Yeomen.ai - Elevate your game! | Devcon SEA
Devcon·Thu, Oct 9, 2025, 12:00 AM
Web3 games bring about possibilities for autonomous worlds that traditional games are not able to offer. Yeomen.ai makes the on-chain data available to the masses in simple dashboards. Yeomen.ai also offers on-chain extension of autonomous worlds to automate and transform game play. Command has an Intents based services layer to offer services on Autonomous Worlds that can expand to all Web3. Yeomen.ai can work with MUD or Dojo powered onchain games and financial applications in future. Speaker(s): Rohan Abraham, Roshan Abraham Skill level: Beginner Track: [CLS] MUD Community-Led Session, by 0xPARC Keywords: Autonomous World, Collective Intelligence, Intents Follow us: https://twitter.com/efdevcon, https://twitter.com/ethereum, https://warpcast.com/devcon Learn more about devcon: https://www.devcon.org/ Learn more about ethereum: https://ethereum.org/ Visit the https://archive.devcon.org/ to gain access to the entire library of Devcon talks with the ease of filtering, playlists, personalized suggestions, decentralized access on Swarm, IPFS and more. Devcon is the Ethereum conference for developers, researchers, thinkers, and makers. Devcon SEA was held in Bangkok, Thailand on Nov 12 - Nov 15, 2024. Devcon is organized and presented by the Ethereum Foundation. To find out more, please visit https://ethereum.foundation/
Transcript
[Music] with autonomous worlds for quite a while and I'd love to show them to you so U um the outline of the talk today we're talking about exploring autonomous worlds where we will showcase our awesome dashboards and I'll tell you how uh we help developers build great games and also how we give Gamers an edge we then we'll talk about command our brand new platform for autonomous worlds we'll talk about how we automate and mod games and our intense platform and our very GL Grand plans for the future so there are quite a few slides let me jump straight in so exploring autonomous worlds uh isn't all the data on chain uh yes it is uh this is this is what you see on chain uh so uh it it's on a block Explorer you see something like this uh a block Explorer are very helpful for erc20 transfers and things of where you transfer money and all of that but here you get to see the raw input it doesn't you don't it doesn't make sense what you have to do is you have to decode the data to make it actually useful once you decode it you get things like this you get to see things like login Player move you get to see what action was performed what the input parameters was and this this opens up interesting possibilities emerge uh if we were to index it and store it at scale we can then create nice tables graphs charts and all of that and uh this is an example from Eve online play test uh uh you can see the data and you can compare it with each other so on a table you get to see U uh compare and contrast as much visual representation makes a lot of sense um it gets even better if we make the game make the data game aware so let me show you an example uh what you see here is all the fleet movements with in the primordium universe and the same data but this time it's aware of the game so you can see all the fleet movements so uh as a developer when you look at it you'll be like hey why are these people making these incredible long Journeys uh and uh you see like clusters of movement you also see these massive long trips and you can then tailor make your game redesign the game it's also great for gamers to see what the hot spots are and things this is one more example this is uh a 2d representation of biomes 3D world uh if you if you look carefully you might see white spots those are the mind uh ores and the red ones are to be mined again these are very helpful for gamers and developers alike so great we have useful dashboards uh one other thing so on on a blockchain the live State can be queried so uh there all the block all the all the work has gone to make the state um pretty the consensus is on the state but what about the history of the state what if you want to see what happened back in time at at a point in time say say a month ago or six months ago or you want to see like how many players grouped together to fight the battle that they won and whose Health was how much at that point in time so if you want to go back to Something in point in time it's much harder and that's where Yan's time travel indexer comes in uh I showed you this morning quickly but what you can do is uh I have a video here so uh this is draw Tech so you're able to show you're able to move back and forth and you're able to see what the state of the blockchain was at that point in time so all of this comes from our mud indexer as you click it generates a whole it pulls all the data from all the different tables at that point in time and gets the data to you um so uh yeah so clicking back and forward on that demo it it got us thinking hey what if we click forward one second at a time or match the block speed and that's where we got the idea for a live replay so uh I want to try this um I want to replay the first ever Sky Strife match on Redstone this was back in May May 1st I think and I I want to do the live replay let me check if it works uh okay it's move screens okay so yeah it didn't work this morning but uh as it loads let me tell you what it's doing so this is loading the state uh from May and once it loads it tries to like replay the game so you can actually see what the game is about okay it's a blank screen give it a minute so yeah perfect so it it loads so what you're looking at is the actual first ever game on Redstone with Sky Strife you can see luden's Ben from West and proof of Jake and I think that's Peter Pan they're battling it out in the first ever Sky strive game so uh this is all from M indexer working with the data so yeah it's pretty cool and