Establishing Token Engineering at your Startup

Transcript

[Music] okay we are good to go Fox it's so cool to see you all chatting because there's such a lot to discuss uh we'll have two final sessions and they are really relevant who for everyone who want to make token engineering a practice in a crypto startup and I invited two guests I'm very happy that Lucas shimanski and Tom Allen are here thanks for coming have a seat and we'll discuss establishing tokenengineering at your startup I'm glad that you are here because you

are both token engineers lukashu founded your own company it's a service firm for token Engineering Services and Tom you're working at crypto econ lab as a research scientist so meaning we can share different perspectives probably and I guess we should start with a brief introduction uh your background what brought you actually to token engineering yeah hello everyone so my name is token engineering agency let's go like that we are focusing on modeling and simulating a different token based system crypto economics and my background before before running

this company and my background is software engineering and then managing software engineering teams so before I started economic Pro I was director of engineering in olix group which where I was running uh the teams of hundreds of software engineers in Europe I was also working a little bit at Spotify so I have this experience with startups and also big companies that are around not only software but also data science and all of these different processes which are important in token engineering context and what brought me to the space the possibility to basically create novel economies and deploy them at scale and try them out which is not possible

at let's say in real world nice thanks Tom sure so um as Angela said my name is Tom Ellen I work in crypto econ lab and I'm which is a part of protocol Labs um so how did I get here I had kind of a security this route really I originally trained in computational physics and did my PhD and worked for some some years in that field and then gradually moved more into a more data science machine learning slash kind of large-scale incentives but not in not in the crypto blockchain world but more infectious disease epidemiology and worked in that for a few years and then eventually it ended up switching to blockchain and why

really just because there's like a huge sense of optimism and possibility to combine effectively a new way of computing with incentives and yeah it's just that's a that's a very powerful idea so I've been I've been working in this field for a couple of years and I run a token engineering slash crypto economics mechanism design protocol design call it what you want a team in crypto econ lab and we primarily focus on the foil coin core protocol but we also do work outside of that and advisory and Consulting too nice maybe you can both give us a picture of a how a typical project looks like in terms of the steps the duration

a typical outcome a typical starting point yeah sure so for me and it might be a bit different for Lucas uh it really depends what the type of project is whether it's an uh an internal filecoin project or it's an external advisory project if it's an external advisory one the kind of steps are there's some like sourcing and meeting the clients like how do you do that when you talk to people places like this and then after that there's a process of onboarding clients so there's part of this is due diligence from our our perspective to see are we aligned in terms of values and where we want the projects to go uh scoping out the project

[Music] talking back and forth with the clients is kind of continuous process of iteration and coming up with a pricing proposal and if everybody's finally happy with their milestones and timeline timelines that you move forward with that and a project could be scoped uh anything from like a month up to six months or you could have a more kind of permanent retainer structure uh or you're working kind of long term with a project like file coins core protocol which is what I focus most of my work on and like how do projects arise crypto economic projects arise for example in falcoin core protocol hmm uh many ways but I mean you're kind of really embedded in the ecosystem so uh I

mean you're kind of you have this process of continuously auditing the protocol to say hey what could be improved are there any uh you know potential approaching icebergs any potential bugs that we've got to think of but you're also really kind of embedded within the ecosystem of the community so ideas will bubble up and come to you for example in follicle in the storage providers might say hey like we really want to onboard more storage but this is our bottleneck uh is there some incentive design structure that can help us like alleviate that bolnik and so that's like a kind of typical way that a project comes up in Falcon okay it makes sense so yeah sorry the question about the process and can sound like simple but

actually it's really challenging to come up with the proper process especially when we are doing simulations because I'm always saying that you can model and simulate basically it with everything that you want but then you have two constraints main constraints which is time and budget right so of course that I'm I'm talking from this perspective of the person who is working with external clients of course that client would like to always have an answer to all of the questions but then we don't have probably enough time and enough budget to do all of that that's why I will process a consists like kind of the the foundational process which takes around three to four months I would say

