Future Work and Roadmap
For a more detailed look at the problems unsolved by Rakis (yet), see the launch blog post (opens in a new tab).
Rakis has a lot of moving components and layers, and is currently entering what we think is best described as a Stability Test. This section outlines some of the planned features, performance optimizations, and the long-term vision for the project.
Planned Features
Sybil Resistance and Incentive Mechanisms
One of the major challenges in any decentralized network is preventing Sybil attacks, where a single entity creates multiple identities to gain an unfair advantage or influence. Currently, Rakis relies on a commitment-and-reveal system, but this is not a complete solution. Future work includes implementing robust Sybil resistance mechanisms and introducing incentive structures to encourage honest participation and discourage malicious behavior.
Potential solutions being explored include: - Proof-of-Work or Proof-of-Stake systems - Reputation-based incentives - Slashing mechanisms for bad actors
Heterogeneous Inference Models
Currently, Rakis supports a limited set of pre-defined inference models. However, the true power of the network lies in its ability to leverage a diverse range of models from different providers. Future iterations of Rakis will enable participants to contribute their own custom models, further increasing the network's capabilities and resilience against prompt-engineering attacks.
Improved Consensus Algorithms
The current consensus mechanism in Rakis is a simple clustering-based approach. While functional, it may not be optimal for all use cases. Future work includes exploring more advanced clustering algorithms, semistructured consensus, and other things discussed in the launch post. network's utility.
Performance Optimizations
As Rakis grows and gains more users, performance optimizations will become crucial. Some areas of focus include:
- Improving the scalability and throughput of the P2P network
- Optimizing the inference and consensus algorithms for better resource utilization
- Implementing caching and other techniques to reduce redundant computations
- Exploring sharding and other techniques to distribute the workload across the network
Long-Term Vision
The long-term vision for Rakis is to become a decentralized, self-sustaining ecosystem for Verifiable AI inference and compute. This ecosystem would enable developers, researchers, and organizations to contribute and access AI capabilities in a secure, fair, and transparent manner.
As an open-source project, the success of Rakis relies on the involvement and contributions of the community. We invite developers, researchers, and enthusiasts to join us in shaping the future of decentralized AI inference.
To get more involved in Rakis, see the active TODOs here. (opens in a new tab)