAMPstartup growth

Frequently Asked Questions

What is the intelligence trap in app design?

The intelligence trap is a design pattern where apps build personalized AI models based on user data that cannot be exported or replicated, creating deep user retention beyond engagement.

How does Midjourney use AI to retain users?

Midjourney builds a personalization profile that learns each user's visual tastes through their interactions, making the creative experience unique and hard to switch from.

Can users export their data and AI models from products like Oura Ring?

While users can export raw data due to regulations like the EU Data Act, the personalized AI models trained on that data remain locked within the product, preventing full transfer of intelligence.

Get More with the Söz AI App

Transcribe recordings, audio files, and YouTube videos — with AI summaries, speaker detection, and unlimited transcriptions.

App Store Google Play

Or transcribe another YouTube video here →