Finding our Identity Block by Block
A few weeks ago, I remember watching a video by DJ Trev about building the foundation of a house. The foundation of a house is the most important element of the entire structure. It serves as the base upon which everything else is built, and provides stability and support for the entire building. Without a solid foundation, the entire structure is at risk of collapsing or becoming unstable.
The problem that we seem to be facing is that we are trying to build this foundation and find our identity on sinking sand. In the past couple weeks, every day seems to bring new challenges and obstacles to the ecosystem that pulls the ground out from under us. Our proximity to regulatory enforcement actions and collapsing financial institutions are earthquake-like events that we are having to endure.
It is during these times - when we are shaken to the core - I’ve been forced to reflect on who we are, what is our value proposition, and where we aspire to go. I’m reminded of Alex Forshaw’s January article about LUNC’s product-market fit. UST used to be Terra’s defining product, but faces massive financial and regulatory headwinds. Do we go down that path again? Over the past several months, I have been thinking about my role in all of this and would like to share my struggle embracing my/our identity and how to innovate in light of this.
The Cosmos Identity — Internet of Blockchains or Fat Protocol
At the macro level (Cosmos in general), I think we are struggling with an issue of identity. Cosmos is a decentralized network of independent, parallel blockchains, whose main goal was to create an ecosystem where different blockchains can communicate with each other efficiently, known as the “Internet of Blockchains.” It focuses on application-specific blockchains (also known as “app chains”) and their integration with the Cosmos and Tendermint ecosystem.
The idea behind app chains is to create specialized blockchains tailored to particular use cases or applications, rather than having one-size-fits-all blockchains like Ethereum. This allows for better scalability, security, and customization, as each app chain can be optimized for its specific purpose. Cosmos facilitates the creation of these app chains by providing tools like Inter-Blockchain Communication (IBC) protocol that enables communication and transfer of assets (and accounts!) between these app chains, enhancing the utility of the entire ecosystem.
However, thus far, the “fat protocol” thesis has played a dominant role. The fat protocol thesis envisions a single, comprehensive blockchain that supports all kinds of applications and use cases that drives demand for the native token and protocol. This idea is similar to early days of the internet when some believed that a single supercomputer was the best approach and everyone else would just open a remote terminal on that central machine. In fact, many of the blockchains in the Cosmos aspire to do similar things, and thus, there doesn’t seem to be much of a clear differentiator at the protocol level. The call for utility on LUNC similarly seems to bolster the drive for the fat protocol, where Terra Classic can support many different L2 applications on top of the L1 layer. I do not want to downplay the importance of this. Getting the chain upgraded with the latest cosmos and wasm support is absolutely critical to LUNC’s future; however, given that the concept of application-specific blockchains is a core principle in the Cosmos ecosystem, exploration of alternative approaches to the fat protocol are warranted.
The LUNC identity — Meme coin or Utility chain
Pre-depeg Luna was a top L1 utility chain with a bustling ecosystem; clearly Terra was a fat protocol with a solid user and developer base. However, post depeg, everything changed. LUNC’s utility dropped to near zero, and the meme coin emerged. The “burn” narrative drove investment early on at the chagrin of utility maxis, and continues to cause fundamental conflicts in governance. This ping ponging of burns, taxes, staking, and funding are a result of LUNC’s split personality, which probably will continue on for the foreseeable future.
While this conflict can become heated, and sometimes brings out the worst in us, there is also an opportunity if both sides can accept our identity and use each other’s strengths. Bringing utility to the chain requires developers and development teams to build here. This requires investment in the infrastructure, dapps, and L1 layer. This also requires incentivization for teams and projects to build here. This thesis requires funding that seems to be in conflict with the burning of potential funding. On the other hand, the burn drives increased visibility and interest, and promotes an attractive investor narrative. The meme-ness creates decentralized growth, with highly engaged community-driven initiatives. While there are a few builders on chain, given the current state of our ecosystem, it would be fair to say we are a meme coin that is working to evolve to include more use cases and utility.
My identity — Machine Learning Scientist or Blockchain Engineer
For the past 15 years, I have been researching computer vision — a subfield of artificial intelligence. I teach machine learning at my university, publish peer reviewed academic papers, and receive external funding to investigate open questions and build applications in ML. On paper, I am a machine learning scientist.
But for the past two years, I have been obsessively moonlighting in the world of blockchains. In the beginning, it was building GPU miners and gaining an understanding of code bases. After the collapse of UST, I became much more involved with Cosmos and Tendermint, deeply diving into the application level, network components, and consensus mechanisms. I’ve been working on the L1 team and contributed to five chain upgrades in the past year. When I introduce myself to people I meet, I often times throw in the fact that in addition to machine learning, I also work on blockchains.
