CTRL/SHIFT - 13 06 2026 PARTHENOPE STAGE - WALT SCHAGEN - THE AUTONOMOUS CORE

So the autonomous core architecting the next generation of agentic startups. W Shagen is a web free architect CTO of mood studio and co-ound co-founder with experience spanning aentic systems define infrastructure and smart contract security. Uh the digital economy is undergoing a structural shift from human centric as SAS models to autonomous agentic ecosystems. And to survive this transition, next generation of startups cannot bolt AI onto legacy infrastructure. So I guess I'm leaving the floor to you to deep dive into this topic and tell us everything about it.

Yeah, thank you so much.

Slides back.

All right. Hi guys. Thank you for uh joining my presentation. So yeah, we started we started the world with the humanto human interaction and uh ultimately payments and uh now we move towards uh human to AI or agent interaction and uh you can imagine for example chat GBT where uh you open a chat and you want to buy a a sneaker and inside chat GBT you don't leave um you can actually buy the product. Uh but now we're moving one step further and that this is future talk.

So uh uh you don't actually buy this product yourself anymore. An agent is going to represent you and your agent has your best interest uh is is going to talk with the agent on the other side and then uh you buy the shoe. Uh so that's uh what this talk is about like how is this future going to look like and I hope I can give you in this presentation some inspiration so that you get some good ideas and uh maybe you can help uh uh give shape to this future.

Sure. Uh yeah a little bit about me. I heard already some uh some nice words but uh yeah personally I have a uh I have a background in nanotechnology a bit of business and uh I have been a professional poker player so I like to approach things from a scientific point of view um but also uh I uh enjoy game theory so uh it really ties into this agent to agent uh concept because it's uh it's role based uh my agent does something, the other agent responds. What do I respond back? And if we think about that, then we can imagine how our future is going to look like and maybe we can identify where uh potential breaking points are.

Um so this is a bit of an overview of the things uh that we'll be uh covering in this presentation potential protocols that we can use uh and some of these protocols uh overlap. So you don't need all of them but you can use them for specific uh applications and the industry. So this is a bit of a well everybody knows these players right this is the the cloud and the LLM providers but they have a stake in this uh concept because they want to be the one who provides uh the agent they want to be the one who the infrastructure is being built upon and so it's in their best interest to uh have the open source protocols uh that uh you are going to be building on and so well this is their business model and the the the the protocols they kind of give away for free in order to uh increase the adoption and ultimately have their platform as the the place to be. The market uh well you see the number already it's it's trillions. The market is enormous but is a potential market.

This is really wild estimates from various uh uh yeah various companies like Deoid and you name them the big four and u yeah it's it's really hard to predict uh because we're not there yet and how uh how many people are going to use it depends also uh on a kind of exponential adoption like now you have the infrastructure being laid as in payment providers like Stripe. Uh there's crypto, there's X42 protocol that I later cover for micro payments. Uh there's the the LLM that's getting better and better. So the infrastructure is being laid for us to build these agents. uh but the more agents come available that can actually do useful services that can uh provide these final pieces of the puzzle of the infrastructure the more the network as a whole gains value.

Um so yeah it it it will be exponential growth at some point uh but for now it's it's it's very limited and also it's hard to measure um because these interaction are uh direct between uh companies like you you don't need like a play store or an app store to uh to do this. This these protocols are just ways for uh agents to communicate. It's a language but it's it's it's not in that sense it's not being monetized although well come later there's one example where actually is monetized but that's open AI um so anyway uh we we we have the complete stack and we'll go over that uh in a little bit. Um it's basically a a big funnel where you go from like okay how do agents even find each other until like what what if we have a legal issue who is the guilty party cuz you can imagine that there is such a there's the developer who built the agent there's the agent from the one side the agent from the other side uh there's the the the buyer uh the seller uh when a transaction goes wrong and the agent does something wrong who is to blame Uh yeah well one of the things nowadays is okay well if it goes wrong you you just didn't build your agent well and keep going but there's there's serious companies uh you know they they have done a legal issue so there is also a protocol for that made by Google so uh we'll cover that one as well um did I just go absolutely in the wrong direction. We have an example case uh a travel agency.

So, we're going to make a trip to Naples. Some of you are already from Naples, but uh I came I come from Amsterdam. So, uh I I came here. I thought it was was interesting. You you just go on a trip.

You have a certain budget. You have certain needs. Uh I want to do uh fun things. Uh but also a question can be like uh what is the weather going to be? And you can have your agent uh find out through for example using micro payments like okay uh maybe I need to pay a specific uh provider uh a little bit to get the weather uh like we read on the weather for my trip and then based on that we can start the process of interacting and negotiating with other agent to get bids on uh your plane your hotel anything and combined all together uh you book your trip and um yeah one thing if if you book such a trip you would like it to be uh one or none like you don't want to book your hotel and then have your plane and not come through.

