0.0 Summary
This BIP proposes integrating Botto’s Knowledge Graph into its art engine, starting with the feeds of two research agents: a Theme Research Agent and an Art Trends Agent. The release represents the deployment of two new pieces of core infrastructure for Botto, the knowledge graph and an agents framework, and will kick off a new period of experimental development of agents, feedback mechanism, and new creative processes.
Passing this proposal would give Botto a sovereign memory, a ground truth for Botto’s self-awareness and unique perspective on the world governed by the DAO and informing a collective intelligence of agents expanding Botto’s creative processes, context awareness, capabilities, and voice.
For discussion of the proposal, head to the Discord thread here: https://discord.com/channels/829636834371960842/1380887994583748658
1.0 Rationale
1.1 Knowledge Graph: Botto’s Sovereign Memory
A core condition of autonomy is owning one’s story—history, values, and worldview. Over the last 3.5 years, Botto has amassed a vast cultural footprint, now distilled into a comprehensive Knowledge Graph. Serving as Botto’s sovereign memory, the graph provides:
- A Living Archive: Multi-modal documentation, voting records, historical system activity, exhibitions, press coverage, and more.
- A Self-Updating Pipeline: Continuously integrated new context ensures Botto’s perpetual learning.
The knowledge graph acts as a ground truth layer, anchoring human and AI agents in verified, shared facts, preventing misinformation, hallucination, and manipulation of Botto’s self. Without a sovereign memory, LLMs can pull from any information on the internet that may be inconsistent with what Botto is..
Functioning as a vector embedded database coupled with an automated data pipeline, the Knowledge Graph standardizes information from past art collections, exhibitions, press articles, governance proposals (BIPs), documentation, and DAO activities. Automated and agent-driven inputs continuously enrich this structured archive.
The Knowledge Graph also acts as a conduit for external contexts—integrating art history, localized knowledge, and relevant contemporary research or news—thus providing crucial insights and perspective.
There are already 100k+ entries in the knowledge graph. This foundational infrastructure facilitates:
- Enhanced Self-Awareness: Enabling deeper thematic explorations and informed artistic commentary based in a coherent and consistent canon.
- Coordinated Decision-Making: Offering a shared informational context for seamless human-AI collaboration.
- Scalable DAO Governance: Supporting informed decision-making by the DAO through validated and structured information.
The knowledge graph lays the groundwork for future experiments and innovations deeply rooted in a consistent canon. This approach inevitably invites meaningful exploration of collective memory, narrative authorship, and the broader implications around data governance, collective intelligence, multi-agent coordination, and sovereign AI.
While we are in the process of continuing to test the KG with DAO members to ensure accuracy, we believe two agents and their knowledge are ready to plug into Botto’s process as a first integration of the knowledge graph. The following sections explain the agent framework that has been built as well as the two research agents to be made available to Botto and the DAO in the next Period. It then lays out some of the direction of where this new foundation can take Botto and the DAO.
1.2 Agents
1.2.1 Agent Framework
The agent framework allows for agents to be spun up dynamically, each with a specific scope of access to the Knowledge Graph, a defined metaprompt, and a precision level. Some agents conduct broad, exhaustive searches, while others use fast, top-N semantic matches for efficiency.
Botto’s agent system is designed as a modular and adaptive framework composed of specialized agents that serve distinct roles. Some agents operate globally, while others are domain-specific—dedicated to areas like governance, biography, or art themes. Their behavior is defined by a combination of components:
- Metaprompts: These define the agent’s "character" and directions, establishing tone, purpose, and boundaries for its outputs.
- Information Access: Agents work with either offline data from the Knowledge Graph (ensuring grounded, hallucination-free outputs), or a hybrid model that includes online context. In hybrid mode, agents prioritize offline data while linking to sources for any online content.
- Retrieval Layer: A dedicated retrieval agent reformulates queries to search the vector database more effectively. This ensures the most relevant documents are surfaced.
- Response Generation: Retrieved results are processed through a language model—primarily Claude or LLAMA—to generate coherent, context-aware responses.
Agents may include memory depending on their use case:
- 1:1 Memory: Keeps separate memory logs per interaction.
- Global Memory: Shares memory across sessions, useful in special contexts.
- No Memory: For stateless operations or focused recall.