So herein lies my identity crisis, where the first step is to be honest to myself and those around me. I am effectively a blockchain engineer on a “fat protocol”, meme coin. Some of you may disagree with this sentiment, which I respect, but this is what I believe at the moment. Let me emphasize that none of these are inherently “bad”, it is just not where I thought I would be today when I embarked on this journey with all of you almost a year ago. After introspection, I’ve also noticed my interest and excitement waning, and my intrinsic motivation decreasing. Personally, this is not where I want to be so — I have decided to take actions to change my identity.
Innovating on our Strengths
Over the past several weeks, the path that I personally need to take has become clear, and it plays directly into our strengths. With recent advancements in AI (especially generative AI like diffusion and ChatGPT), the utility of AI has exploded, but is quickly becoming computationally intractable. For example, OpenAI’s GPT-3 required over over $12M USD in training costs, and the carbon footprint of training this model was equivalent to the output of 126 danish homes for an entire year. This does not even cover the cost of obtaining and labeling massive amounts of data. At this point, there is no way that an individual or even small/medium sized organization can develop, verify, or even run state-of-the-art machine learning models.
On the other hand, there is no lack of compute power on the blockchain. The computing power used to run Bitcoin is around 40–50 exahashes per second. This is several orders of magnitude greater than the computing power of the world’s most powerful supercomputers combined. Although
the Bitcoin network utilizes specialized single purpose hardware which cannot be repurposed, the Ethereum network was the largest GPU decentralized network in the world and the computational power available on this network was undeniably massive in scale. But approximately 80% of Ethereum miners have gone offline following the Merge due to unprofitability when the hash rates overwhelmed other PoW chains. This compute still exists, but remains untapped. Can we somehow capture this network for AI?
Given my expertise in both machine learning (ML) and blockchain technology, I have gained unique insight into an ideal convergence of the two. I am excited to share a research project I have been working on, a Cosmos Proof-of-Stake machine learning application specific chain that decentralizes the power of AI. I have developed a custom machine learning Cosmos module that interacts directly with the Layer-1 infrastructure to tap into the compute power beyond the consensus layer. By retrofitting proof-of-stake validator nodes with high powered GPUs, we can utilize these nodes to provide access to state-of-the-art machine learning models for anyone interacting with the blockchain. I have been able to repurpose my Ethereum miners by upgrading the RAM and SSD disk space (the 1x PCI-E risers are fine). With interchain block communication (IBC), and interchain accounts, LUNC users will have access to a suite of the latest machine learning advancements. While I am still figuring out details, using the app chain thesis, I can develop and iterate at lightning speed. Below, I am demonstrating some of the capabilities already built into the app chain hosted by multiple testnet validators.
In the short term (next 3–6 months), I am planning on building and releasing a public beta. Long term, the goal is to have a blockchain that can be used to perform large scale distributed learning across many nodes. We can potentially deal with the intractable nature of some of these large industry monopolies surrounding large language models and foundational image models. We are also looking at privacy preserving or federated learning techniques and SNARK verification of the ML models. The tokenomics of the side chain will be especially important to consider for LUNC. How can we utilize the app chain to help fund LUNC development, help reduce the supply of LUNC and UST, increase utilization of LUNC, and incentivize miners to contribute their compute. I am in talks with seasoned experts to get this part right from the start.
In summary, I am flipping my script and forging a new identity for myself to align with a vision where I can have greatest impact. I am a machine learning scientist building app-chain utility for LUNC connected via IBC. My personal goal and contribution to this community has shifted to drive a recovery narrative that iterates on the identity and strength of Cosmos, Terra Classic, and my own personal experience.
Make no mistake, there is no guarantee of success and this is a risky pivot from the existing status quo. A lot of things need to go right for this to capture sufficient traction and network effects. But, if the stars align, I foresee LUNC could become a Chimera blockchain cryptocurrency that is a mix of fat/app, meme/utility, and ML/Cloud/Blockchain that has never existed before, with a massive utility value proposition. The nice thing about the app chain thesis, is that other community members can drive their own recovery narratives and help in their own way, i.e. USTC.
Other App Chains for USTC
While I am only debuting my ML idea, I have been in discussions with others in the community about how to rapidly innovate using a side chain connected to Terra Classic. We can quickly prototype applications or risky algorithms in a controlled sandbox and connect to Terra Classic via IBC. I have offered my assistance to help with the app chain technology and am also recommending that they consult legal counsel for execution. For this, we have already been searching for a good partner.