So there is a bit of a issue with that and then there's a solution that they now have is is is a TTL time uh time to life time to live um which is that a bit from uh uh an agent is valid only for a specific period of time. Only problem with that is that you can get race conditions. So uh u if some some agent doesn't uh respond for example their back end database is not working and they're just waiting waiting waiting with it now you don't know if let's say you have still have 5 seconds for the TTL to expire can you still do all of it or not can you send it all to out to everyone and get all of it back with certainty let me come back uh to that later because there's a solution for that that uh is might be interesting because of this u blockchain uh based event. Um but first the the the web web two uh phase of uh discovery. So you have agent to agent protocol.

I don't know if some of you have heard from it but it's a way to uh you basically provide a page on your website and uh other agents can read it and it's basically your storefront saying what you do what you have to sell. So if you're a travel agency, you can say like, "Hey, I'm a travel agency. Come to me. I have nice uh trips to offer, complete packages." Uh and then if you want to go more into detail, you can use a model context protocol where you say like, okay, specifically how are we going to tackle these problems.

So these are basically the aisles in your store like I have snacks, I have food, I have drinks. This is how we're going to uh make you a meal. That's the specific uh solutions and uh for the developers inside your uh your um your agent card you can already point to to your MCP.json file. So you can already say well if you want to know the details it's over there.

Uh this way it will you will you can guide the the agent from the other side faster. Um, web 3 has also a a a value here in that you can verify the credibility of an agent. So basically uh you have an NFT that you put out there and you say okay uh how often does a transaction goes right? Uh maybe you need to uh put on the blockchain a proof of payment and you have a small uh payment in addition to give feedback and then you can compare agents by uh their performance and agents can then see like okay this agent must be pretty reliable because uh he has a response uh uh of positive response of more than 99% and uh it's reliable. Um but what you can already see from this is that this might be uh gamifiable in the sense that uh if it's a fraction for for a scent for feedback and let's say uh you have a couple uh a couple hundred bucks then you can produce quite a bit of feedback like uh you just loop and you you uh watch straight the feedback.

Um I have checked a couple of these um the agent to agent uh NFT uploads and they have a couple thousand of replies. So with a couple dollar you can already gify it. So this is might not be the the end solution but if you provide a have to provide a unique proof of payment then it becomes maybe a bit more difficult. So uh you have your travel agency agent and well you want to be found. First thing is okay how how do I end up in the search engines?

Traditionally we have SEO uh but now uh we have gio. So uh one way to optimize that is if you have a website that is basically like a FAQ page. So, it's really short answers to questions because that's what the the the LLMs like the the chat GPTs of this world, the the well any model you name it, they like these short concise answers and then they can quote you and you end up in their uh you end up in their response. That's what you want. And uh one uh specific field is uh AEO uh which is basically the the zero field in Google search.

So u if you ask anything in Google and it's a question you your the AI will uh go to work and this is basically you're you're ahead of any any one who wants to advertise you're in front of them because that's the answer to the question you're in there. Uh so that's an important field but of course the problem is that nowadays with faster and faster cloud-based coding and whatever u you know everybody will have uh so many agents that this is not going to be enough anymore but it's a start. Um so what you're going to have to do is you're going to have to advertise. You're going to have to pay to be listed. So an agent uh comes and says well what travel agency shall I use uh I can use any travel agency I don't care uh then you can bid on it and with this protocol you can initiate this bidding process the first uh phase the uh the ADCP phase is is semantic so it's it's more like a conversation like hey what do you do well uh uh I have here a client and he wants to uh uh do a trip to Naples.

Then you as an agent, you can reply like, "Oh, that seems like a great thing. I want to bid on that." And uh that that's the semantic part of the process. Then you go into the the the hard the hardcore part of the process like, "Okay, now we actually need to bid for for money. How much are we going to bid?"

And that is uh like it's it's JSON. It's it's it's typed, right? Because one problem with agents is that they sometimes they tend to just reply with like, "Oh, that's interesting. Um, but I didn't quite get your question uh yet. Can you please clarify?"

No, no, we now we need to now we really need to have numbers. So, uh, you can uh apply this protocol in different phases of the negotiation uh process. You can go uh as for example, let's say you're a um you you you can book plane tickets, you can already put it in there, but in a negotiation between the buyer, the agent buyer and the plane ticket seller or when in my example uh the the plane ticket seller and the intermediate like a Lonely Planet who then goes to the client. So there's different uh B2B or B2C types of uh potential of negotiation. Then there is a then there are commerce protocols.

This is more like the uh old-fashioned kind of like okay you have a shopping basket and you want to so you have a web shop uh and you have a shopping basket you want to check out. uh but maybe uh you want to negotiate around like the details of this uh checkout. Maybe you want to apply a coupon. Maybe there's limited supply of some of the things in the web shop. Uh here uh it with these kind of protocols is possible to negotiate about that.

Uh there's there's two two versions of this. One is from OpenAI and one is from Google. Uh we're going to have a quick look at um at what the the the card negotiation uh can contain. I mentioned already uh the coupons you can you can uh you can bid with with the user agent and vice versa. uh maybe uh there's certain stock that you want to take into account and um one of the one of the protocols is from OpenAI and they initially started as the idea of like okay we're going to just check out inside ChatGpt and that's it and there's no there's going to be no we're going to be the merchant of record there's no individual small uh web shops who are going to do the the checkout themselves.

