Dear Community Members and Team,
Currently, community education and developer support within the ZK ecosystem rely primarily on manual content creation and human responses. While these efforts have been valuable in helping users understand ZKsync as a network and reducing misunderstandings, as the ecosystem continues to grow, relying solely on manual work faces limitations in terms of response speed, coverage, and proactivity.
I would like to propose introducing a ZK AI Agent as a supplementary tool to enhance the efficiency of community education and developer guidance. The goal is not to replace human work, but to use technology to make these efforts more scalable and sustainable.
The core idea is to train an AI Agent using high-quality materials such as official documentation, technical proposals, and code repositories. This agent would continuously learn about ZK-related technologies and communication styles. It could operate 24/7 across platforms like Discord, forums, and X (Twitter), responding to user inquiries and proactively engaging with potential developers to provide technical explanations and development guidance.
To implement this in a controlled manner, I suggest a phased approach:
Phase 1 (Validation Phase): The community and team would oversee the training and supervision of the AI Agent. It would start by handling frequently asked questions, gradually learning ZK’s technical content and community communication style. The focus at this stage would be on passive responses while accumulating data and improving answer quality.
Phase 2 (Assistance Phase): The AI Agent would begin assisting with daily inquiries and developer guidance. Human contributors could then focus more on complex issues and in-depth technical content.
Phase 3 (Proactive Developer Outreach): The AI Agent would take on a more proactive role in developer recruitment. It could monitor platforms such as Discord, forums, and GitHub to identify potential developers, initiate conversations, and provide targeted technical introductions and development pathways to help lower the barrier to entry for the ZK ecosystem.
Phase 4 (Long-term Exploration): As the AI Agent’s capabilities mature, we could explore allowing it to assist, under human supervision, with development-related tasks on the ZK platform. For example, generating basic code frameworks, providing development templates, or assisting with code reviews. Additionally, we could explore incentive mechanisms where a portion of the value generated from AI-assisted development is used to buy back ZK tokens, creating a positive feedback loop.
Regarding costs, based on different levels of activity and model usage strategies, a preliminary estimate is as follows (using current market prices):
- Phase 1: Approximately $6,000 – $8,000 per month
- Phase 2: Approximately $9,000 – $13,000 per month
- Phase 3 (with increased proactive engagement): Approximately $12,000 – $18,000 per month
These costs can be managed by using a combination of different models. While the numbers may seem significant, they should be viewed in the context of long-term ecosystem growth.
In terms of expected outcomes, even if only 4%–10% of the developers engaged by the AI Agent eventually start building on ZK or contribute code, the cumulative effect over time could still bring meaningful growth to the ecosystem. This is a long-term initiative that requires continuous iteration and feedback.
I believe that introducing an AI Agent in a phased manner could help the ZK ecosystem improve its developer friendliness and educational efficiency. I welcome any thoughts, suggestions, or discussions from the community and the team.