depending on the how the big project is is that the first phase is um kind of phase of system requirements which we are using in order to not only get the requirement requirements from the client because in most of the cases clients don't really know what they want and we need to help them to figure this out so you can treat it as a more like Discovery face system requirements face but the most important thing is that at the end of the day the deliverable is what we are calling model design document which is this first Milestone where we are saying okay so this is how we end up this first step and this deliverable is our input for the next

step which is creating a model in our case it's it's got cut right now but we were chatting before about the bottleneck of cardcards not being ready for a high performance at running as we needed it right now so we are hacking it a little bit but anyway then the second step is modeling where we are just basically providing modern verifying does it work with the assumptions and the third stage is simulations we're running simulations that their main goal is to answer the questions that were in the model design document at the first place so this is the foundation and additional comment is that you always

need to be ready for the fact that you will switch between these phases so you always need to assume that you can go from the third step back to the second one or even to the first one and then of course that client project wants to take these simulations further so answer more of what if questions and then what we are doing we are just basically estimating it again we already have all of the tools that we can run on top of it so we are estimating it kind is accepting this stuff and we are running it again just briefly all right I wonder um what are the typical challenges in in scoping a project you've mentioned sometimes you have to help the projects to really find out what they want or

need what are the typical challenges in this setting defining a scope to to define the scope so it kind of depends from what perspective you're coming towards uh in in terms of of scope uh so I mean if you come from the perspective of like running a team of several engineers and research scientists uh I mean I I tend to bias more towards or or like hiring people who have a lot of potential who tend to be on the younger side but then like a pro something you often see is they uh like greatly um uh underestimate the scope of a project and how quickly it is to do so that's like

one thing to watch out for in terms of scope be careful for add your own team to know it yeah yeah I mean so I mean there's like different perspectives of like thinking about like how do you run token engineering or startup this is like more from the perspective of like running the team uh rather than the perspective of hiring the team which is another perspective um yeah I've had this own experience you're always feel too slow um you mentioned cat cat and uh for those of you who don't know cat cat is a cat cat is an open source tool it's actually a python Library open sourced for the in the token engineering community to run

simulations now it's one of the tools and one observation during ECC from myself was that more and more teams are start to build and come up with tools that they build internally for their teams or that they provide for clients what are the most interesting developments here yeah so that's kind of interesting that's like well Lucas and I were just speaking about before we came on so we kind of a slightly different approaches um the approach that we've taken for at least through the file growing work and and partially this because you know this is like a long-term engagement so you can spend a really long time building custom software and really building everything from the scratch grinder so I

mean the kind of typical stock we use is the standard python data science stake but then we build up our simulations on top of that so we generally build our simulations using junks which is a accelerated linear algebra library for Python and this is really good so this is allowed us to build models that are very easily vectorizable parallelizable but also differentiable so this just like opens up a whole new world of being able to do experiments so the way we kind of approach it is we have a uh for I mean like let's take for example the supply simulations we have a kind of basic skeleton model for supply simulations and we can run this a few different ways you can kind of take a

signal processing approach where you have some signal driving it and then this feeds into the rest of the model and you get some output or you can take an agent-based approach where you have some subpopulations of agents and they're making some kind of rational decisions and this is what's driving driving the network or you can take another kind of approach which is like the the one I which is where you have some say unknown parameters or some like structure in the model that you want to test and then you can optimize this or you can like take like an mcmc approach and even like inferior distribution for a parameter so that's kind of how we tackle the simulation side actually already pretty

broad right yeah yeah we got like a good arsenal of weapons I think to tackle you know the economy but how does that how is that different or similar to you yeah so besides like standard python data science stack and yet this needs running something more sophisticated than like Jupiter notebook because also like most of the materials that you can see available on the Internet or tutorials whatsoever are on top of Jupiter notebooks right which I mean it's fine of course it's the industry standards in whatever industry like in general data science standard but then when you have this need of

