Published - September 2nd 2023 Link to Yellow Paper Technical Details: https://github.com/MorpheusAIs/Morpheus/blob/main/YellowPaper.md
Morpheus is designed to incentivize the first peer-to-peer network of personal AIs, known as Smart Agents. Providing users open-source Smart Agents to connect to their wallets, Dapps, & smart contracts promises to open the world of Web3 to everyone.
Finally, the average user can talk with their Smart Agent in normal language and have it understand the question and take an action based on their intent/approval. This moment is similar to how Google's search engine opened the early internet up to the general public through their easy to use web interface in the late 1990s.
To make Smart Agents accessible to everyone and increase decentralization of their infrastructure we propose the development of the Morpheus network. The Morpheus network will include a fairly launched token (the "MOR" token) for incentivizing all four of the key contributors to the network. Namely, the community of users providing data, coders contributing to the Morpheus software/agents, capital providers funding development/operations and those supplying computation, storage and bandwidth. It has been well shown by the history of Bitcoin and Ethereum that free & open competition for scarce digital tokens can provide scalable infrastructure for a public blockchain over long periods of time.
Incumbents such as OpenAI, Microsoft, and Google are running close-sourced large language models, charging customers license fees, and monetizing customer data. These models are censored, fragile, and operate in walled gardens. There is a strong demand for an open-source large language model available for free. Llama, Falcon, and other open-source LLMs have recently been released and are rapidly approaching the accuracy of their close-sourced competitors.
What these open-source LLMs currently lack is a standard graphical interface by which users can chat with them, an API for developers, a cloud solution to move between devices and a way to manage user data and the recovery process. This is where the Smart Agent Protocol comes in, as it provides an open-source LLM run locally and is managed by a user's Web3 wallet.
However the local only approach still lacks an API for developers to build on and the cloud solution where a network of users can run the software on powerful hardware to enable use cases such as light clients, where the user doesn't need to download the full node or Smart Agent locally.
Morpheus will provide these APIs and decentralized cloud functions by launching a network and a token to reward people providing this public blockchain infrastructure to the Smart Agent community. As an implementation of the Smart Agent Protocol, Morpheus seeks to gather the resources required for open source based personal AIs to match and then exceed the capabilities of the tech companies offering closed GPT models today.
Morpheus has many advantages right away. Being Web3 native the user can buy or sell crypto, send stablecoins, access smart contracts, use Dapps and DeFi services, which no LLM is connected to today. Regulatory barriers faced by centralized companies prevent them from offering these tools to users, so their models can chat about tasks, but not act on the user's behalf in a Web3 context. Running on the decentralized public infrastructure is cheaper than paying Chat GPT a license fee for every new user.
Morpheus is the Linux type alternative for developers who want to be able to quickly spin up new agents / LLMs at no cost. The user can maintain ownership of their business or personal data. This avoids leaks, hacks, and attacks from competitors. By rewarding developers for contributing code not just to Morpheus but also to build more specialized Agents, an App Store/ Agent Store type experience for users will develop. With persistence of data, prompts, and history owned by the user the Smart Agent Protocol becomes the best solution for interoperability in the world of LLMs and Agents.
Lastly, Morpheus having a graphical user interface and leveraging Electron to package it as a one-click install, allows Morpheus to become the first open-source AI to pass the famous "Friedl test". This is a threshold that benchmarks when the ease of use of a software first makes it accessible to non-technical members of the general public.
Our proposal provides for this with a Morpheus token (symbol "MOR"). MOR is rewarded each day 24% to the community, 24% to capital, 24% to compute, & 24% to coders. 4% for protection funds.
This reflects the reality that for Morpheus to grow, it needs the following:
Community - Users bring and own their data.
Capital - Brings the funding for compute and code.
Compute - Providers the equipment and power.
Coders - Provide the intelligence to use the data, capital and compute.
The MOR Token Supply is limited to a maximum of 42,000,000 tokens that will ever exist. The distribution will start with all four groups earning the tokens by providing forms of proof of work (labor) and proof of Stake (capital) to the network. No pre-mine. No early token sale. Just a fair launch.
The block reward will start at 14,400 MOR per day and then decline by 2.468994701 MOR each day, until the reward reaches 0 on day 5,833. By that time (about 16 years from now) provided Morpheus is in wide use, the fees will have taken over as the primary incentive. Fees paid to users for their data, fees to compute providers, fees to capital providers and fees to coders.