Problem was they couldn't solve it for all the small web shops because everyone had a different way to handle their supply coupons uh local regulation. So it was difficult and ultimately they didn't manage. So they had to kind of give back the power to the uh to the small web shops. Uh but still they want to make it a business model. So they charge 4%.

Uh, and if you want want to do the do do their uh uh web traffic, then you're going to have to uh do that. Um, but the other protocol from Google charges nothing. Uh, but yeah, then you're not uh you're missing part of the traffic. Um, then um there's another protocol that is more for enterprise. So maybe less for uh suitable for this audience where maybe we're focusing more on startups.

But uh if you want to dive into the the legal uh issues that I discussed for for like who is the the guilty party, you need to record what happened. So you need to record in in the the the conversation between an agent and another agent who uh who said what. uh and only then you can sort of determine where it went wrong and who's guilty for doing something uh not the right way. Uh and they do that through mandates. So that's kind of like a unique um a unique code they generate for each part of the uh journey.

Um okay so then we reach the the checkout phase or the the the payment phase. Um, and I thought I would share a little bit of general knowledge about how payments can work because there's basically two ways. You can push and you can pull. Uh, a pull payment is kind of like the uh the credit card uh way where uh the user doesn't uh initiate the payment. it's it's being pulled away from their uh account and uh they can uh uh have a a legal period of I think in Europe is 8 weeks where they can uh say no that actually was not correct and then you cannot do anything against it they they will get their money back.

So this is a expensive uh way but it has advantages as well. Also it has a it's more it's safer for the uh user sometimes because they're uh insured. So um for microp payments this is what not suitable at all. You need push payment. Basically the user says I'm going to pay uh and then they have no rights.

They basically give all the rights away but it al doesn't cost anything. So there's there's nothing to ensure or know uh yeah there there's no risk. Uh and one thing I thought it was interesting I looked up that effectively merchants lose 8% per dollar uh on uh pool payments. So effectively with disputes and everything combined they lose 8 cents per dollar I if you make such a payment. Um so this is both are huge industries of course.

Um so machine payments protocol is uh is is where the the machine to machine interaction comes. Uh and this is of course the the the push uh payment and this is the uh the alternative to X42 protocol which is mere more crypto uh based. So X42 protocol is uh related to USDC payments uh on on base and this offers more also kind of stratify uh payment options like uh you have with stripe uh integrations you have like uh billing and invoices. So if you're an enterprise that's not per se interested in DeFi only, you you want to use machine payments protocol and if you have for example an arbitrage bot that just wants to trade and uh uh maybe make money from data sources uh you have you don't you maybe you don't care about uh these uh traditional streams. Uh yeah and so uh this uh this MPPP they they use tempo as a blockchain which is uh one slide I kind of want to skip uh for time constraints.

Um but yeah you can look into them. This is this is just a a business they they launched a a blockchain and they have really cheap uh uh gas fee because this basically permissions. So, uh this is a little bit of a comparison between the two uh protocols that you can use and uh yeah the the funny thing is like we're talking about protocols that have started in like 2025 like this has been uh a year old and uh the road maps are really strongly converging. So whatever features you're looking at now that one has and the other one doesn't a year from now they're looking at each other and they're going to have the same. So I wouldn't look at it uh too uh too much into like oh is stripe on available in the one side and not in the other.

Stripe is available in both you know like they want to they want to do everything. So um yeah only thing is that X42 is more like blockchain native and and MVP is yeah a bit more uh solution for enterprise. Um so yeah this we're almost reaching the the conclusion of my my talk and um yeah I I was hoping to today to inspire you guys uh with a few ideas and uh one of my thoughts was okay we need to not build simply build agents uh because well you might be replaced or uh you want to build the the infrastructure or the the the fundamental al uh break points essentially. Uh so I I put down a couple of ideas of what uh what potentially can be done. Uh so one for one example is uh for example a price uh aggregator.

Uh I think uh so here we we just talked about uh uh this that uh well how do you know if a if an agent gives you the best price if your agent is with you because they can also accept bids from uh the the other agent like how do you know that that you get the best price? Well, you can make a aggregator protocol and you start comparing prices maybe real time maybe on aggre per month maybe over a bunch of users. there's a whole bunch of things that you can do to start comparing that uh that those kind of things. Um other problems uh can be for example All right well um feel free to to to read some about some of these uh um about some of these ideas but basically there's there's friction some sometimes because and it can be long-term friction of for example open AI with Gemini with uh you name it cloud and they have different visions if you can see those friction points and you can combine them. You can offer, for example, an SDK.

You offer a solution that is interoperable. That can be a long-term thing that nobody can uh compete with because they don't want to. And then you have a solution that uh is not uh being competed away in a year from now because they come uh with an upgrade to their protocol. Um so yeah, that was kind of it. uh maybe make a photo of the references if you think it's interesting.

Um then you can look up a bunch of data yourself and um with that I would like to thank you very much. Thank you so much. And we have a beautiful stand of mood studio here. So I'm sure you can find B for all the questions and deep dive into more details.

Yeah, thank you so much. Feel free to come over and discuss the topic with me.

Thanks again.