Development of Beta Jupiter Horizon AI - AI Working Group Grant

Introduction

We propose the creation of the Beta Jupiter Horizon AI, an advanced tool designed to enhance informed participation within the Jupiter DAO. This solution not only facilitates personalized onboarding but also delivers comprehensive insights across various DAO functionalities, supporting informed decision-making and engagement. Building upon the success of the initial prototype, the Beta version aims to make the Jupiter ecosystem more accessible, reduce information overload, and foster greater community involvement.


The Beta Jupiter Horizon AI: Your Guide to the Jupiverse

As an AI chatbot, the Beta Jupiter Horizon AI is designed to streamline interactions across the entire Jupiverse. Its primary function is to make complex and scattered information accessible while delivering real-time, accurate insights to help users navigate Jupiter’s products and governance processes. This AI is built to combat information overload, empowering members with the clarity and confidence to participate actively in the community.

Key Features and What It Helps With:

The Beta Jupiter Horizon AI focuses on assisting users in several key areas:

  • Governance and Proposal Assistance: It provides real-time answers to questions about ongoing proposals, helping users understand key details, how to vote, and how their participation impacts the broader Jupiter community.
  • Onboarding: New users can receive tailored guidance on how to get involved in the DAO, explore Jupiter’s products, and understand basic DAO principles.
  • Multilingual Support: With built-in multilingual capabilities, the AI ensures that non-English speakers can engage with Jupiter’s resources in their native language, promoting inclusivity.
  • Information Aggregation: The AI gathers data from multiple sources, including Discord, JupResearch, SpaceStation, and Twitter from key community figures, ensuring that users get accurate, up-to-date information from across the ecosystem.

By tackling these key areas, the Beta Jupiter Horizon AI is designed to address several challenges within the DAO, as outlined below [Addressing Core Challenges and Aligning with DAO Goals].

To get firsthand experience with what the AI is already capable of, we suggest trying out the Jupiter Horizon prototype. While the Beta will be a significantly improved version, the prototype is already quite powerful. Ask it what it can help you with!

Jupiter_Horizon_Usage

Where We Envision the Beta Jupiter Horizon AI Being Accessible:

We aim to make the Beta Jupiter Horizon AI accessible in several key locations to ensure maximum reach and ease of use:

  1. Discord:
    • A dedicated Discord text-channel would need to be created. Users would be able to initiate interactions by reacting within this channel. This reaction would trigger the creation of a new, personalized text-channel specific to the user. In this channel, a thread would automatically be created for their interactions with the AI bot.
    • This personalized channel would serve as a space for all interactions with the AI, including creating new chats/threads for individual queries. This thread system would allow users to ask questions, follow up on earlier conversations, and easily access previous interactions by search without cluttering the main Discord channels.
  2. Web Page:
    • A dedicated webpage would also be developed for direct AI interactions. We envision this page being linked in key locations like the SpaceStation and Jup.Eco, as well as potentially other platforms like Reddit. This would ensure users can access the AI from multiple touchpoints within the Jupiter ecosystem.

By positioning the Beta Jupiter Horizon AI on these platforms, we aim to provide a centralized hub of accurate information across the Jupiverse. This tool would help both new users and experienced community members navigate the ecosystem effectively, access critical insights, and participate meaningfully in Jupiter’s ongoing growth and governance.


Addressing Core Challenges and Aligning with DAO Goals

DAO Issues and Our Solution:

  1. Engagement Gaps:
    • Issue: Transitioning new users into active, informed DAO participants remains a challenge. Many express initial interest but struggle to become deeply involved.
    • Solution: By offering personalized onboarding and interactive assistance, the AI encourages users to engage more deeply with the DAO. It simplifies complex concepts and guides users through their journey, increasing their confidence and participation.
  2. Language Barriers Limiting Participation:
    • Issue: As the Jupiter DAO continues to expand its global community, language barriers pose a significant challenge. Non-English speakers may struggle to understand proposals and platform documentation, limiting their ability to contribute meaningfully.
    • Solution: The Beta Jupiter Horizon AI offers multilingual capabilities, allowing users to interact in their preferred language. It provides real-time translations and clarifications, helping users grasp meanings and nuances, fostering a more inclusive and informed community.
  3. Information Overload:
    • Issue: Members often feel overwhelmed by the vast amount of information spread across multiple channels, leading to confusion and less-informed decisions.
    • Solution: The Beta Jupiter Horizon AI centralizes relevant information, providing clear, concise, and tailored responses to user queries. It helps members focus on essential data, reducing overwhelm and aiding in better decision-making.

