Agentic Token Systems are Replacing StartUps in The Post Web era

Transcript

Hi everyone, this is Akim from Outlier Ventures and I'm pretty sad that I cannot be with you in person this year in KHN at ETHC at the token engineering track. I would really have loved to be with you but yeah I'm pretty confident that we will have other opportunities to see us soon. Okay, let me share the presentation for today about the postweb thesis and the main title here is the agentic token systems are replacing startups in the postweb era and before we go into the first part of this title like the agentic token systems and startups how they are affecting startups I first want to set like like the stage

for our postweb thesis so you understand a little bit more what we mean by this postweb thesis And before I dive in, um, let me quickly share who we are as Outlier Ventures. So, Outlier Ventures is and was the most active web 3 accelerator. Um, today since we are working in the space for more than 10 years now. We have a big investor network of more than 500 investors, big investors. And yeah, we our portfolio consists out of 400 companies at the moment, which is yeah, pretty great. And I'm pretty proud also to um have such a big network around us. Okay, then let's

start um talking about the post web. So what is the post web? A quick recap, there was web one, web two, and web three. Web one was just about reading static internet pages in the beginning of the internet. Web two was not just reading but also writing. So people could share their content, their own content like on social media platforms. And now we are basically in web three which is why we are all here. Um where it's not just about reading, not just about writing but also about owning digital assets.

And this is all great, but we believe we are now in this transition period from web three, not to web 4, but instead to the post web, where basically the web layer will become thinner and thinner over time because AI agents abstract away a lot of the interaction that is happening currently via UIs, which means we are not just reading, writing, and owning digital assets, but also delegating our intents towards AI agents and we can already see this happening right when you think about when you work with chat GBT or other chat LLM platforms where you will be visiting less and less traditional

websites because you just ask the LLM and it will do it for you. And this brings me now to the four core pillars of the post web. The first pillar is that it is intentbased which means that we as humans are expressing our intentions and AI agents are fulfilling them instead of us just manually going through any websites, clicking on buttons, filling out menus and um yeah, navigating through different pages. Now an AI agent can do this for us and we just express what we want and they do it. And second, it is hyper contextual which means these AI

agents understand what the current context is in which we are communicating with them. And they can then also for example spin up websites on the fly that are showing basically what we want and what we need to see exactly in the format and the shape and in yeah the structure that is best for our specific intent or like the specific situation we are currently in. So they really understand what's what matters and they can show us information in the most efficient and best suitable way for us.

The third pillar is that the post web will be deterministic yet adaptive where basically the two advantages of DT so distributed ledger technologies such as blockchains are combined with the adaptive nature of LLMs and AI agents. So you can combine these dynamic behaviors of agents with a very deterministic um layer of execution which also feeds into the fourth point which is that the post web is verifiable which means that the post web basically can be like in the post web all these

actions can be tracked because they can be submitted to the chain. locked on the chain and whatever an agent is doing can be followed up on chain and this means that we are having a economic shift a transition from an attention economy towards an intention economy. In the past everything was about our attention being attracted to certain platforms and especially to ads like this whole ad economy. But now it is about AI agents fulfilling our intentions the on the best and most efficient way. Which means in the old economy in the attention

economy the process was maximally extractive. So there were ads and um entities just extracting as much as possible from us while the actual outcome for us was more suboptimal. And in the post web, we believe this will be different because now the extractive nature will be less of a thing and instead these AI agents will be optimizing to a more optimal outcome for us humans. And basically the platforms and the solutions that are optimizing for optimal outcomes in terms of different intents will be the most um

competitive ones. And this brings me now to the death of classic startup theory. So in the postweb era, the foundational assumptions of traditional startups, how they are conceived, built and scaled become obsolete. Why is that? Why do we believe um that in the death of this the classic startup theory? It is because in this new world, in this postweb era, the advantages of user experiences will be limited due to the reasons I just expressed because UX, user experience and UIs will be less relevant because AI agents will just abstract them away.

Which also brings us to limited product modes because every almost every product can now be built in a matter of very short time like days maybe weeks but in many cases in days where in the past it required a whole engineering team to build a product over many weeks and months maybe sometimes years and this means that also loyalty and retention will be less important because AI agents don't care about loyalty and retention. They are just optimizing to fulfill the intent in the most efficient way. um as opposed to the past or to this current attention economy where people are gravitating and sticking to

certain brands. AI agents don't care about these brands which means traditional flywheels also break and therefore in some cases minimize network effects and therefore also limiting impact of growth capital because um yeah it depends really how it's used. So I think investing in capital obviously is still needed but it is it will be very different from what we see in the moment how it's used in the moment because it's not important to use capital to build like a startup team that is building a certain product but instead what you want you want to support and seed self

bootstrapping and self-emerging systems what I will talk about uh in a minute as Well, and this whole paradigm challenges this winner takes it all games. Why? Because almost everything is getting commoditized. But when traditional startups are dead, the question is what to do instead? The answer is systems. So success in this future era requires designing, building, coordinating systems and not startups or status products. So founders must shift their mindset from company building to creating ecosystems, co-creating

ecosystems and finding system fit instead of product market fit. Product market fit will still play a role but it's basically a subcategory of a larger thinking of the whole system. how the whole system should work and the system needs to be designed with the right incentives, alignment and participation models rather than just extraction because margins will be diminishing in this postweb era and they will be replaced by more efficient systems. Um this reminds me also on the already existing Dows decentralized autonomous organizations that we are having currently but I think they are pretty