really running a far much bigger model or more computational heavy simulations um but then like for example in our case what we are doing we are we are using jupyter notebooks maybe just only at the beginning just to check something out like like initial verification but then what we are doing we are using our own let's say technical infrastructure that we have where we are running our simulation on top of the on top of the cloud where yeah we were discussing the the performance issues with that that even yesterday night I was doing uh the performance test and comparison about running cut cut with different instance

types and so on and so on so there is like it it is enough to some degree but the ones you really want to put it at scale and make sure that your simulations are running fast like you have different ways so whether you are building your own tool which is basically from the get-go design to be as parallel and as let's say computation performance as possible or but then again it probably is the case where you have just you know you are working inside some client in the long term but then for net for company like ours and we are thinking more about contributing to cardcard where we will be able for example to to contribute in a way where it will be able to

address our needs that are not addressed right now so in terms of of stock and tooling this is what we do and yeah mainly also I would say in terms of doing something that Transit today all of this I do believe that there will be more and more tooling about around the verification and that's why also we are building our own standardized tooling for verification on on top of the cut card and also models on top of the cut cut which I think at some point in the industry we will finally reach some probably agreement how to do that and then we can just you know easily exchange between each other

with the models that we created which is currently not possible Right it's possible but it's not happening basically there is no I even seen that recently cardcard announced some changes some I think some major changes in the framework itself which is not helping uh because again you know for me it's uh if you are like really looking like down to earth for me it's again the cost just basically of time of people that needed to align to that while we already you know prepare all of the stuff on top of the actual version and and then it's not also helping in developing the the standardized way of how to approach cut

simulations because again there is some major change that can break things that were in the in the past so it's possible but not happening I do believe that this event is a great start to actually start to talk about it maybe who knows in future you will organize the conference only about this stuff where we can deep dive into more like technical topics and bring some consensus to the space what's your point of view here yeah that would be really cool I mean I'm just like drawing in on the last point you made where we would be good uh also like dive more deeply into like the differences in the technical approaches that we use uh I mean I was kind of curious like do you find this to be a big limitation uh or like a bottleneck

for your work the fact that it's quite computationally intensive when it comes to work with clients no not really and I mean we can afford ourselves to you know run a simulation that can take up up to one hour a simulation that consists of hundreds of Monte Carlo runs and hundreds of parameters and this kind of stuff which we already optimized because normally it took like if you would like to run such simulation on your local machine then like you can count something around one day probably up to one day uh if you will put this in the in a Docker then forget about it because probably your memory will be not enough

at least this is what we and this will kill the process and this will be after the simulation so for the clients it's not a big problem if you're running it in an optimized way especially if you have infrastructure for that which is an important point from like people considering perspective of should I build team internally or should I hire external team the one thing is first of all it's a team so it's not one person so the one thing is a team that conceptually can like do this stuff can have this knowledge but then you need to provide infrastructure so they can basically operate around it which then performance is also the case but for our other needs where we are running basically

tokenomics audits at scale and we are simulating like hundreds of projects and you want to have iterate on this model and you want to have results as soon as possible then it's an issue yeah yeah I mean it is it's really for being able to iterate on the scale in terms of like which we think and like work if you can run models on a scale of like 15 minutes and keep it reading on ideas that's really powerful uh like before whenever I worked in computational physics we'd use high performance Computing thousands of course and run a job for a week and it's just so painful um we don't want to be going on that scenario although I suspect there's always like something in us that's drawn towards that it's you know it's like the the kind of principle or like whatever it is the first law of high performance