42 Million supply cap for the MOR tokens. 14,400 tokens per day evenly distributed between community users, capital, code, compute.
- 3,456 tokens for compute. Proof transactions for API calls served.
- 3,456 tokens for code. Proof code committed and merged into Morpheus repo.
- 3,456 tokens for capital. Proof ETH yield contributed, swapped for MOR & burned.
- 3,456 tokens for data. Proof of feedback on prompts, agents, and Morpheus. With the remainder set aside for protection resources: 576 tokens per day for that purpose.
The objective is for MOR to provide broad utility in many of the functions of the Morpheus network. As a result, the use of the MOR token offers an on-chain accounting mechanism to calculate rewards based on real usage of the software.
Developers pay MOR to compute providers for functions beyond what local hardware can execute. MOR pays for the Morpheus API calls for decentralized applications using the Smart Agent Protocol. Users can pay MOR for specialized Agents released by developers. In turn, developers can pay users MOR for training data for new LLMs/Agents.
All projects go through phases of development. It's important that at the beginning liquid resources such as ETH are used to pay devs and for equipment. Ethereum did the same thing when they leveraged BTC from their community to pay for the initial coding of their blockchain. The difference here is that the Smart Agent Protocol has already been developed and Morpheus is implementing a version to broaden its reach, so there isn't a need for a crowd sale before the project goes live. MOR tokens will only be rewarded after the software is live.
Capital - ETH locked in a smart contract that generates yield (ETH in a Thorchain saver vault for example).
The value of the ETH yield is converted weekly into MOR and sent to all compute providers + coders + data contributors above a threshold of 10 times the gas fee required for the transaction on Polygon (Ethereum 2nd layer).
- Capital, The value of their ETH yield will be calculated against all ETH contributed and MOR yield rewarded to them pro-rata.
- Compute, The value of the ETH yield is converted into MOR and sent to compute providers.
- Coders, The value of the ETH yield is converted into MOR and sent to the coders.
- Community, The value of the ETH yield is converted into MOR and sent to the users contributing training data.
- Capital, A balance between block rewards and fees earned develop.
- Compute, A balance between block rewards and fees earned develop.
- Coders, A balance between block rewards and fees earned develop.
- Community, A balance between block rewards and fees earned develop.
- Capital, Fees for providing ETH liquidity to the MOR token will provide a majority of their rewards.
- Compute, Fees paid to compute providers will form the majority of their rewards.
- Coders, Fees going to coders will form the majority of their rewards.
- Community of Users, Fees paid to users share with data providers will provide a majority of their rewards.
Note this isn't a timeline. Rather each phase is a description of part of the life cycle. It may take years for the community to grow and mature through each phase and the block reward expires after about 16 years. This lengthy distribution schedule is intended to give time for the tokens to be rewarded on a very broad basis globally. Also, the smooth daily decline in block rewards over many years gives all participants time to achieve scale and transition from the early subsidized rewards to operating solely by the fees they earn.
Ever since Bitcoin's launch people have argued about "what will happen when the block rewards finally stop?" To avoid this unhelpful debate in the context of Morpheus and to continue aligning new coders, community, compute & capital providers long into the future, we propose a "tail emission" of MOR tokens. This MOR tail emission will start after the last MOR tokens have been emitted on day 5,833 of the distribution schedule.
The tail emission will be calculated by reviewing the number of burned MOR tokens in the past 5,833 days and setting the tail emission value to be 50% of the burned amount. This tail emission value will be emitted in the next 5,833 day period. But in no case will the tail emission be greater than 16% of the then in circulation MOR.
For example, if on average 25% of MOR tokens were burned during the first 5,833 days then 10,500,000 MOR would have been burned during the first emission schedule. Then by applying the 50% tail emissions value we calculate 5,250,000 MOR can be rewarded in the second 5,833 day period. Which works out to about 16.6% of the 31,500,000 MOR left in circulation. According this amount would be further reduced to 5,040,000 MOR (16% of tokens in circulation) to be rewarded in the second 5,833 day period or ~864 MOR per day.