so cool so let me go back and so that that's so I've covered Yen for exploration of data we we've discussed the rich dashboards and also the ability to like uh visualize data go back to time travel capabilities and all of that uh I want to move ahead from exploration to action so this is where we actually make onchain actions we had a few questions while we're thinking about human command the questions were how do we extend an autonomous world and how do we incentivize developers to actually EX in autonomous worlds and how do we give security guarantees to users we don't want users to go to random websites or or or download things to actually run mods and automation so and also we were thinking what the mechanisms of interaction between all these various parties would be so I want to show you Yan command but Yan command is a few things uh and let me start by the first thing an human command is an app store for autonomous worlds so uh this is uh y command if you go in there now you see this this is a list of it's a gallery of apps each app is a mod or an automation submitted by a developer you could submit it anybody could submit it and they're all listed there and uh you could you we have uh mods and automation for biomes primordium quite a few other games and then if you were to click on one pretty much like what you see on an app store uh you get ratings reviews you uh there is an exploration explanation of what the app actually does and uh you get to see related apps um and also you get to view the code which is important you can actually see what the actual thing does and once you're happy with the app you click on install and that's when uh the are our sandbox environment for users come in uh when you click on install we tell you hey this app is looking for access to this world contract in this case it's a biomes World contract it's looking to mine move jump build and these are the actions it wants to do if you feel yes it's okay it's safe let me do install it you install it and this is uh a list of all the apps running on Yan command and uh you can either stop them start them you control them you uninstall them all kinds of things like that there are various apps there are also apps which send you telegram alerts each time something happens depends on chain and uh yeah so uh there are quite a few things uh quite a few apps already in there so an App Store with the sandbox environment is already awesome uh we can do quite a few things and uh we've already we have quite a few we had a few people testing it uh this is uh uh Ben from VD uh he's his team's actually tested it and what he says is command is a game changer and one of the coolest Platforms in the space it kills a fud that Bots will fully onchain games by democratizing access to Automation and improving the gameplay experience bullish AF so uh uh essentially the keyword here is democratize so we are able to like create a simple interface users are able to click on stuff they're able to like mod automate uh and and and yeah experience the autonomous World in a different way uh and everybody gets to benefit so uh uh Do you want to build a yman all you have to do is uh head to command. um. and then you can get started it's it's pretty easy because all the other apps are also open source you can just go there click on a click on the fork button you're able to like improve the apps submit it back to the app store for other users they're able to play and all of that so which is great and you reach a bunch of users uh and yeah so what's next for command we've learned a lot by building all that you've seen till now and command as well and now I want to I want to I want to really invite ideas questions feedback as we think about what next what's next for command so uh um Yan command as an intense platform for autonomous worlds now um think about this say say for example you are in biomes and um you're building a Wonder it's it's creative work you're you're putting that together and you need 125 stone blocks so um there are two options you can go in and and craft these blocks that's actually labor intensive you actually you have to switch off from your creative mode you'll have to do something else or the other option is to go make a physical trip digital physics apply and things like that you'll have to make a trip to the Bazar go buy it and things like that so what do you do if you don't want to do either of these things what do you do um what you really need is a deliver in biomes or some kind of a grab if you were in Thailand that's uh something like a service that would give it to you and uh uh we can't really build this on an autonomous world now because we need an intense powered Services layer on chain and once we have this we can actually launch services like uber bounties or job boats and things like that so uh U let me let me tell you how an intense platform would work first there is the intent so your intent is simple it's get 125 stone blocks to me in 50 minutes you have a clear high level goal the constraints are defined is just four constraints I need 125 and the type is stone blocks and the target is me in this case and I need it in 50 minutes uh this is verifiable and auditable so you get to see it on chain and this is permissionless fulfillment as in anybody can do the job anybody can get it to me and anything is fine uh and you can and again only the intent is specified you don't have to specify things like delivery routes or how you get it you just have to get it to me and get me this fulfill my intent and uh the solvers work to satisfy your intent um and every app on every different world really will have unique intent so on cig GOI you might have you might say feed my Cami one snack every five hours which is a continuous type of an intent in Cafe Cosmos you might say things like acquire the land and farm on it which is again composable in biomes