# End-to-end internet games by Small Brain | Devcon SEA

- Speakers: [Small Brain](https://streameth.org/speakers/small-brain)
- Channel: [Devcon](https://streameth.org/devcon)
- Date: 2025-10-07
- Duration: 24:47
- Watch: https://streameth.org/watch/yt-4Cblt2gOIas
- YouTube: https://www.youtube.com/watch?v=4Cblt2gOIas

## Description

For the past 1.5 years, I've been building fully onchain games–games where the entire state is onchain for some reason (have launched 7!). 

There is lots of cryptographic data floating around the internet. New primitives are allowing all this data to be interoperable with each other... and even verifiable on-chain. 

I'll discuss some of this tech (tls notary, app attest, zkml, etc.) and discuss what new wild games we can build with them.

Speaker(s): Small Brain
Skill level: Intermediate
Track: Applied Cryptography
Keywords: Gaming, Mechanism design, Mobile

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] hi everyone thank you for the intro welcome to Devcon um end to end internet games that's what I'm going to talk about but before I talk about it I am going to go on a quick degression um when people who have like regular brains come and enter crypto and we talk to them one of the ways we describe it kind of endearingly is Magic internet money and I want to focus on this term a little bit specifically the the the internet part of this term um because when we say magic internet money we usually mean a couple of things we mean like culturally and community-wise a lot of the things that drive Trends in crypto are very very similar or born out of the internet we kind of also mean that like crypto isn't money backed by something physical it's backed by something digital and so it's internet native in that way but the definition I am most concerned with is the one of Internet native money which is a definition that actually like was brought up and and you know people have been talking about it since 1995 when they've been criticizing the fact that there's no like form of payment native to the internet that flows across the internet the same way packets of data do um and in some senses crypto maybe has met this goal right like I can do programmatic payments uh the payments happened 247 like right now in some cases they happen nearly instantly but I think the point I want to put in your heads is kind of an obvious truth that we know um- which is that interacting with blockchains just like feels very very different than interacting with the internet like in in fact interacting with blockchains feels like you're in this weird parallel reality that's kind of next to the internet um with different physics and different rules and like gravity Works upside down it's like super weird like even as somebody who's like internet native and grew up on the Internet like when you walk into the world of blockchains um you you start feeling very different and you need to relearn a bunch of stuff and I think like in that way crypto as it is now like does not meet the definition of Internet native money um and and to take like something Albert KN said yesterday which I really really liked is like ethereum is like an island like everything works super nicely on this island and things are great and there's a few narrow Bridges to the real world and those bridges are getting like larger and better but they're still like Bridges and there's still few and far between and maybe it's more so like an archipelago you know whatever um the other point I want to make is that when you do cross this Gap lots of value is created like two of the most consensus valuable things this cycle stable coins and poly Market both kind of come from applications where you know in some sort of very simple way blockchains are connected to the real world um in stable coins uh you know somebody custodies local currency and gives out token in poly Market markets are resolved by an optimistic Oracle somebody says this is a resolution to a market and then there's a voting period where governance can decide whether that's okay or not in both of these cases there's like kind of significant operational overhead um but but there's a lot of value that's created kind of when these two separate worlds touch one interesting thing I'll note is like and this is just like kind of a Vibes based Point honestly this talk is kind of just a Vibes based talk uh we'll get into some specific stuff soon but it's going to be mostly Vibes one interesting thing that I'll note is like if this is you know the real world and then this thing is like the world of of bits and the internet um stable coins and poly market and things like that like crypto applications which jump out of the blockchain kind of miss the internet entirely I'm not saying that this is good or bad I'm just saying that it's interesting like as somebody you know thinking from first principls I would have expected a bunch of Internet data end up on the Chain like before we have like onchain Vapes and stuff you know just surprising that we ended up in the physical world before we ended up in the internet world and that's what I want to talk about today is I want to talk about like how we can make real magic internet money and ethereum can write more programs than I own two and you own three and you transfer me one so like maybe more largely like how can we make magic internet apps and by this I mean apps which don't feel like like this but which feel like this like ingrained with the fabric of the internet I I don't have a good answer but I do have three different Sparks and these three different Sparks are just three different Sparks of tech which I think enable people to build magic internet apps which we're just starting to see now and some of you may heard of and some of you might not have heard of but I'm going to spend some time um talking about these three different Sparks and what their trade-offs are and then we will open up for questions oh also I think like just shoot your hand up if you're confused I'm down to do that as well but anyways here we are so the first Spark is TLS notary and TLS notary how many of you have heard of TLS notaries raise your hand yeah sweet most of you I think like this is the most common common one out of the next three things that I'm going to talk about but TLS notary has a lot to do with this little uh lock you see in a browser when you visit any website on the internet like at the top of your browser you usually see