Computing that whatever Computing resources you have you're like filled to expand it so sure we'll like make our models like really efficient and paralyzed but then we just want to test more parameters and more questions right yeah you Rohan go no no no I didn't spend time yet but if there are some like better performance solutions then I would be happy to check this out okay my Approach was like okay cut cut I

think that the version It's 0.4 something now which is it didn't yet establish the let's say first version when everyone can say okay it basically Works uh so then I was like I'm not going to jump into another tool that is kind of the the fork of it because then like the development probably is even more in Alpha stage so you know for me I can't afford this kind of let's say bugs that are coming from the tools and not from the human mistakes we mentioned teams so not only tools are relevant also the people who are part of a token engineering team so

to that it's not only a single person um let's jump on it would you recommend any crypto project to hire a token engineering team to build internally uh from scratch or higher external yeah okay of course I'm biased we also were talking about it um but to be like totally objective I do believe that okay if there is like for example project like file coin which I don't remember but I assume that go research driven then probably like the only way to go is to build the team from scratch because you already have people that are knowledgeable about the space but for let's say most of the cases I

would I would say this is an exception to the rule where the rule for me is that most of the places is occupied by you know startups or mid-sized project projects where they still need to validate their own proposition and they need to find their micro Market fit and they need to do it fast so then that the most let's say objective or standardized way how we are doing with the clients I think it's even similar in the software engineering world is we are starting the whole thing in the client situation which is we are bootstrapping and everything around token token engineering so teams tooling processes and so on and then after we had the

first results we are starting to kind of handing off the whole stuff after the project verified they Market fit and they have for example money to expand more so of course they will think about hey let's build team internally and Lucas help us to transfer the knowledge and to make sure that this will work inside so this is I think the the best model for startups we mentioned it a couple of times crypto econ lab has been grown in-house for firecoin at protocol Labs how was this journey yeah for sure for sure um so in crypto Lab at the moment there

are nine research scientists and we built that up over the past two years um and if you're going to have a full-time uh in our team of research scientists and Engineers yeah for sure you need like uh more funding for that to sustain it long term it's also the uh you know non-trivial to set up in the sense that hiring people is not easy uh if you get a false positive that's a big cost um and it's it's you know it is difficult to hire uh good people I mean there's a lot of talents out there but not in the in in the same way as in other fields where people have been doing it for 40 years so there's like quite High variance so going through that whole process of hiring

um uh engineers and research scientists the same way as you would do in any kind of tech businesses is like a slow process uh there's like many stages really so that's one thing but I mean we've got ourselves into a grid position now we've got a very strong team with a whole range of different skills so what we try to do is like set it up were such that there's a real sense to kind of complementarity like some people have a background in machine learning and control theory I mean if I just like literally go through the people in the team and and this guy has like he's an excellent research software engineer as well we've got other people who are more into Game

Theory and governance which is like a super important part it's like fine to say okay here's your tokonomic design here's the research this is what you should do but if it's before an established ecosystem then you need to be able to pass this through governance somehow so that's really important and then we've got other people who work more on the theoretical side so more pen and paper analysis of incentives we got some PhD stats people that's very useful and like yeah more machine learning and so like some of the people we've drawn from come from Academia but some are much younger like 20 223 and they're

just like so smart and excellent domain knowledge and you know really on an upper trajectory so like having a whole range of skills and abilities that's very important um but it's also not sufficient I would say just to draw on people who have these very strong technical skills you also need people who have a bit more of a business sense and can run in a startup culture and like not get too deep into the weeds but step back and see the bigger picture and really keep things moving forward that's really important yeah and how relevant are I mean we think crypto econ lab since minimum a year invests in sharing your work your publishing articles about your work your even open sourcing models

you are visible with your work how relevant is it in terms of finding talents yeah that's true uh I mean I was kind of pleased to see some crypto wecon lab work uh in the previous uh slide um we do try to do everything open source so we open source our models generally and we open source all of our all of our research in general and this has been good I mean this is how we've attracted some people they said oh hey we're reading your article and they come up and talk to us at a conference and then they end up working in the team so that's really good but it's also useful um from the perspective of working with external partners because they can see the kind of research that you're doing it's all up online and you can say hey well we can give you similar insights