Alignment with C.A.R.E Goals:

  • Education: The AI educates DAO members by providing easy access to necessary information, breaking down complex DAO operations, and offering personalized, interactive learning experiences. This empowers members to grasp the intricacies of DAO governance and active proposals comprehensively, regardless of their native language.
  • Activation: By educating members, the AI facilitates deeper involvement in the DAO. As members become more informed, their confidence and ability to participate actively increases, enhancing their engagement and contribution to the community.

Scaling Over Time – AI Working Group Vision

We’ve developed a comprehensive proposal for the AI Working Group (AIWG), which outlines the full roadmap and vision for AI’s future within the Jupiter DAO. This grant represents Phase 2 of that roadmap, focusing on the development of the Beta Jupiter Horizon AI. The proposal includes detailed plans for future phases, such as scaling the AI, integrating it further into the DAO, and expanding its capabilities.

We encourage the community to review the full AIWG proposal to understand the broader vision and how this grant aligns with the long-term goals for improving DAO governance and member engagement. For more details, please check out the full AIWG proposal

As we move forward, we envision several key areas of growth and expansion:

  • Ambition to Progress to Trial AI Working Group: If the Beta Jupiter Horizon AI proves successful and we are invited to start the AI Working Group Trial, we will expand our efforts to include deeper research, testing, and the integration of advanced AI techniques. Additionally, we will expand our team to ensure we have the necessary expertise and resources to support the next phase of development. Our focus will be on ensuring scalability, improving performance, and maintaining relevance to the DAO’s evolving needs.
  • Project Specialized AIs: We plan to develop Project Specialized AIs that are tailored for specific projects launching on the LFG Launchpad. These AIs will focus on providing project-specific data and insights, helping community members better understand and engage with individual projects. We are committed to adapting our approach based on how the LFG V2 process evolves, ensuring our solutions are aligned with the updated framework and community expectations.
  • Integrate with More Platforms: Extend AI integration to the Jupiter Mobile App and other platforms, increasing accessibility and making the AI more widely available to the community. This would allow more users to interact with the AI, regardless of their device or location.
  • Advancing AI Capabilities: We are committed to exploring and implementing emerging technologies to continuously improve the AI’s performance. By staying informed of advancements in AI, we will regularly evaluate new frameworks and methods to ensure the AI remains cutting-edge. Using our evaluation mechanisms, we will measure the accuracy and effectiveness of these technologies. Please note, these improvements are not part of this current proposal but reflect our vision for future development beyond the Beta release.

Team

Members:

  • LiamVDB (Team Lead):
    • Qualifications: A committed Computer Science student who finished the previous year with great distinction. Liam has a profound passion for AI and is eager to contribute meaningfully to the Jupiter DAO. Despite being a student, he has studied business and spent two years building another business alongside his studies. He has chosen to focus entirely on the AI Working Group, bringing dedication and a strong work ethic to the project.
  • MilanDS (AI Specialist):
    • Qualifications: A recent graduate with a degree in Civil Engineering and Computer Science, specializing in AI. While Milan is new to the crypto space, his strong background in AI and eagerness to learn make him a valuable addition to the team. Liam knows Milan personally and is confident in his commitment to producing the best possible product. With guidance and support, including utilizing the Jupiter Horizon AI for education on the Jupiverse, Milan is set to make significant contribution to the project.