Given the ever-changing regulatory environment, especially given the proximity of Luna Classic to recent enforcement actions, we are proposing that it is now important that the Terra Classic DAO (decentralized autonomous organization) retain legal counsel to advise some of the core contributors and the wider community about best practices and basic corporate structuring to ensure that we can operate efficiently while maintaining our decentralized structure. Because there is no centralized authority, the legal representation of the DAO will need to be structured in a decentralized, democratic, and highly collaborative way.
As mentioned before, internal structures and teams have been set up to build the foundation, but with the recent developments in the regulatory environment, it has become increasingly important for Luna Classic to align its internal structures and proposed ideas to the external world. We believe that the community is increasingly realizing the need to limit potential legal risks and structure ourselves with a view toward long-term sustainability. This will also enable the community and protocol to execute on some of the proposed ideas for revitalization.
Next Steps
Here I outline some concrete next steps that I will be taking in the short term.
- Re-focusing all of my efforts to build out the ML Cosmos App Chain for beta release in the next three to six months. I will likely be asking the L1 team for help when needed.
- Help gain consensus on eXperimental team (“X” team) to work on USTC issues that can innovate quickly via additional App Chains. If this is the path they want to go, I will volunteer to assist with app chain level technology.
- Continuing funding L1 team in Q2 to upgrade Cosmwasm and IBC Interchain Accounts. I will volunteer to help when needed.
- Pass over TGF operations to Marco. TGF report on activities, spending, and future plans will be shared later this week.
- Gain consensus from the community on legal representation.
- Iterate and Innovate to find and deploy LUNCs value proposition.
I imagine now we can “move fast and break things” on app chains safely firewalled to Classic, and see where the future takes us.
Addendum — Interchain accounts
“Interchain accounts allow one blockchain to securely control an account on another blockchain, using IBC.” — Josh Lee, Co-builder of Keplr Wallet and Osmosis (https://medium.com/chainapsis/why-interchain-accounts-change-everything-for-cosmos-interoperability-59c19032bf11)
“Enabling composability in IBC allows innovation in distinct applications to be deployed without needing to upgrade the entire Interchain, providing a more scalable innovation model that preserves compatibility” — IBC Product Lead at Interchain Gmbh, Charly Fei (https://blog.cosmos.network/interchain-accounts-take-cosmos-interoperability-to-the-next-level-39c9a8aad4ad)
Interchain Accounts is a relatively new development in the Cosmos ecosystem that I think few people are aware of and deserves some explanation.
Inter-Blockchain Communication Protocol (IBC)
Inter-Blockchain Communication Protocol (IBC) is a protocol that allows different blockchain systems to communicate with each other. The most recognizable use case of IBC has been cross-chain asset transfers. In this case, IBC enables the secure transfer of assets between different blockchains, allowing users to access and manage their assets across multiple networks.
That being said, IBC is just a protocol and new functionality can be added to this protocol. One functionality that has recently been introduced is Interchain Accounts (ICA). ICA is a module built on top of IBC, specifically designed to enable cross-chain account management and interoperability. ICA allows an account on one chain to control an account on another chain. ICA streamlines cross-chain interactions, enabling developers to build more complex and feature-rich cross-chain applications.
In essence, I like to think of IBC as networking two machines, where the most common use is transferring files between the machine A and machine B. However, ICA is more akin to one machine (A) giving control to another machine (B) and allowing users on machine B to run applications on machine A.
Let’s look at a generic description.
Now let’s look at a specific example — stable diffusion. Stable diffusion is an image generation technique that has recently been popularized in the machine learning community. It aligns text embeddings to visual representations, so you can simply type a caption, and the generative diffusion model will attempt to create an image that matches the prompt.
Here, we can initiate an app-chain msg from the LUNC chain, by wrapping in a generic Msg format. This is a sample command on might run, where the msg.json holds the tx to be executed on the app-chain.
terrad tx intertx msg.json -connection-id connection-0 -chain-id columbus-5
The intertx is a custom ICA controller middleware that handles the wrapping of the Msg to send to the app chain. When this message is broadcast on Terra, gas fees are incurred just like any other normal transaction. The relayer picks up the message, and sends it to the app chain. On the app chain side, the ICA account receives the message, unwraps the tx and executes.
In summary, the envisioned utilization of the app chain would be seamless to a user with ICA. You would connect via terra station wallet, execute ML tasks, and generate gas fees for the LUNC chain.