for your project so that's useful about you how do you find your talents yeah so maybe just say I will add the comment on an open sourcing and showing also researchers and so on so in our case it's more challenging because you can imagine that when we are working with external clients you are working under NDA and most likely when you are doing some audits you are funding some stuff that clients don't want to share with anyone obviously um or then even if you provided some interesting novel model and so on like all of the IP is going to the client and obviously client paid for that and like unfortunately most of them don't

want to open source it and share it but I'm working on that so we will able to see recently that we started to share more and more of behind the scenes of our work because when I'm playing this string of hey it will be also good for your marketing like it's a win-win then then it started to to work well in terms of hiring Thailand's um yeah I think the obvious answer will be token engineering Academy that's for sure this is where I mean this is you know if if you are looking for I I have like plenty of years of experience in looking for people because I was also building these software engineering teams from scratch and I know the struggle of hiring and especially like investing a lot of time with but trying

to understand that our people that are reaching out to you are like proper fit for for your company and for the position for this case that you require uh I don't have this problem with token engineering Academy because I already know that there are people there that are interested in this field which is already a lot for me at least if I was always fun of a model that high lens let's bring on board people with huge potential and huge interest to the field and then we can like teach them the rest of the stuff it's it's that easy so we are actively using token engineering Academy we are running our local study group and we are using it in a both ways

so one is looking for new talents and the other one is onboarding new people because when we are hiring somebody that is from different source which we we do have case it was referral from our employee then he is joining our token engineering study group to onboard himself on crypto World more because the background is mathematics statistics and so on so yeah that's just kind of when it comes to filling this gap of the of the crypto knowledge so that's for now that's my main source I'm not looking for anything else because it's enough for me and someone told me that the token engineering crowd is the least crypto enthusiastic crowd in crypto I I wonder how important is it do you

Source people or find people who are crypto natives or how relevant is it that you have a good understanding of crypto smart contracts blockchains yeah that's a good question I never heard that phrase like that before um so I mean I I think that am I mean one issue with hiring people for token engineering for crypto economics mechanism designers is so multifaceted uh you can hire somebody that's an expert in cryptography but that's that's not enough you need somebody who knows how to build models and simulate rational agents you need somebody who knows about policy uh there's just like so many different

strings to it so I mean it kind of depends I mean what are you thinking whenever you say uh like uh the crypto economists or token Engineers are less interested in or in uh like crypto native like how are you thinking about that yeah the thing is probably that we need a lot of skills that are not crypto native that are stemming from other engineering disciplines computer science game theory and then we we see a TA Academy that the potential of token economies gets them excited um for the sake of of this idea not necessarily for being practically

involved and that sometimes causes issues um because you have to understand the end the technical environment you're designing for yeah that's very true so I mean this is something we like come across whenever we're working with other core devs for example at Falcon um like different people understand the system different ways some people take a very like top level approach and they come to it with some like principles like hey this is the behavior we want to see this is our objective how do we go about designing an incentive to get there other people will be much more into the nuts and the bolts sort of like the smart contracts and the actors saying okay this this piece of code this line interacts with this like how can I change it to you know improve the the outcome it's always like two different

ways of understanding it I don't think either is right I think they're like complementary and really unique both it's like very rare to get like a single person who can do absolutely everything so yeah it's like collaboration is the name of the game makes sense one additional comment maybe but just so people don't get this perception of uh you need to be like highly Technical and so on because actually what I do believe and we've talked about it also earlier is that the the skill that we are looking for the most is communication so how much you are able to translate this whole complexity that we are talking about to a potential client or the receiver of your