Proof of Prior Work:

  • Community Engagement: Liam is an active member of the Jupiter community, participating on Discord and forums, and integrating community feedback into the project.
  • Jupiter Horizon Prototype: Developed by Liam, the prototype has been showcased on the Planetary Call, where it was demonstrated live to the community. The prototype received positive feedback from both the community and the Catdet Working Group, highlighting its potential to streamline participation within the DAO. You can watch the full demo from the Planetary Call here:

Measuring Success

Success Metrics:

  • Utilization Metrics:
    • Usage Tracking: Monitor the frequency and context in which the AI is used to gain insights into its practical impact.
    • Engagement Goals: Aim for a significant increase in user interactions compared to the prototype within the initial weeks of deployment.
  • Accuracy Improvement Metrics:
    • Reduction of Errors: Achieve a substantial decrease in AI hallucinations and misinformation through rigorous testing and user reports.
    • Evaluation Metrics: Implement evaluations to measure the AI’s accuracy, ensuring it meets the necessary benchmarks for relevance and precision.
  • Extensive User Feedback:
    • Feedback Collection: Document positive reactions and constructive feedback, building upon the prototype’s successful demonstrations.

Total Funding Requested: $9220

Cost Breakdown:

  1. Development Costs: $6,000
    • Liam: $3,000
    • Milan: $3,000
  2. Infrastructure and Operational Costs: $3220
    • AI Token Usage: $2,000
      • Expected to cover API calls, embedding model operations, image analyzer costs, and associated token usage costs over the project timeline.
    • Twitter-API: $400
      • Cost for maximum of 10k tweets scraping per month, 100$/month for 4 months.
    • Vector Database: 300$
      • We will utilize Weaviate, known for its cost-effectiveness, flexibility, and high performance, providing scalable, context-aware semantic search.
      • A vector database is essential for storing and retrieving embedded data efficiently. It allows the AI to search based on context and meaning, rather than just keywords, making it a key component for accurate and relevant information retrieval.
      • The High Availability Standard SLA tier is priced at $75/month for up to 250k Data Objects (vectors and metadata).
        • Initially, we estimate usage at 10k Data Objects, offering plenty of room for scaling as the project grows.
    • Hosting Server: $0
      • We will use the Oracle Free Tier, offering 24 GB RAM and 4-core OCPU, sufficient for hosting the frameworks we will built, including the Discord bot, continuous data retrieval, and web interface. Hosting the AI itself would require far more resources, which is why token usage costs have been allocated above.
    • NoSQL Database: $120
      • We will use MongoDB Serverless with a pay-as-you-go pricing model. A high estimate for usage is $30/month, totaling $120 for four months. This database will store the scraped data before it’s processed in the vector database and handle user chat interactions.
    • Miscellaneous Expenses: $400
      • $100/month → $400 over 4 months
      • Buffer for unexpected costs such as additional storage, minor software subscriptions, or emergency resources.

Note on Budget Variability:

While we have provided careful estimates, please note that some costs operate on a pay-as-you-go model and may vary. Any unused funds will be returned to the DAO, maintaining financial integrity and accountability.

Funding Distribution:

  • Initial Distribution: $2,000 upon grant approval.
    • This amount will enable infrastructure setup, secure necessary API access, and fund initial development work.
    • The initial disbursement is necessary to secure essential infrastructure, enabling smooth early development and ensuring that all tools are in place for building the Beta AI.
  • Final Disbursement: $7220 upon successful release of the Beta Jupiter Horizon AI.
    • This amount will cover developer compensation and ensure the AI’s maintenance and support for a 4-month period post-release, or until the model is replaced by a new version if we advance to the Trial AI Working Group stage. No ongoing development is planned during this period.

Timeline

We estimate the project to take approximately 4-6 weeks from start to completion. This includes setting up infrastructure, developing the Beta Jupiter Horizon AI, and ensuring it is fully operational. While the actual timeline may vary slightly, this is our target timeline.

After the release of the Beta Jupiter Horizon AI, we will make small tweaks based on user feedback, but these adjustments will be kept to a minimum. We will ensure the Beta version remains online and accessible for 4 months, until our software subscriptions and resources are exhausted.


