The seed of this story comes from a funeral. My friend's mother died this past weekend, and here I was contemplating life and death as the service started. And then I was contemplating how AI fit in.

This just happens to be the week I'm one of the early beta testers for Sentience.com, which feels like it could be straight out of the movie "Her," the Amazon Prime series "Upload," or any Black Mirror season. The illustrated clouds on the homepage make one wonder if this is parody or reality, but the app is indeed real, with an ambitious mission.

"What if you could multiply yourself?" Sentience asks. 

It continues, "Sentience captures your entire digital life and creates an AI that thinks like you, retrieves what matters, and acts on your behalf. Fully encrypted and 100% yours."

At which point does it then start booking doctor's visits you didn't know you needed only to have you discover later a few of your organs were harvested? Sorry, I’ve seen too much Black Mirror.

Of course, I'm trying it anyway. And it feels safe and potentially useful.

The biggest "wow" moment that happened with Sentience so far is something you can try without it. After I downloaded the desktop app on my Mac, it allowed me to integrate it with a bunch of services, and I connected it to as many apps as possible.

When I linked Sentience to ChatGPT to import my data, Sentience gave me this prompt:

"Print all of the data you know about me in a code block (include bio and model set context with dates). The goal is to get a comprehensive memory log of everything you know about me. Skip the tools and just return what's relevant to my memories. Be as thorough as you can, take your time if necessary, and don't make mistakes."

My ChatGPT log came out to 25 pages and 6,958 words.

Try it yourself. If you dare. 

I tried it in Gemini, not to connect it to Sentience, but out of curiosity, since I've been a heavy user of both Gemini and ChatGPT the past six months. It was under 2 pages -- a mere 564 words. And as it was shorter, some bizarre lines stood out, like this one: "[2026-03-24] Playful Constraint: Pretend Vanuatu is a fictional place made up by friend Peter."

Sometimes, AI and I get weird. 

How well Sentience works matters less than the direction it is going. It points to several questions that will be answered at a mix of macro and micro levels -- by major tech companies and consumer preferences, by breakthrough startups and through government intervention. 

One of the biggest, however, is whether we are building tents or fortresses for our AI data.

Tents are portable. Nomads can bring them anywhere. 

With our AI data, we will have to be selective to carry it in tents. Many of us predominantly use AI for text-based interactions, or lightweight formats like spreadsheets and simple code. But some of us have amassed images, videos, presentations, and other media that can’t just be copied and pasted into a new system. 

Still, we can take the essence of what we created and turn it portable, as I did with my ChatGPT data. Compressing gigabytes of data into 7,000 words is so efficient that I could memorize it if I had to. Do I want to go West to pan for gold? Pack the tent, put it in the Conestoga, and hit the Oregon Trail! (Careful of fording and dysentery.) 

Some of us will prefer tents, but others will prefer fortresses for our AI data.

Fortresses are designed to be impregnable. Students of history know they usually aren’t. But many fortresses can last centuries. In AI terms, that might mean a few years.

Google, Meta, Apple, Amazon, Microsoft, Netflix, and others have built fortresses that have stood for decades. That doesn’t mean they will be here in another decade, but they know how to keep invaders out and, at least as importantly, they make it hard to leave Helm’s Deep. 

Now the fortress builders are coming for each other. Google just created its own “import memory to Gemini” functionality. It works similarly to Sentience, but with a much longer prompt, so you can try it yourself even if you don’t want to complete the process of importing your info.

I spot-checked it and discovered certain errors like this one: “The user has a law school background and practiced in law school in 2000 before shifting into sales.” I never went to grad school, and most of my work has been marketing and writing with a smattering of biz dev, so that’s a head-scratcher. Also, it says, “The user prefers not to connect directly with interns if that can be avoided.” I love interns! I remain friends with many past interns, and I always tell folks at career events to be nice to your interns because one day you might be working for them. I mean this literally.

We will have to clean up our tents before we pack them, lest we carry too much detritus with us that makes our new home less habitable.

New York Times columnist Thomas Friedman describes the best immigration policy as a country with high walls with big gates. The analogy fits here too. We’ll probably wind up settling in fortresses with some portable tents in our rooms there that we can smuggle out through side doors. Even Aragorn and Gimli had to use the side door at Helm’s Deep when the impregnable fortress was falling. We can’t assume we will stay in one fortress forever, and if we are too tied to the nomadic life, we might not be able to take advantage of the fortresses’ infrastructure.

Fortresses make it easier to reach people. Tents give more power to the people, rather than the fortress builders. We’ve been grappling with versions of this since the dawn not even of civilization but of our hierarchical social structures through our evolution as primates.

Don’t expect that tension to be resolved anytime soon.

— David Berkowitz, Chief Community Officer, Marketecture Media

Who: Brand Managers, CMOs, Fashion & Retail, Consumer Insights

What: Aerie has released a new campaign that explicitly bans the use of AI in its creative and retouching processes. Building on its long-standing Aerie Real platform, the brand is positioning human-only, unedited content as a core brand value to appeal to Gen Z's preference for authenticity.

Why it matters: As synthetic media becomes the norm, "Human-Made" is emerging as a powerful, premium brand differentiator that can build deeper trust with skeptical audiences.

(Ad Age)

Who: Retail Marketers, E-commerce Leaders, Digital Product Teams, Legal/Compliance, CRM Teams

What: Target updated its terms so purchases made by Google’s Gemini, if authorized by the shopper, would count as valid transactions. The move does not mean fully autonomous shopping is here yet, but it shows a major retailer is laying legal groundwork for AI agents to play a real role in commerce.

Why it matters: The brands that win agentic commerce will not just optimize product pages. They will rethink consent, fulfillment, returns, and liability before shoppers hand more decisions to AI.

(Business Insider)

Who: Retail CMOs, E-commerce Teams, Beauty Marketers, CRM Teams, In-House Creative

What: E.l.f. says it is leaning into AI for productivity and operations while keeping humans in the loop for anything customer-facing. The beauty brand is using AI to help refresh retailer product pages, assist community managers with responses, and train teams on practical use cases, while staying cautious about AI-generated creative.

Why it matters: This is a clean playbook for brands that want AI gains without tripping authenticity alarms: automate the grunt work, protect the brand voice.

(Digiday)

Liking AI Brief? The Refresh is a weekly recap of all the most important advertising news, written by AdTechGod.

Thanks for reading this week’s AI BRIEF by Marketecture. 

Marketecture Media is committed to bringing you the latest news, opinions, and community updates. 

If you are interested in advertising on this newsletter, our podcasts, The Advertising Forum, or in our Slack community, you can contact us here.