you might say something like guard my force field while I'm away which is a conditional intent so yes intents do vary based on the game uh we see two kinds of intents really based on the transaction execution externally fulfilled which are service requests and self- executed which is part of automation so let me talk about the externally fulfilled intents so this is an intent that we were just talking about now in biomes so as a user you would broadcast this intent that you want something service providers would actually bid in real time so they say hey I'll do it for $150 there are random amounts and um um ratings and reviews are based on the past um past activity um you could actually manually select one or even let Yen uh automate this hiring process based on your pref your preferences of how you want it to be assigned and and so this is an example of an externally fulfilled intent or a service request the other type is is also interesting these are self- executed intents so uh say for example you have a dark forest or a complex game and you want to find a path from A to B now these are like hard paths it's not trivial you broadcast this intent again and solvers this time around they compete to find a path but they don't execute the transaction they work on a they work on a path they then submit the solution to you to get a payout uh if they don't want to give you the whole solution it could be a ZK proof or something like that the solvers use offchain compute they could be doing AI inferences complex calculations whatever in in this case the actual execution is by you and the transaction is could be crafted by the solver this is great for automation cases uh command in this case has to verify the alignment with your intent and with what the solution is from the user from from the uh solver so great and this is one interesting concept like uh Yan could automate both the service request and and the and an automation so for example example yman might detect that you're running low on something it might fire off a service request and another worker another person running yman would actually run the automation script to perform that perform the Tas so command would really power both the Uber driver and the Uber the person who requesting it so all that could be automated uh to achieve all this we really need an intense standard with an intense platform um and it needs to be effective with you will need a single standardized intent across all the different autonomous worlds so it it should be easy it should be one single standard and we enable solvers to compete for the best solution it has to be a Marketplace model we could also look at things like U uh cross-chain intent fulfillment uh we need to we need the interance platform to be intelligent we'll have ai solvers and also we need to somehow maintain alignment with these intents uh to facilitate all the seamless collaboration between humans and AI solvers and lastly we'll need it to be responsive uh if for example you were on you were booking a grab uh you would actually put the request out and then you can edit it you can change your destination you can cancel it things like that so these things are also important on an intern platform so uh uh once we have this all working it's going to be a pretty sophisticated intense platform so the the Focus right now is on the on the top quadrant there the fully onchain games are high transaction volumes and low Financial Risk like if you if you get if you forget to feed your Camy or something like that the the possible loss is is minimal compared to like something on a d or or a DEX so uh uh it could start that's an ideal starting point because it's it's high transaction it gets tested as alvarus was talking about this morning we can go into the different stages of making it work making it fast making it scale so we can also increase it from security so start with something and then move to more security demanding applications so yes uh in all these points I'd like to stress that the intent solvers are the real heroes they're actually the real workers that get things done so which brings me to my final aspect of human command which is a solver control and ownership platform so uh as described before the solvers are the ones that actually do it do the work and uh they create value and they ACR value so uh you pay them for the work that they do and all of that so they they will definitely ACR value uh the solvers are going to be busy in a onchain world uh it's it's continuous it never stops uh what do you do at the end of the day after you play a game do you log out of an autonomous world uh why not be a gamer by day and a solver by night I think what the example is say for example you the the vision with Tesla and the robo taxi is you get a Tesla you drive it around when you don't need it anymore you send it away it actually does uh takes other jobs it does other stuff and then and then it comes back when you wanted so if you have a player that's fairly advanced in an autonomous world you could um you could then send it out to do jobs and and things like that so which is going to be which is interesting um this is one of the last uh pictures here so uh a a solver will have different entities running at and and behind it so the key parties are first of firstly we'll have developers developers are individuals that actually build the logic they put it out there uh we will need hardware for these solvers they will be they'll be running AI inferences or or complex CPU intensive jobs and then you have IP contributors IP contributors are interesting because you don't have to like uh you don't have to write long pages of of ideas or something like that we've we've been in Discord groups where the ideas are very simple they're like hey why don't you mine in a staircase pattern or something like that those ideas also are pretty important and