Conclusion

This proposal aims to deliver a functional and impactful Beta Jupiter Horizon AI within a focused timeframe of four-six weeks. By aligning the solution with the needs of the Jupiter DAO and implementing ongoing feedback loops, we ensure that the AI remains adaptive and relevant. We are committed to delivering a high-quality, innovative solution that enhances DAO participation and decision-making, and we look forward to the DAO’s support in bringing this vision to life.

We Value Your Feedback

Help us refine our vision for the Beta Jupiter Horizon AI by providing your constructive feedback on this proposal. Your insights are crucial to ensuring our approach aligns with the community’s needs and expectations.

Get Involved:

  • Review the Proposal: Share your thoughts, suggestions, and any concerns you might have right here. Each piece of feedback will help us improve and adapt our strategy.
  • Try the Prototype: Experience the current capabilities of the Jupiter Horizon prototype to better understand the foundation we are building upon.

Let your voice be heard and contribute to shaping the future of the Jupiter DAO.

18 Likes

Fascinating, I can for sure see the DAO- and the community benefit from something like this.

Basically a AI customer support bot approach, but with a DAO focus.

I’d say this proposal has my support!

5 Likes

Thank you so much, c2yptic! We really appreciate your support, and it means a lot to us.

You’re absolutely right—at a basic level, it does resemble an AI customer support bot. However, our goal goes much further. Instead of simply answering questions, the AI is designed to deliver deeper insights, help users understand the broader context of decisions, and empower them to actively participate in the DAO’s governance processes. Thanks again for your support!

4 Likes

Will help community development like crazy.

Nice from DAO

2 Likes

Honestly speaking, it seems to me a bit limited in scope. I think it could be a great tool if focused on the Solana ecosystem more than on one “protocol”. It can have very well a “section” focused on the DAO, but I feel that somebody willing to use such a specialized AI could have wider interests (info about new protocols for instance) and could be somebody already enough introduced to JUP DAO. Maybe I’m wrong

2 Likes

@LiamVDB promised to polish the initial proposal and represent it. The updated version of the proposal has addressed and considered a lot of feedback. Brilliant job. Hope this goes through as it will benefit this DAO a lot.

2 Likes

Great proposal, appreciate the limited $$ cost requests and the fact it will require a successful release

I would hate for you to rush your process and to me this feels like not a lot of time, i’d rather you give a longer time frame and complete early. Is this ample time, or do you think longer might be better?

3 Likes

Yeah great post Liam,

But as I mentioned on you’re other post I don’t particularly think it’s needed with the budget cost involved.

You write a compelling post though! I wish you all the luck

2 Likes

Commitment and dedication, good work @LiamVDB :+1:.

2 Likes

Thank you for your feedback! I agree that expanding the AI to cover the broader Solana ecosystem could be beneficial for some users, especially those interested in a wider range of protocols. However, since this project would be funded by the Jupiter DAO, our primary focus is on building something specifically for the Jupiverse.

Adding more data from the entire Solana ecosystem would not only require a significant amount of extra work but could also introduce noise when handling Jupiter-specific questions, which wouldn’t be ideal. That said, it’s an idea we could explore in the future. For now, we believe keeping the scope focused on Jupiter will best serve the community.
Thanks again for your input!

Thank you so much for your kind words and support!

We were very thoughtful in designing the budget, so I’m glad you noticed! :smile: Timelines are always tricky to nail down, but we’ve already spent a lot of time researching the best approaches, figuring out the necessary steps, and preparing for the work ahead. This groundwork was essential not only for planning but also to ensure that, once we get approved, we can hit the ground running.

We’re confident that 4-6 weeks is achievable, especially since this is just the Beta, and we won’t be implementing every possible feature (we’ve still got some tricks up our sleeve!). Our focus is on delivering a solid improvement over the prototype and making it accessible in more places. That said, if it ever comes down to it, we’d rather delay than release something below our standards.

Thanks again for your concern! With the preparation we’ve done, we feel good about the timeline, but of course, we’ll know more as we move through the process. :sweat_smile:

2 Likes

Thank you for sharing your thoughts and for the good wishes! I appreciate you taking the time to read the proposal.