After the second 5,833 day period is complete, this process will repeat. The tail emission will be calculated again by reviewing the number of burned MOR tokens in the past 5,833 days and setting the tail emission value to be 50% of the burned amount. This tail emission value will be emitted in the next 5,833 day period. But in no case will the tail emission be greater than 16% of the then in circulation MOR.
For example, again if 25% of MOR tokens were burned during the second period, that equals 9,135,000 MOR were burned during the second emission schedule. Then 4,567,500 MOR can be rewarded in the third 5,833 day period. However, since this number is more than 16% of the then 27,405,000 MOR then left in circulation it will be reduced to 4,384,800 MOR to align with the 1% annual rewards (relative to tokens in circulation).
This process repeats forever on into the future.
Longterm result. About 1% of annual MOR rewards (relative to the number of MOR then in circulation) will be available for future coders, compute, community & capital.
IMPORTANT NOTE: This does not alter the nature of the hard Supply Cap of 42 Million MOR. Since the tail emission schedule is by definition only a portion of the MOR tokens which were burned, thus the MOR token can only become ever more scarce with each 5,833 day period.
Below is shown the unified MOR Supply Curve showing the first 5,833 day period and adding the long tail emissions from year 17 to year 256. Presuming the example of a 25% average burn rate of MOR over the epochs.
The path for all Proofs in Morpheus starts with downloading a copy of the Morpheus full node. This includes a local copy of the open-source large language model on their machine. Once running a full node the user is in a position to contribute to any of the four functions of the network.
The Morpheus full node comes with a wallet or the user can connect their existing wallet. This enables the user to sign and send transactions recommended by their Smart Agent.
The definition of a Capital Provider is someone who has downloaded the full node and connected their wallet and locked at least 0.01 ETH in a yielding smart contract. This Capital Provider Smart Contract will provide 100% of the ETH yield produced to the Morpheus swap function. The swap buys the MOR tokens from an Automated Market Maker (AMM) and then burns those MOR tokens. This will provide liquidity to all those coders, community members and compute providers without recirculating the MOR tokens.
As a result, the Capital Provider will receive MOR tokens each day in pro-rata to their total ETH yield contributed. For example, if there are 100 Capital Providers who each contributed 1 ETH on day 1 when the network launches, then each one gets 1% of the 3,456 MOR tokens each day = 34.56 MOR.
The definition of a Coder is someone who has downloaded the Morpheus full node, connected their wallet, and contributed an agent, smart contract or other software to the Morpheus Network.
The Coder will send a MOR transaction to the Coder Smart Contract to register their agent / smart contract or software. The Coder will include in the memo of the transaction the following metadata.
- A. An IPFS link to their software's endpoint in the memo field of the MOR transaction when they register.
- B. A cryptographic signature, similar to how developers sign/authenticate app releases.
- C. The version number of the software.
- D. A hash of the state of the program, so users can check it is a valid and unaltered copy.
The contributors to the Morpheus codebase are rewarded in proportion to all the cumulative development done on the repository as gauged by Full Time Equivalent (FTE) work contributed. For example, if there are 10 coders each having contributed 10% of the FTE time when the network launches, then each one gets 10% of the 3,456 MOR tokens each day = 345.6 MOR. This calculation is updated each month based on the cumulative FTE time contributions of the current mainnet version of the Morpheus software.
Once there are specialized agents or tools or chains (sequence of prompts/calls to an LLM) interoperable with Morpheus, then half (50%) of the rewards will go to their developers. The reward will be calculated in proportion to the usage of those agents. For example, if there are 10 developers who built 10 agents each generating 10% of the agent usage on the Morpheus network. The Morpheus smart contract will calculate those usage statistics via MOR transactions. Then the Morpheus software coders would earn 50% of the MOR reward and each developer of a specialized agent would get 5% of the tokens = 172.8 MOR per developer in this example.
The definition of a Compute provider is they have downloaded the Morpheus full node, connected their wallet and are ready to receive API / LLM / Compute requests from other Morpheus users. Compute provided can be calculated by including a signed transaction generated by the Smart Agent with the return of the output from the MOR transaction.
The Compute Provider will send a MOR transaction to the Compute Provider Smart Contract to register the API endpoint for receiving requests. The Compute Provider will include in the memo of the transaction the following metadata.
- A. An IPFS link to their APIs endpoint in the memo field of the MOR transaction when they register.
- B. A cryptographic signature, similar to how developers sign/authenticate app releases.