this little lock like if I go to google.com there's like a green lock lock and that lock makes me feel good it makes me feel good because I know a couple of things because of it I know that I'm talking to actual google.com and I know that my messages with google.com are encrypted like nobody in the middle can see them only me and Google can see them this is because of Transport layer security and generally how this works I'm going to be very reductive but generally how this works is there two steps there's some sort of handshake between me and Google where I verify their SSL CT to like know they're actually Google and then we create a shared key um and then it just submit symmetric encryption the whole way from there uh TLS notary is a way and so like with with the context of what I'm talking about what you should be thinking or what you might be thinking is like all right there's some cryptography going on when I visit any website on the internet when I get any response from a website on the internet like what if I could hook into this cryptography to bring it onchain um and TLS notary is a way to do that with a bunch of caveats so before we get to how this thing works we're going to start with a simpler problem which is like if me and David are doing something together and like I want to convince David that I got a valid web response from a website what TLS notary would do is instead of me engaging in TLS with a web server me and David jointly engage in TLS with a web server and we we do a multi-party computation and we create that symmetric encryption key together and we do it in such a way where I don't need to reveal most of the data in my TLS connection I can only selectively reveal what I want so for example like I might like only show David that I have like $110 in my Wells Fargo account by like logging on to Wells Fargo and David can like kind of know beyond a doubt that that like that is true um the problem is like once we have this how do we bring it to the chain and this is where the notary comes in the notary is essentially a hack because I can't do this sort of multi-party computation with a chain itself and so we just do it with another server which we call a notary and then the server like sticks a signature on this and then the chain just does a signature check and like this is fine but like obviously weird for for you know the the the obvious reason which is that you're trusting the notary and so we see interesting things that people do kind of from this jumping off point which is like for example reclaim protocol on their white paper talks about how they can create a network of different notaries with economic incentives to make sure they're honest there's people doing things like running TLS notary on like layer AVS is there's a bunch of different things you can do to make this better like maybe the dumbest one is just checking with multiple notaries to make sure you're okay um anyway so we have this Tech and it's running and what this means is we can take any data that's on the internet and kind of efficiently cleanly Bring It On chain we don't need to go through an oracle process we can just be like hello blockchain I have a valid response from this web page and now it's a fixed Point people can collaborate around and it's somewhere we can move money around based off of um a brief degression on this point is uh we built a product based off of TLS notary that I'm going to talk about for like a minute and then move on just to give you a little taste of the stuff you can do um the product is called tomorrow. news what tomorrow. news is is it's a prediction market for tomorrow's New York Times headline and unlike the most prediction markets where you have to choose either A or B like kamla or Trump or whatever um in this prediction Market you can type in any sentence any natural language sentence and the market will result based on how close your guesses to the truth specifically in this case based on how close your guesses to tomorrow's New York Times headline now how do we pull this off um we pull it off with with two different tricks one trick is we use sentence embeddings sentence edings are a trick from AI where like you put in a sentence you get a vector and you put in another sentence you get another Vector if the vectors are close to each other the sentences are close in meeting like I like cats and cats are good of a of a short Vector distance and like I like cats and you know the Moon is blue of a much further Vector distance and so we can use this to resolve the prediction Market um but how do we bring that result and the result of the New York Times onchain is we use TLS notary so specifically when I make a prediction I put an embedding onchain and when I want to resolve the prediction um anybody can resolve the market to resolve the market they provide two proofs one is a proof that this sentence is tomorrow's valid New York Times headline that's a TLS notary proof and the second um proof is that this this is a valid response from open Ai and vetting endpoint and the nice little hack here is like you can use zkl to get uh inference on chain but like TLS notary provides a nice little hack to get the output of an ml model on chain with a different trust model which is just like check the open AI endpoint um and so we actually put this together and launched it you can play right now at tomorrow. newws if you want to see an example of that but that's like one sort of thing I'm very very excited about and I think is totally underexplored is you can take any data from internet where there's obviously a ton of meaningful data and move money around based off of it and even compose on chain the same way you do in like defi based off of it this seems to me like a very ripe area for exploration okay breath spark number two has anybody heard of appest cool all right couple of people so what appest is appest is a protocol from Apple that's used used to prevent fraud in mobile apps um all it is is it's a way for a server when it gets a request from a client to know two things one that this request is from a valid uh Apple piece of Hardware like an iPhone or an Apple Watch or or whatever and two that the request it got is from an app running an unmodified version an untampered version of a certain app binary and so like the iPhone hasn't been jailbroke or anything and so developers use this to like give out premium content you know things that are gated by payments