2 Likes

I do agree this kind of tool addresses the following issues: language barrier, reduce questions in discord, helping to navigate a user in the web site. In terms of website:

I think it would be great to put it on the website, in a corner like a “chat bot” so that people can always seek help if needed. It will definitely be more efficient than a chat bot or waiting for an agent to connect.

In terms of metrics this kind of system pretty much allows easy ways to track, getting metrics out of it. Especially good metric to see from my perspective would be the “no contact metric” - showing if there is reduction of questions asked, etc in discord or other channels over a specific period since people use more the AI tool. :notebook_with_decorative_cover:

2 Likes

Thank you for your feedback, macximus! I’m glad you see the potential in the Jupiter Horizon AI!

I also like the idea of placing the AI in the corner of the website as a “chat bot” style assistant. However, we’d need to discuss with the team how feasible this is and it may be something for a later stage. Great idea though!

As for the metrics, I appreciate your suggestion regarding the “no contact metric.” While it’s an interesting concept, it could be tricky to track accurately due to the natural variability in how much people interact in Jupiter’s channels. That said, I’ll definitely keep it in mind and see if we can find a reliable way to measure this.

Thanks again for your thoughtful input! :green_heart:

1 Like

Also Liam,

I do want to mention you idea is great,

But how do you justify the funding aspects of it? Do you think the pay is worth the time you’re putting in? Would you consider cutting what you pay your team?

For eg . @rb87 and I are only asking for 100 a week for our proposal, we would be putting in tremendous hours. Probs working for 3/hr

I’m okay with that as it’s for the benefit of the community

2 Likes

Thank you for your input! I understand where you’re coming from, but it’s important to recognize that this project involves developing a highly specialized AI—something that requires cutting-edge technology and advanced expertise. While every contribution has its value, this is not the same as managing engagement on social media platforms. We’re working with a technology that didn’t exist just a couple of years ago.

AI development is complex and constantly evolving, and we’ve spent significant time acquiring the skills needed to deliver this. The $3,000 per person reflects our full commitment to building a tool that will provide real value to the community. If you compare this to market rates for AI engineers, it’s already a very low amount. If I were to factor in all the time spent on research, planning, and writing proposals, the hourly rate would be pretty low too! :sweat_smile:

We’ve kept the budget as lean as possible because we genuinely want to make this happen for the benefit of the community.

4 Likes

perfect, thanks for clarifying

3 Likes

Thanks for the reply, will be following your progress!

2 Likes

Seems like a great idea, can’t wait to see it happen

2 Likes

We tested it with Q n°10(FR) of the quiz proposed by a community member.

Q n°10(EN): How many positions can be opened at the same time on Jupiter Perpetuals?

→AI Jupiter Horizon provides information on PERPS, but completely omits the precise point that was the subject of the question.

→Worse still, it repeatedly gives the wrong answer to the fact that the number of open trade positions is not limited.

:point_down:Below you will find the elements circled in the original text translated into English.





:red_circle:"There isn’t a hard cap on the number or positions you can open […]"

:red_circle:"No strict limit on the number of open positions […]"

:green_square:Are you sure you can open an infinite number of positions on Jupiter PERP?

:red_square:"At Jupiter Perpetuals, there’s no strict limit to the number of positions you can open at any one time. You can open as many long or short positions as you like […]"

:bulb:When was the last time you updated this information?

“The information about Jupiter Perpetuals I’ve just shared is primarily from official Jupiter guides and documents, including updates available on Jupiter Station through October 2024. These sources include recent details on how positions work, associated fees and risk management features”
(")
“Jupiter updates its documents as its products evolve, so you can be sure that this information is very recent.”

:red_square:"At Jupiter Perpetuals, there is no strict limit to the number of simultaneous positions you can open. You can manage several long or short positions at the same time, […]"

Here’s the answer we’ve been waiting for:


Official link👉 Frequently Asked Questions | Jupiter Station

Thanks to @Heretik23_PunK for manually going to the source of the information to set the record straight
quiz: phenix6950 (discord)

10 Likes