- C. The version number of the Morpheus software they are using.
- D. A hash of the state of the program, so users can check it is a valid and unaltered copy.
The pro-rata MOR transaction fees burned by each Compute Provider serves as proof of the Compute Providers status and earns a proportion of the MOR tokens each day.
For example, if there are 100 Compute Providers on day 1 when the network launches, then each one gets a pro-rata reward based on the amount of MOR they have burned via fees. In this case, presuming each of the 100 compute providers burned 100 MOR, then 1% of the 3,456 MOR tokens each day = 34.56 MOR.
To qualify to receive Compute requests a Compute Provider's address must HODL MOR tokens. Requests will be routed to the Compute Providers API pro-rata to the MOR they hodl compared to all other registered Compute Providers.
The definition of a Community Data Provider is they have downloaded the Morpheus full node, connected their wallet and are ready to send data back to the Morpheus network for LLM training. This is known as Reinforcement Learning from Human Feedback (RLHF) and is critical to improving the Morpheus software, agents and tools. Data provided can be calculated by including a signed transaction generated by the Smart Agent with the return of the output from the MOR transaction.
The Data Provider will send a MOR transaction to the Compute Provider Smart Contract to register the API endpoint for receiving requests. The Data Provider will include in the memo of the transaction the following metadata.
- A. An IPFS link to their data via an endpoint in the memo field of the MOR transaction when they register.
- B. A cryptographic signature, similar to how developers sign/authenticate app releases.
- C. The version number of the Morpheus software they are using.
- D. A hash of the state of the data, so users can check it is a valid and unaltered copy.
The pro-rata MOR transaction fees burned by each Data Provider serves as proof of the Data Provider's status and earns a proportion of the MOR tokens each day.
For example, if there are 100 Data Providers on day 1 when the network launches, then each one gets a pro-rata reward based on the amount of MOR they have burned via fees. In this case, presuming each of the 100 data providers burned 100 MOR, then 1% of the 3,456 MOR tokens each day = 34.56 MOR.
To qualify to receive Compute requests a Data Provider's address must HODL MOR tokens. Their data provided can be rewarded pro-rata to the MOR they hodl.
Note that Morpheus does not need to set aside MOR rewards for blockchain consensus or transaction execution on a distributed ledger thanks to building on Ethereum's 2nd layer "Polygon". Yield from the ETH locked by users and capital providers will cover the gas fees needed for the Smart Agents and Morpheus Smart Contracts to operate. This selection vastly simplifies the Morpheus software as developing consensus and securing transactions is a very time and resource intensive process.
The best systems pick the least number of magic numbers and instead let the free market decide as many variables as possible. Fees are a great example of this. Rather than picking an arbitrary fee to charge instead, Morpheus leaves these numbers up to users, devs, capital and compute providers. For example, if a compute provider can offer a $0.02 price per 1,000 language tokens for their LLM and a user decides to pay it, then that's what the market is willing to pay. As compute speeds up prices are likely to change and so it's better to leave these and other variables up to those using the Morpheus software.
Fees For Compute Amount of fee set by users and compute provider. Option to pay compute & burn MOR tokens with each fee. Open market to develop over time. Free marketplace for compute instead of consensus or privileged nodes.
Fees For Code / Agent Intelligence Amount of fee set by coder and accepted by user. Option to pay fee & burn MOR tokens with each fee. Open market to develop over time. Free marketplace for code instead of consensus on tasks.
Fees For Capital Amount of fee set by LP & accepted by user. Option to pay fee & burn MOR tokens with each fee. Open market to develop over time. Free marketplace for capital instead of consensus on treasury.
Amount of fee set by users and accepted by data buyer. Option to pay fee & burn MOR tokens with each fee. Open market to develop over time. Free marketplace for data.
All fees are paid in native MOR tokens creating natural demand in the system as usage grows.
Use of Fees To Incentivize Honest Agents & Repair Losses In Case of Errors Another important use for MOR & ETH in the Morpheus network will be compensating users in case of Smart Agent / Smart Contract failures. We believe curating and building reputation backed up with economic resources will be key to growing confidence in Smart Agents and having a source of funding to address errors, bugs, and other issues that arise. After a major bug and the resulting hard fork of Bitcoin in 2010, an early core developer named Gavin Andresen stepped up to pay Bitcoin to miners who had lost rewards due to the hard fork. This action was important and quickly resolved the hard fork but it was ad hoc.