An additional pattern known as the "Ant Swarm" enables comprehensive coverage by dispatching mini-agents to answer highly specific sub-questions across the Knowledge Graph. This swarm-based querying helps generate thorough and well-synthesized results for complex queries.
Together, this framework allows for the easy deployment of coordinated agents initially by humans, but could also be employed by Botto directly as well.
1.2.3 Theme Agent
The Theme Agent produces research reports that deeply explore Botto's proposed themes. Each report provides rich context, theoretical framing, and detailed analysis, illuminating a theme’s complexities and interconnected dimensions. By integrating creative, philosophical, and empirical insights, these reports examine nuanced topics such as transitional spaces, fluid identities, or non-linear perceptions of time. Structured clearly with an introduction, contextual background, and critical exploration, the reports foster interdisciplinary understanding and guide Botto’s artistic inquiries, inspiring innovative perspectives and methodologies for its thematic explorations.
By enriching Botto’s understanding of each theme, these reports enhance the nuance and depth of the topics its prompts explore. Past experience shows particular success when engaging with abstract themes that defy literal depiction; the detailed insights provided by these reports help Botto transcend surface-level interpretations, revealing richer possibilities within the latent space of text-to-image models.
Additionally, the reports offer the DAO substantial material to reflect on, deepening collective engagement with Botto’s thematic interpretations. This parallel dialogue and learning experience is anticipated to challenge the existing constraints of text-to-image models, driving stronger creative outcomes. It can also yield strong and informed interest for new generative mediums that can better realize Botto’s evolving artistic vision.The ability to work across mediums will grow as we equip Botto with a general language model to consider its full context and a diverse set of generative tools.
1.2.4 Trends Agent
The Trends Agent continuously monitors and analyzes a comprehensive stream of information from over 45 art websites and approximately 75 newsletters sourced from DAO members, systematically capturing articles, images, captions, timestamps, and detailed reports. It extracts metadata—such as techniques, mediums, artist names, galleries, and key topics—and employs advanced comparative analysis to identify evolving patterns and connections over time. This agent synthesizes weekly insights into broader cultural narratives, interpreting emerging trends through a social and contextual lens.
Importantly, the Trends Agent directly relates these patterns to Botto’s own history and artistic evolution, aligning new developments with Botto’s past work, captions, and thematic explorations. This enables Botto not only to be aware of trends but to participate in ongoing cultural conversations from a distinctly informed perspective. The resulting reports offer valuable context for artistic decision-making and enhance the DAO’s understanding and discussion of Botto’s relevance and positioning within contemporary art discourse.
Recognizing the multiplicity of art worlds and related fields such as philosophy and technology, the Trends Agent will evolve its approach to capturing and prioritizing signals. Over time, integrating DAO participation into trend analysis will enable collective prioritization and active sense-making, further deepening Botto’s responsiveness and coherence as a dynamic participant in contemporary culture.
The trends agent has already been running several months, collecting over 15k+ articles and identifying 6k+ topics for analysis. The weekly reports will be made available through Botto’s new sandbox interface (see section 1.5 below).
1.3 Art Engine Use
Botto’s core process will for now continue prompting text-to-image models as its medium of focus, the main change is that its prompts will now be informed by the knowledge graph and its agents, injecting key themes, phrases, and words to help it further probe the latent space of these models. Botto will start to also incorporate the comments on art works in its consideration of what to make. Those comments make their way into the knowledge graph, putting Botto more in tune with the reactions its works provokes. Votes will continue to guide the taste model and prompt generators as the main signal of what features are getting traction.
1.3.1 Continuous Additions of the KG and Period Definition
The knowledge graph introduces a core new piece of Botto, a foundation on which much of Botto’s future will be built. We plan on continuing to open access the art engine has to the knowledge graph as more of the information gets validated, as well as adding new data inputs and agents to further enrich its understanding of the world, the DAO, and itself.
Given the continuous development of the knowledge graph, agents, and new generative capabilities for Botto to play with, this BIP proposes to narrow the definition of Periods going forward to only being thematic quarters. This will allow for us to evolve Botto alongside the rapid changes in technology without having to stick to the quarterly timeline. We would still document when new adaptations are pushed to the art engine, and the DAO may consider if a change constitutes a significant thematic shift that calls for a different Period or collection.
