The End of Extractive Apps: Designing for User Agency

Transcript

Hello. The awkward silence until I start just eating me up. Okay. Um Hi everyone. Are you all bored? It's the last day of the conference. Okay. I need to warm up a little. I was running. I couldn't find the stage. Okay. Um One more second. >> [snorts] >> Let's try this. Okay, back.

Um I feel like this is such a calm room. So, I'll start with something active where we are all like kind of talking to each other and not sleeping. I heard that the room is so dark, you're just kind of you're pretty much like ready to sleep. Okay. So, this is I am Alby. I'm Albiona. I go by Alby Codes on socials and currently I'm an entrepreneur in residence at Flashbots X.

And yeah, what what do I do? So, we build consumer products on TEEs and Flashbots X is basically Okay. This is like too many inputs at the same time. It's like having three for you pages in front of you and you're not sure where to look at. But it's good. It's stimulating me. It's fine. And that's my that's my main point of the talk, which is great. I'm experiencing it live.

Okay, so I'm going to talk today about the end of extractive apps. What does that mean and how do we design for giving back agency to users? And I was sharing that I'm part of Flashbots X, which is this accelerator arm of Flashbots and also contribute to the teleport team, in which maybe you kind of are know people like Andrew Miller and Shin. One of the first products that they built was teleport.best, in which you could kind of delegate your tweet your Twitter tweets and through an LLM restriction on what you could post and all of that was on a TEE or like each also the tweet was basically an NFT at the same time.

But yeah, today I'm going to talk about how do we design applications in this day and age of AI and socials that we can get back to we can give back users agency. And I have a question for all of you and this is very relatable and I need for all of you to raise your hands. The question is when was the last time you didn't take your phone to the toilet? Do you remember? Do you remember if you just didn't take your phone to the toilet?

Okay, I see four or five out of maybe like 12 people, which is nice because you know, we are in the conference. Uh you're rushing around and like what you're walking around. But yeah, um if you can remember that, that's not a failure of discipline. That's basically someone designed that behavior in your day-to-day life and you never opted in. And um what did that make me feel? Kind of you know, when you go through those doom scrolling sessions and you go out of it and you feel foggy and everything, but you also don't really remember anything what you're going through. It's basically like I'm going to talk about these apps, but

basically I'm talking now from the power user's perspective. Like I do consume a lot of socials on my phone and that made me do something about it. So, I took all of my Whoop data. So, I have a Whoop on my arm, which tracks everything and what I did, I was like Okay, because during the night, whenever I was like having late night sessions on my screens and then my because of being so tired, I would replace my rest time to screen time on TikTok or on YouTube or on Reels on Instagram and that kind of got to create a loop in in itself. So, I got all of my Whoop data for like 6 months, downloaded that and I then went to my

TikTok watch history. I said, let's download that as well. And at the same time, I have this extension that I use on browser, which is called the nutrition facts and it basically labels all of the tweets that I consume on browser by a nutrition facts label. So, it's like, oh, you consumed a positive tweet, a negative, aggressive, snark and all of that. And then I also connected each video of my watch history to like an emotional content label to see, okay, what is that information diet that I'm going through that I'm feeling in this way or like it makes my body just like feel different.

It's it doesn't make sense to kind of be in a state of like needing to touch grass every here and there because that the grass is just so far away even though we are walking the same streets and we are in nature most of the time maybe outside of our laptops, right? So, what did that tell me is that um Yeah, this is how the nutrition facts look like. So, basically when at the time when I did this calculation in the back that I got 22,000 tags post on X and then 17,000 TikTok videos that I consumed.

Um and I also measured So, basically I measured those two data sources with my heart rate variability on Whoop because I wanted to know how is my heart rate like acting up whenever I consume whatever kind of content on socials. And the nutrition facts label basically is done by, you know, like a community member who's very into information diets. It for God that I installed it just yesterday, I saw that it's now 29,000 tweets that I've consumed on browser, which is like that's how I consume on Twitter most of the time.

Um don't use it that much on my phone. And um you by the end of this talk, you might be more aware of how could your information diet look like on a day-to-day basis. That's the goal. And um Yeah, what were the results of of this calculation is basically that on the days when there was a lot of rage content, my body was basically recovering as if it would recover from a night of drinking alcohol tomorrow. My heart rate would like consistent like mis- was mismatched in a way like changed that drastically from the other days versus when I would consume something that was very like intentional

um and in different times of the day as well, that was very important. If my scrolling time sessions were during the day versus at night on TikTok, that really made a difference on my heart rate. Uh it basically um was more stable. Like I was more stable and kind of more like better improved on of course taking in all of the exercises and things like that. And that made me realize that, oh wow, I was not just me like consuming something content, but content was literally consuming me physically or even mentally. This is the mental load that is one of the most important parts.