they like use appest to to limit fraud um in in those capacities let's dive into how appat test works again I'm being kind of reductive to to get the point across but how appest works is your app generates a key pair your app then engages in this like fun Black Box Apple doesn't tell you how it works like back and forth protocol with apple uh in order to convince Apple that it is indeed one a real iPhone or Apple device and to an unmodified version of your app's binary then your app can generate attestation objects which look something like this um that your server can then verify to make sure that this to to say that this app this request came from a real app running this binary and like the important part of this adastation object is like this x5c it's just like a certificate chain and um some friends actually recently figured out a way to to verify the certificate chain in sp1 which means we can now make CK proof proofs on it which means we can now verify those proofs on chain and so this is really really interesting because now we can verify onchain that an iPhone a real iPhone running a specific app said something and I I feel like this's just a huge opening of gates because there's like there's like a lot of stuff on an iPhone um there's like location sensors there's like audio sensors there's like local ml models there's like you know Health Data there's your contact list there's all this insane stuff on a phone that you can now like really cleanly verify on chain um and some fun things you can do with this or like you can make an app that airdrops tokens like based on how far somebody's phone drops um and you can also do useful stuff like I feel like this like kind of opens the gates for you know there's still a lot of problems to solve here but I feel like this is like a meaningful step in opening the gates for decentralized versions of like Airbnb and Uber which can verify location data you know some sort of way to generate maps in a peer-to-peer way like social consumer apps that bootstrap off of contact lists that are onchain and like onchain versions of games like Pokémon go like I think this sort of stuff paired with like the fact that right now we're at an insane place in terms of like consumer onboarding to blockchains means we can build like really new and wild and creative experiences that are fully on chain Leverage The Chain in interesting ways all right deep breath spark number three um spark number three needs another little little degression like the one we started off at the beginning uh here's a a black line that is the the kind of like bottom layer of stuff on the internet yeah so like here are all like the services we commonly use on the internet that like hold a lot of interesting data like Twitter you know whatever you get it like there's a couple of nodes on the internet on the graph of the internet they're like very important a whole lot of of like very meaningful interesting stuff um there's some businesses built on top of those nodes like honestly not very many businesses I had a pretty hard time thinking about them and finding them and it makes sense why uh like for example like tweet bot like you know was built on top of Twitter I don't know if it exists anymore because of the API um to help you write tweets like better Twitter clients like games were built on top of Facebook Etc but it makes sense why there aren't that many businesses like this and that's because like there's kind of risk in building a business like that right like you're just totally dependent on the provider um and on top of that you're dependent on the provider having good apis and open apis Etc that are actually useful and you literally see no businesses like even one layer above that because like not only is the risk multiplicative but it's very hard to understand data you get one layer out from the bottom layer of the internet in fact to like verify and build something real off of data that you get one layer out from the top layer of the internet like if I get something from tweet Bots API response I also need to like connect to Twitter to make sure the original data is valid in a lot of cases and so not only is like the risk multiplicative um because you have like risks at two levels of apis uh but the data is like very like gets more and more annoying to deal with as you go up the layers because you need to deal with everybody below you some friends at Xerox Park have been working at a solution to this um called pods and like I'm I think there's like many more better talks happening at Devcon about this so like definitely go to those but the one thing in here that's I'm excited about and that is relevant to this talk is the fact that pods provide really easy ways to create pieces of data that not only like have verifiable claims in them but can then be computed on to create another verifiable claim that can be verified in itself without checking the everything checking everything that came before it it's like recursive in that sense like I just said a bunch of words like what does that mean it means like I can get a pod that is like a verification an attestation to like my age let's say like my age is like over 20 and then like somebody else does some compute on that and it's like all right my age is over 18 based on this compute that I did I can look at that give that my age is over 18 with the proof that it comes with it I can give that data to somebody else and they can be convinced of its validity all the way back to the root after just verifying one succinct proof and like that is really really interesting because what it does is it like enables you to build build higher on top of this this Tower and like turn this like kind of like very low like we have very small buildings on the internet it lets us build like really tall towers of businesses on the internet again this is like very Vibes based I don't even fully know what's like crazy and interesting here but it does feel that like another dimension of of building stuff is open um and this also is like very related to the other things I talked about in the talk because many of these proofs are easily verifiable on chain um and so you can kind of see all of these things coming together anyways those are the three Sparks and to summarize like again reductively of like my own talk um blockchains are a very complimentary type of computer to the internet like they're annoying to deal with but they're they're very