inefficient and inaccessible for people overall. Um, I think AI agents can help a lot by abstracting away complexity of Dows, making them more accessible and democratizing them in a way um more. And I really like also the the work of Angela and her team like for example metrics based voting, how quantitative metrics shape better decisions and other articles that yeah she already published. So please have a look into that because I think this is actually going into the right direction where we find frameworks on how to build the right coordinating systems that are not just looking to coordinate human

decision making but also involve AI agents um in a native way. Yeah. And to build these systems, the token engineering discipline is one of the most important ones of all. So token engineering in our opinion will become very crucial and fundamental in this postweb era because it's combining economics, game theory, computer science, systems engineering. All this is required to create these systems that are also robust, adaptive and that they are antifragile. So they are adjusting for constant systemic alignment. So they

are not just self-emerging but they are also adaptive to external changes. And for this we can use tools that we already know such as game theory, intention theory, incentive and mechanism design, governance and Dows. So in other words, these cryptonative concepts become the backbone of AI systems. And some of you remember that in the past we at Outlaw Ventures developed this token utility canvas that is still being used until today and it's a great tool to help founders in it in the traditional startup world to build and

to define their token. And we looked at it again through the lens of the post web and essentially extended it and changed it with respect to these post web attributes. Overall we would say it will not change all too much but some things will definitely change from the token design perspective. So first of all the stakeholder side will change drastically because now we're not just looking at humans but we also need to account for AI agents. So AI agents need to be part of the equation and for the incentive modeling AI agents need to be part of the

equation like essentially stakeholders willing to get out as much or as many incentives as possible is something that we are already aware of. But I think this will be even becoming more important in this agentic era where these AI agents might also try to optimize for maximum incentive extraction. Which brings us again back to this question on how to distribute incentives in a way that it bootstraps essentially the most value of a product or a system in this case. Um yeah the in the most effective way. And for this we also need to account for the

whole system fit. So we put another tile here system fit evaluation where we essentially look into how the whole system get works together how um robust it is and again like how all these parts work together under the lens of the post web. We will be publishing a dedicated chapter three on the post web. So the first two chapters have already been published but in the third chapter like this whole token engineering part and the changes with respect to token designs what I just teased here they will be um yeah talked about way more into in way more depth than I can cover

in this talk unfortunately but yeah you can really look forward to to this third chapter that we will be publishing soon. Let's have a look at the 10 principles of the post web. First, built for hypermodularity over monolithic platforms. Second, designed primarily for agents but with people in mind. So, AI agents are not and bots are not an accident to your protocol, but they are a core aspect of it. Three, optimize for solving systemic intention efficiently with marginal rent seeking. So you want AI agents to plug in and to essentially optimize any inefficiencies out and that

you already expect that margins will diminish over time because as I said a lot will be getting commoditized. Four, focus on system fit through alignment and overall improvement like setting the right alignment through the right incentives and participation models and compensation models for participation. Five, leverage game theory to incentivize positive emergent behaviors that we already know from the yeah blockchain networks that we have been developing in the past. So we can leverage this here. Six, don't rely on static modes and business models.

Prepare for continuous disruption. So in the best case the system is built in a way that it can react and adapt with respect to external changes. Seven prioritize contribution and participation versus passive consumption and extraction. Right? So when you build um a highly efficient contribution and participation system, this will then out compete essentially other models that are more monolithic and um just are willing to scrape in the most profit.

Right? So these decentralized contri contribution and participation models are way more efficient when they are done the right way and when we leverage AI agents to make them yeah quicker more accessible as I said before eight are organizationally built for multi- aent coordination over centralized control so AI agents will play a bigger role also in decision- making and coordination over time nine adapt evolutionary incentive design and token engineering. So incentives should not be static but they should be dynamic. So in the past I wrote also an article about adoption adjusted vesting

which goes also in this direction. So I think it's it's really a bad thing that we just give out incentives on a static cadence. This does not make sense at all from an economic perspective. But instead you want to make these incentives as effective as possible for the system and therefore you need the right monitoring um approaches and adjustment algorithms or also decision- making from AI agents in combination. 10 ensure robust data creation, collection, sanity and usage.

So data is still very crucial in this postweb era because as all of us know these AI agents are only as effective as they have visibility on the the right data and instructions they need to fulfill an intent. Right? Without the right data without the right tools they can not do all too much. So this is very important whenever you build a system. um ensure that you have a very clean um data collection um in place and that you make the right data accessible to the AI agents um that they need to do the and fulfill the intents.

So these are the conclusions. The post web is not about creating the next super app or unicorn startup. It is about participating in intelligent decentralized systems that align incentives self-organize self-optimize and evolve together. So founders who understand and adapt to these principles will be best position positioned to create lasting impact and sustainable positions uh in in its economy. I think the founders who get these kind of thinking and principles first, they will be the ones who will um essentially be the winners in the end.

So the future belongs to those who build for the system and not for their own startup in the end. Okay, this was it for this session. What are the next steps from here? So when you are a founder or you are an ecosystem that wants to dive into this thinking more and need some advice how to do it like for your specific case, you are invited to contact me um or to visit our website on outliadventures.io or to contact me on LinkedIn. You can find me under aim and I'm really looking forward to get in contact with you. If you want to build

your own system or you want to make your ecosystem post web proof, then I would love to talk to you. Thank you so much and I wish you a great rest of the token engineering track today with the token engineering unplugged sessions and all the other talks. Um yeah, see you soon. Thank you.