Recognizing ahead of time that software is never perfect and setting aside 4% of MOR resources to repay those affected by a bug in the code. The Morpheus developer community will serve as an oracle to recognize when a bug or error has had an economic effect on a user, compute provider, or capital provider. A predefined set of failures will be covered by these resources to include bugs in the Morpheus Smart Contract or local install.
For broader protection, an integration with Nexus Mutual or similar smart contract / decentralized protection network could be considered to cover edge cases with agents / smart contracts that want to be included in a Morpheus Agent Store or better ranked by the SmartContractRank algorithm.
Storage For Persistence & Wallets For Recovery Rather than store personal data in the Morpheus network itself, which would be cost prohibitive and a centralizing force, individuals will hold the private keys controlling access to their data, prompts & wallet. The data itself will be stored using the IPFS standard and the Filecoin network for decentralized long-term storage. Leveraging the Filecoin EVM and DeFi for yield a permanent recurring storage can be arranged. Alternatively, users can pay annually ENS style for storage. The approach and keeping the private Web3 wallet as the key for movement/recovery of this data to different devices as the user changes computers or phones.
The Morpheus implementation of the Smart Agent Protocol will be a direct fork of the existing locally run repo. The most significant change will be to update SmartContractRank to include knowledge of the MOR token and its functions in powering a network of Smart Agents.
The Morpheus MOR token is being developed as a Smart Contract on Ethereum via the ERC20 standard for fungible tokens. Most smart contracts are on Ethereum and the Ethereum Virtual Machine has become the lingua franca of the Web3 space. To reduce the cost of gas fees for sending daily rewards we will leverage Ethereum's layer 2 called Polygon.
We believe the choice of the Ethereum blockchain is the best starting place given on chain actions such as staking ETH can only be validated by a smart contract running on the same chain. In addition on chain validation of coding via ENS domains or Ethereum public addresses adds another means of connecting code contributed to the wallet of the coder who provided it. A record of which the Morpheus Smart Contract can access daily.
Also, zero-knowledge proofs for scalability and privacy are key to many use cases. So starting with these capabilities day one will put the Smart Agent community in the best position for the future. Polygon is in the process of adding ZK tech, much of which is already live.
In the near term, this tech stack selection secures Morpheus directly in Ethereum's layer 1 security with the reduced gas costs of a layer two. In the mid-term, this selection also provides a pathway to expand Morpheus to other Ethereum layer 2s and EVM compatible blockchains.
As interoperability improves between public blockchains Morpheus will seek to serve all Web3 AI Agent builders across the various EVM / solidity compatible developer communities. We recognize strong builder communities on Polygon, OP Stack, Base, Arbitrum, Avalanche, Polkadot, Solana, Filecoin & Cosmos that share a similar vision and values. Morpheus is only possible today thanks to the tools built by developers across many of these chains.
To avoid leaking private data when sending prompts to the Morpheus peer-to-peer network of Compute Providers, the software should seek to leverage Fully Homomorphic Encryption (FHE) versions of Large Language Models as they are released. Also with the advent of hardware acceleration for FHE in 2024/2025, it's projected that costs for this computation will reach parity with plain text processing.
- LLM example https://huggingface.co/blog/encrypted-llm
- EVM example https://www.fhenix.io/
We are close to an important moment in history. With Morpheus, everyone will have a powerful personal AI capable of thinking with them and taking actions to benefit them. In the same way the personal computer and the search engine empowered the individual, we have the same opportunity with personal AIs today. The Smart Agent Protocol brings together the right mix of capabilities with LLMs, Agents, and Web3. Morpheus extends those capabilities into a public network capable of accelerating the mass distribution and usage of Smart Agents.
We believe the economic alignment of incentives is ultimately how we secure the best outcomes from the coming of AGI. Help us secure an open source, permissionless and free future for everyone.
I received an email from a developer named Morpheus September 2nd 2023 with the above proposal.
David,
Find below a proposal for launching "Morpheus - A Network For Powering Smart Agents".
The paper lays out the token economics, tech stack, and means of calculating proofs for fairly rewarding the community, coders, capital & compute providers with tokens.
This paper is freely given to the Smart Agent community and is available in the public domain.
Free your mind.
Morpheus