And uh there is this feedback loop basically that when you have um so much intense consumption times period um it basically drops the recovery like score on your Whoop if you are a Whoop user in this case and um this architecture basically made it feel like the reason why we are doing all of this consumption like journey in a way or like usage on all of these socials is basically the architecture made it so low resistant or even no resistance at all for you to just flow into that state of consuming

things. And it got me thinking and basically it there was a point in time even when I would meet with my friends that we would say something like yeah, like I'm doom scrolling and the never ending the never ending feed is just eating me up and it was so interesting to notice that especially with with a lot of a lot of AI tools that are being used now to even build better things and to get closer to the end user's experience. And it felt like we are all delegating our thinking to AI and our attention to socials media that

is just happening automatically. Not because like we wanted that to happen, but because the architecture kind of gave it gave in to like our life experience and there's no no other path to go other than some other points that I'll talk in a bit. And then here's a question that I I would ask myself or even my my best friends is that if the algorithm because at some point in earlier I I felt very good when the algorithm would show things that I was very interested in.

Uh but the question is like if the algorithm can predict what you will watch next, what you will click next, and then what you will feel like, what side of you can it can't it predict? Like what is there a side of you that the algorithm doesn't really know? And that is like this never-ending cycle of like experience in all of these uh current apps that are a part of our day-to-day life. And uh this is where uh >> [laughter] >> the game shows up that we are stuck in.

In a way or another, the two best friends are caught. They are set in a pri- two different prison cells. They are given like options, as you know, like the prisoner's dilemma. Oh, if you betray your friend, you get to walk free, but if they betray first, you get 10 years. Or if you stay silent, 1 year. Or if you both betray, then you get 5 years each. And in this case um uh what happens, like where do we fit in this side of things, is that you we both both sides know that oh, it would be better for us to stay silent, but the betrayal happens because you never trust that the other party wouldn't betray you or like would betray

you. So, you just keep doing the same thing all over again. And then, um where do those social platforms fit in? Those are the actors that are basically in different cells. TikTok, Instagram, um all short-form media, YouTube are basically on a race to extract as much as possible from the end user. And in this case, if one of these platforms gives up on the engagement stuff, they would immediately know that the other party would just get most of them market share after they have given up. So, what do they do? They don't really give up.

And where would me and you fit in in all of this system? Uh this is where basically we are the commons. This is where this resource is being depleted again and again. Our our attention, our energy, are basically glued glue feeling to all of those screens. And the interesting point is that each person in this room uh carries something that no model can replicate. There is a specific combination of experience instincts knowledge relationships that make you you.

And you can call it a soul, you can call it an edge, whatever. It's a thing that makes you not an NPC by default. And those extra extractive platforms are systematically trying to like flattening that side of yours. Like, okay, turning your uniqueness into this like targeting profile, compressing your depth in a watch history list, and then making you predictable because predictable is profitable. And there are 15 billion hours being spent every single day on social media.

And the amounts of profits are just like never-ending, like 200 billion dollars on Face- like Meta alone in this case. And they just launched this model Tribe V2, which is basically trained on brain scans. Then now they can uh see, okay, what kind of piece of content does to a human brain without even trying to scan new brains or stuff like that. So, it's basically a model. And now all of the end the the end perspective is like, okay, now we do all of this algorithm, now we actually know how to precisely hit the points that we need to hit, which is so sad.

And also, the government is basically noticing as well uh by increasing the fines, increasing the bans, landmark verdicts across all you EU, US, Australia. But regulation basically doesn't change the actual game. The structure is still there. Uh it's just the penalties are growing, like especially this last 6 months at these are being very active. And what do we do in that case? Uh and why does that happen? Uh basically, we are all functioning in these bubbles that we think that we're doing our best for the best, but then the collective output is i- is bad. No one basically wanted to uh choose this outcome or like the

benefits that we have. Uh and [clears throat] but no one alone can actually stop the thing. Uh stop the effects that we are experiencing. And if you delete all of those, if you remove yourself completely, then it's like uh privacy as withdrawal. You're like, okay, I'm not going to participate, but you're also you're going to is- isolate so much. Um so, how do we do the change um of the structure itself so that we can make it work to for us and not against us?