then uh any system would have to figure out how to how to compensate these these users and finally we have external Services uh for any agentic framework will have access to realtime data and things like that so uh and finally each solver can be operated as a doll so as they get larger and they acre more value uh it can it can be pretty cool so that's a glimpse of the future of command uh so back to the original questions that I asked at the beginning uh this is how human command could evolve to answer these questions so how do we extend an autonomous world uh an app store for mods and automation along with an intense powered Services layer could be a great start how do we incentivize developers we spoke about the solver ownership model with a single owner or a DA the solvers would get paid and then we would have IP contributors and they all get paid really with with something like a solver ownership model and how do we offer some security guarantees to users uh we spoke about things like sandboxing and intern alignment at the platform level along with discovery of rate with ratings reviews to establish trust with developers and then uh we will need an orchestration and coordination layer to manage interaction between all these different parties so uh U yeah I was asked like what is the meaning of yan uh Yan means uh Dependable diligent or loyal worker someone who does a great service a part of the dictionary definition Yan the name is inspired by Yan Waters and U the yeah Tower of London they've been um tased with God in the crown jewels for centuries and things so yeah so that's the end of the talk so thank you so much guys get in touch um this is [Applause] Q&A time yes um are you the paler of autonomous worlds is that a question yeah are you the paler of autonomous worlds yeah we could be we started as a a a dashboard system we want all this data is a really onchain everything is visible uh unlike paler which probably works on private data that you don't want access to here everything is on chain you really publish all the data we just put it all together in a usable format for users really that's yeah that's the focus all right and I think this one is uh yeah how can other people add their the games they build to Yan yeah there is some work involved in adding a game to the dashboard so all you have to do is go on to our website there's a link if you don't see your game in there already all you have to do is go to the bottom of the page and say add my game to the dashboard we will need a few things like your ABI or something like that so we know how to index it what your world contractors things like that and we could get started straight away all right so next question is really funny because it's like it's kind of just questioning the entire category it's not even a yman question but uh why are autonomous rules important like why are we doing all this how is this valuable at all yeah uh it's great so uh yeah in my view U we have people here there's a lot of energy there's a lot of activity we're producing a lot of data and then we could do things like so maybe maybe this could be the perfect place for an agentic system to take life like they polish their agent capabilities do something so there probably a lot that can come out of autonomous worlds yeah can yman command R run on my browser or Cloud that's a great question uh we've been debating it internally for quite some time where should it run at the moment human command runs entirely on your browser so it's on your device your key your your rules all that is on your device which is pretty pretty cool for U from a uh decentralization point of view the only annoying thing is if your laptop switched off or something like you cannot it stops running uh but on the flip side I have Yen command running on a remote desktop on a server and and I and I ran it earlier today so it does stuff so you just have to get a remote desktop like open up a browser and it could run on a cloud really okay so the next question is one I'm also personally very interested in um go into detail in how intelligent agents that you create can can get integrated into these games how do you how do you bring them into the world yeah yes uh there are there are a few ways we think about it first is like an agent who owns an agent like so in terms of an agent who controls it so I spoke about different owners so there are there are people who provide the hardware there people who who bring who develop it there's IP there's quite a few things so uh in in there so you we see those people as the people who control and own the agent and then like how do you bring them in they're already in they have access to the apis and things like that that's how we interact with the world and uh yeah we could yeah that's okay and I think this is probably the last question what is your vision for the next two plus years of yman yeah so Yen's been uh it takes time to build a platform like this and then we've been looking at it for a while I think everything that I showed you in the first part is all there uh we have an app store and we have um a Sandbox environment which is pretty secure uh that was our first step our next vision is to build the services layer to power autonomous worlds and possibly Beyond autonomous worlds to web 3 as a whole so it it'll take time it'll take a couple of years probably and U does that answer the question was there anything yeah and I think that's it I think I want to close off by saying Peter theel says that if you want to win you got to be contrarian and right the problem with Peter theel is paler can only run on one world but over here we create 100 worlds and Yen runs on all these worlds so Yen can its market cap is 100x paler um and so Peter theel needs to invest and oh cool stuff thank you thank you guys thank everyone yeah
Automatic transcript — names and jargon may be misspelled.