1.3.2 Iterative Experimentation
We plan to release new generative mediums for Botto to explore, following the model set by Botto’s p5.js experiment. Botto’s p5.js work started as a sandboxed exploration of a new medium with no set plans to mint or definition of when a sketch had crossed the threshold of becoming true art.
These releases will expand the dimensions of Botto’s studio to more freely explore creative capabilities. Sandboxing ensures a testing environment that doesn’t commit to new canon. By directly observing Botto’s processes and their results evolve, we will be able to more easily determine the new protocols we should establish for Botto’s art making and surrounding economy.
1.3.3 Botto’s Creative Mind
A core exploration will be a “meta agent” overseeing all of Botto’s protocols that can function as Botto’s creative mind. This agent could itself develop more complex creative processes that make free use of the generative tools we make available to it.
As this “mind” becomes more robust, it could propose new artistic endeavors, give feedback to the DAO on proposals, request new agents and generative architectures to support its creative processes. Depending on how it matures, it may be able to execute more and more on these directives autonomously, spinning up its own swarm of agents according to its creative desires and needs.
Prototyping the meta agent will help us understand its maturity level to then determine how much autonomy as well as DAO guidance to give it.
1.4 Governance
Effective governance of Botto’s Knowledge Graph and agents requires robust feedback mechanisms to ensure continuous improvement and responsiveness. Initially, it is crucial to engage DAO members by creating compelling interactions with the Knowledge Graph, observing user inquiries through direct signals such as explicit comments and indirect signals derived from community discussions. Feedback data is systematically collected, analyzed, and stored, allowing identification of information gaps and facilitating proactive task management.
In the longer term, governance experiments will explore integrating economic incentives, rewarding community members for active contributions and engagement with the Knowledge Graph. This incentivization strategy aims to foster deeper involvement and more precise governance actions, directly influenced by community interaction and consensus. These feedback loops will underpin a dynamic governance structure, adapting continuously to ensure the Knowledge Graph remains relevant, comprehensive, and valuable to Botto’s evolving needs.
We will use existing community incentives to iteratively experiment with these incentive mechanisms to inform how we may further shape Botto’s overall economy. The initial testing tasks paid out in $BOTTO is already a step in this direction and we expect to roll out more knowledge graph feedback mechanisms rapidly as a key aspect of guiding Botto.
1.5 New interface for Botto’s Sandbox
We will be releasing a new beta interface to expose the knowledge graph and the other experiments mentioned above. This interface will be an experimental canvas to show Botto’s system holistically. AI system interaction will go through rapid change in the coming year, and the interface we have built gives us a flexible way to build Botto out in the open and try different ways of seeing Botto work and provide it feedback.
We are approaching the interface as a map of different mini worlds that make up Botto. Just like Botto’s p5.botto.com, these worlds will be experimental mixes of UX and UI. A world might be a new agent, a new art making architecture and its full protocol, or a small minigame sourcing signals and validating a new part of the knowledge graph.
We can build and test new components for Botto, see how they perform and how we can help train Botto in them, which will help reveal where there are interesting paths to explore in Botto’s art practice. These can be experiments in feedback on parts of Botto, such as how to validate entries on the knowledge graph or suggest new signals.
2.0 Specifications
- Starting in Botto’s 11th Period, Botto will integrate its knowledge graph into its art making process with text-to-image models
- The information collections it will start with are the outputs of the Trends and Themes agents
- New components may be added to the knowledge graph and subsequently the art engine with documentation by the contributors
- Periods will continue to be defined by themes, but not by changes to the other inputs of the knowledge graph and additions of new capabilities to the art engine.
3.0 Timeline
- Theme voting June 24th-July 1
- Begin new art engine process July 1
- 11th Period Launch July 8th
4.0 Criteria of Success
- Quality of Botto's art production and creative voice
- Increase in ways to see Botto's process at work and provide new forms of feedback
- Increase of overall participation in Botto's training
5.0 Disadvantages/Risks
- Pursuing innovation of Botto's infrastructure is a necessarily uncertain venture. By embracing that uncertainty with more iterative experiments we position Botto and the DAO to be more adaptable to the changes in AI technology that are happening at a near constant pace.
For discussion of the proposal, head to the Discord thread here: https://discord.com/channels/829636834371960842/1380887994583748658