simulations report and so on I had plenty of discussions with token Engineers that they were like yeah I can do modeling I can do simulations that's not a problem when I was asking okay can you do the simulation report and present it to the customer no I don't want to so and this is like huge issue and when there are people that have let's say they're kind of t-shaped people where they have this good enough understanding of what we do here basically where is the value for for the receiver of this information and how we are doing that and have this strong ability to try translate it to let's say human language at the proper level this is our super skill okay

final question now if anyone in the audience a crypto project thinks about okay we need to hire a crypto you can lab we need to hire tokenomia Pro what should they have prepared to talk to you not much really because at the end of the day this is the goal of the discovery phase that I mentioned so a clients reaching out to us with different requests at different stage so you can either have only just high level idea that enough we will just probably have more time into the discovery phase to figure out what we can do you can also be at the stage where even everything is live and right now we are

trying to figure out okay what's next which is also where we are helping so just reach out and we will discuss together what we can do do you have any framework checklist yeah so I mean we have a framework and a checklist for onboarding external clients I mean I would say the most important thing is just to be able to clearly articulate what is your idea of utility and what it is you're trying to maximize uh with the project and from that we can take it from there and help you design the crypto economic incentives that can achieve this okay thanks we have some more time for questions in case you guys have any

like this might be a funny question out of all the tokens out there which one is the favorite design that you've seen so far I mean naturally I'm going to say I think it's a very good design I think there's a lot of uh complexity and sophistication and things to keep us working on and to improve and it is really a protocol that has to keep evolving because because of what because of what the utility is it's growing storage and in many ways uh this is like a decentralized version of a storage company so it's like a complex thing to to work on and design the incentives for but I think it has been good very good so far and I I think there's uh yeah more work to keep evolving that I mean

what's your perception of your favorite token engineering project Lucas I mean that's uh that's a question question about what's the best let's say design token and so on it's uh it's question without the answer because it depends what was the ultimate the function of this token which I think the good example of that was today in the presentation about the uni swap like what's their main formula for what they want to optimize for so first we will need to know what each specific project we're optimizing for and then this will for us lay the foundation of our let's say evaluation model of okay this is a good design because like of

course by default everyone is looking at token prices right and this probably in most of the cases will be not the best idea how to evaluate the project because the token price was one not the most important thing of course it has impact but maybe not at this stage is so important the second one like the multiple ways of how the projects market makers whoever on the market are able to manipulate the price it's endless basically but so then it's not really also a good metric to objectively evaluate like yes this is this is a good token token design I have my couple of favorites but not Financial advice

and I will share it maybe later because yeah it's live streamed I'd also like to point you guys to the first talk today where Trent talked about the evolution of ocean and I recommend to watch this video again because to me it reflects very nicely that there is no final token design and that there you have to adapt Your Design to the evolution of your total protocol of your project and sometimes you might run into problems or in to a dead end and then you have to be really Brave to leave your I mean he talked about throwing away 90 of their code this is

something and you have to deal with this with a token that is already life manage these changes and manage this Evolution and I think it's a really great example to get into this thinking of it's it's an ongoing progressing thing and there is nothing like a finite Perfect Design um you're interacting with the markets around you tokens are composable so we have some really cool challenges to solve and yeah take a look at this presentation I will use this analogy of the tools that we are discussing like okay so what is the best tool for modeling tokens I mean it depends right

like what does it mean the best is it the best in terms of performance is the best in terms of how it's easy to use is it the best for you because of I don't know it's it's the cheapest one so I think the the the the question itself will need to be let's say more specified on top of which we are evaluating the the design model but it's a good question and Trigger for the really good discussions thank you any other questions final one how to establish token engineering at your project shoot it yeah I don't mean to monopolize this I

apologize but a quick question you mentioned that you need to have a continuous design iterations for youth ergonomics how do you achieve that while ensuring decentralization and governance uh yeah sure um so I mean you're heavily reliant on the governance process it's it's there's always a complex trade-offs so trade-offs in any ecosystem where some people may perceive that they're winners or losers from a given change