>> [snorts] >> Um and most people from the end user's perspective is basically they don't recognize that this is a trap. Uh they think that scrolling is just a choice. And they think that the feed is basically for them and for their good, but the architecture was designed to basically make you feel bad to stop. It's like you can't feel good if you stop the never-ending scroll. And there is a little fact on the never-ending scroll that basically was developed by this developer in 2006 that the guy was like, "Hey, I'm tired of clicking the next button. I'm tired of it. Let me just remove it completely."

And that's when like it just happened naturally. No one like knew that that would like end up into this result of like huge platforms of uh being built. So, what do we need? We knew a new we need a new architecture, something that um you don't really negotiate the terms inside the actual game, uh but something that you can take that game and like make it fit into your day-to-day life. At Teleport, uh we basically treat agency um uh um well, uh Teleport, we see agency that as something that can be engineered and can be designed into applications for the user to basically be able to decide for

themselves. Um platforms basically for now, they give this low granularity uh granularity agency, which is like, oh, here's your screen time. Oh, you can also get a passcode on uh apps that you're using. And um you can decide to not use it, but that's like putting a padlock in a in a cookie jar. Uh that's not agency. Um it's like a a shallow perspective on that. And um the high granularity tools are basically the ones that basically get all bits of data for you and that feed that to the algorithms, feed that to the models, are basically being stalked in. Like you're you're not aware of all of that data about yourself.

Um and that creates like this false binary. That's where uh basically you can go out um uh you can delete, go off the grid, uh and get this defensive isolation. Or you can just passively accept all of it. You take the feed and then you be the actual product. Both basically deny access to your own agency. And how do we um move forward after all of this? Uh I call this agency experience. Um not what people want, but what do they want to become? Uh the tech industry started the AX as an identity experience

or agent for to kind of survey our agents and stuff, but I'm using it consistently for designing apps to serve humans. And one gives the power like to the system of all of the AI tools, but the other one kind of gives gives gives it back to the user. And this is not just a cover-up for ethical design rebrand. Ethical design kind of says to developers, "Hey, should you should be better at your work and do this and that." But like agency experience is basically says that the architecture should be uh impossible to be for for you to be extractable from users. Uh that should not happen.

And the three pillars of the agency experience are autonomy. Basically, it supports users' autonomy by the for them to decide what happens with their own attention. Uh to so, having that cross- crossroad where they can like click something then creates friction and also make the invisible visible. All of that whoop data that I shared, all of the TikTok watch history um that I also installed and everything. This is kind of the architecture of what happens. We call it designing information diet applications where um all of the information gets passed as you see feed for you page, watch history, but then you clock that in into this design that only um

uh that serves only inside uh TE. And then, I would just move so quicker to the um uh to the diagram, but remember that the operator literally is blind. And um uh when with this architecture we are basically applying the address serial interoperability where we don't really get to change inside of it, but we take all of that and we serve we give it back to kind of the user uh uh in a way or another in different um scenarios. And this is what we call the Hivemind Core, which is basically that like the query layer built on top of it.

The app store the app stores all of the data in a Postgres [clears throat] database. Watch histories, gold planners, and everything. But the constraint is that no one reads the database directly, not the developer, not the operator, or the agent. Everything goes through this three-stage pipeline, and each stage has less access to the one that it had before. Um there is this like if you can see vertically, the scope, the query, the mediator agent, and the Postgres. So, the scope agent is the one that actually does the thing of like storing it and then creates that function on the stage zero that generates the insights out of the um

data database that was used. And then we have the query that is the second uh agent um separate We're done. Separate container, separate sandbox, and the mediation, which is basically like it just reads the results. Um everything is in a container isolated. And what else? I wanted to give you this just as an idea concept that I did for this event, which is basically how would it look like from the end user using an app when they get on their screen and they have the choice to open Instagram or TikTok.

They see on their iPhone in this case the dynamic island saying something like, "Oh, Vanessa is meditating for the last 3 minutes. Do you want to join her or do you want to open TikTok?" And it's just it gives you that that layer of decision that it that can change the frictionless experience on the actual app. Yeah. Thank you. >> [laughter] >> Yeah, I'm sorry we have to discuss it to Uh I invite you to uh continue the discussion with RB uh just after because it was really interesting. Thank you.