composable because we're all building on the same computer they create these like fixed points that we trust and downstream of that we can move money around based off of them like compare this to the internet which is like this very like Emeral like you know fastf flowing nonfixed Point thing that's like very very useful for like quickly building stuff and iterating and it's super cheap right like very different types of computers and like currently the apps we've been seeing built that are blockchain apps like are kind of like this they exist within the silos of their chains and there's like very limited but very very interesting points were which they touch the world um all the point the one thing I want to leave you with is the next time you're dreaming up what to build I think we've explored a lot of apps like this like I'm very interested in what apps that feel like this look like and I actually do think we have the tech right now to start messing with stuff in that domain um yeah so that's the talk I'm very excited about this I'm very about building stuff like this uh the best way to contact me is to DM me on Twitter or like find me after um yeah 30 seconds left perfect so um I will I will start taking questions thank you all so if you have questions scan the QR code you can ask them you got a few minutes to get them in uh but we'll go ahead and start with the top one um in earlier slide you mentioned uh the headline prediction Market um it appears to be a blank page right now oh oh yeah I'm stupid I put the link up it's tm. news you can go play right now the headline prediction Market uh I think was like a great example of the stuff I was talking about but I think was like specifically a bad product because um it it like it turns out that the New York Times headlines kind of a category like it's like one of four things it's like you know the election like Israel like a hurricane like there's just a couple of things that it is and it's like really hard to ask users to like write a complete sentence um and so I think like the the right path would be like finding somewhere where an expressive prediction Market actually creates a ton more value but you can literally go to tm. newws right now and like type in a prediction for tomorrow's New York Times headline and the market will resolve tomorrow and uh you will do well based on how close you are in meaning to tomorrow's headline um yeah cool so what would it take to actually put an embedding model on chain so the trust rout will be native to the chain sounds like proof we can extract with optimized ZK um I I don't know exactly what the trust route will be native to the chain means I I think there are two ways right now like I would practically put an edting model on chain way one is the one I did which is you TLS notary proof uh the open AI API and they'll call a ZK TLS what the ZK is is like after you get the TLs notary proof it just like a ton of like for open AI it's a bunch of Json from their response and then you like use DK to say like here's the actual embedding that I extracted from it um actually we don't even use ZK for that one we just par the Json on chain but like usually you get like a massive HTML from from TLS not and you need to like do some nice DK to like run a Rex over it and make it useful on chain um so that's the first way I would do it the second way I would do it that's like maybe as a better trust model is like embeddings models are like actually pretty small um and so like Ezekiel or like some ZK ml solution could just generate a ZK proof of valid execution of of a forward pass on the model and you could verify that proof on chain as well um yeah but both are are valid and I think just have different like trust models awesome next question based on all the advances you described in the talk what do you want to build next um I'm actually building a talking rubber duck next that has little to do with this but um one of the things I do want to build next like like I'm very interested in like stupid fun stuff that like makes me happy you know and so the the one idea that I threw in there that like didn't seem that cool that like I I I then like was like we can do useful stuff with like the phone drop thing like I actually love in want to build um there's a bunch of like super interesting ZK appest things that you can do like uh um you know you can build like location based onchain games like there was this missile game where like people would shoot missiles at each other and then like they would literally have to get out of the physical location to not get hit by the missile there's that was like an iPhone game from a long time ago I think it'd be fun to do like an onch and incentivized version of that there's like an Infinity of of notebook of ideas I have with this stuff that I'm happy to talk about but I'll cut myself off there um yeah cool web or mobile like both I don't know it depends what you're making okay so was it better to do polls rather than writing a sentence for TMR news yeah this I think this was the initial point I made is like that that's a a for this specific thing it's like maybe better product to do categories I think like the that's not to say there isn't a place where it's literally better to have an expressive prediction Market where you can type in anything um and that's really really not to say that there isn't like a blue ocean of stuff when you can bring meaningful internet data on chain and then have money move around based off of it like I think I think that's the real point I want to make in this talk okay probably the last question yeah how do you feel about te trusted excution environments as a stepping stone to the zk's another tech for composing real world with a blockchain um I how do I feel about tees um I don't know I don't feel much about tees I have thoughts about them I mean I think like they look I think like sgx gets broken a lot I think they're like very useful and you can do interesting stuff with them like a lot of the stuff that flash spots is doing with like writing is very very interesting right like you can like give up your account people can write based off of your internet accounts because the te owns it I think that's really really interesting um yeah they're like really good for adding private information to onchain games in certain interesting ways um yeah cool a lot of stuff getting the time up signal and hear the music so catch me after thank you you all