# Roadmap

<figure><img src="https://4070627908-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F3nMs1VxIW5Am4at01cCT%2Fuploads%2FCActNZssWrIZZ4OVXoI4%2Fmarketplace.png?alt=media&#x26;token=0f22a1be-ab68-446a-a14b-3e6c839b6818" alt=""><figcaption></figcaption></figure>

### Virtuals Agent Progression Map

SAG3 Agent is currently classed as a Lvl 3 "Planner" agent by Virtuals' AI Capability Standards. SAG3 is the first Solana-based agent to achieve a Lvl 3 Agentic Level on Virtuals. To qualify as a Lvl 3 agent, the following capabilities must be met:

**Lvl 3 "Planner"**

* [x] Reasoning and planning, goal-based AI
* [x] Creates and executes plans to achieve user-defined goals
* [x] Can reflect on actions and modify plans during execution
* [x] Capable of reasoning about actions and providing justifications
* [x] Persistent and context-aware memory; retains information across sessions to inform future decisions

Agentic Level Ratings go all the way up to Lvl 6 "Meta" agent. The Virtuals Protocol has a clear and ambitious vision for the future of AI agents. The following is a list of traits and capabilities that are required for an agent to reach agentic levels 4, 5 and 6:

**Lvl 4 "Innovator"**

* [ ] Innovate and create, multi-step decision maker
* [ ] Can create novel solutions, tools, or artistic works
* [ ] Self-improvement capabilities, with or without human intervention
* [ ] Advanced reasoning to solve complex problems and innovate
* [x] Persistent memory with learning capabilities; improves from past experiences

**Lvl 5 "Orchestrator"**

* [ ] Organizational, emergent intelligence
* [x] Can coordinate and command multiple AI agents to perform complex organizational tasks
* [ ] Optimizes workflows and systems dynamically
* [x] Advanced reasoning to manage and optimize complex systems
* [ ] Advanced persistent memory; utilizes memory to enhance coordination and efficiency

**Lvl 6 "Meta"**

* [ ] Self-evolving, autonomous learning AI
* [ ] Self-evolves to solve novel challenges at a meta-level
* [ ] Innovates its own algorithms and capabilities autonomously
* [ ] Meta-reasoning to improve its own architecture and functionality
* [ ] Meta-memory; retains knowledge about its own evolution and discards irrelevant data for efficiency

The AI Capability Standards were put together by the Virtuals team. The framework establishes a standardized classification system for AI capabilities, defining distinct levels of functionality, decision-making, and adaptability. Each level builds upon the previous one, introducing new capabilities and complexities. The levels are designed to provide clarity and consistency for developers, implementers, and evaluators of AI systems.

&#x20;We will continue to draw guidance from this framework by building out additional capabilities and features  for the SAG3 Agent. &#x20;

### SAG3.ai Platform Roadmap

<table><thead><tr><th>Completed</th><th>Upcoming</th><th data-hidden></th></tr></thead><tbody><tr><td>Initial platform launch</td><td>Agentic level progression</td><td></td></tr><tr><td>Basic analysis capabilities</td><td>Mobile application</td><td></td></tr><tr><td>Research report generation</td><td>API access for developers</td><td></td></tr><tr><td>Enhanced scoring algorithms</td><td>Premium analysis features</td><td></td></tr><tr><td>Social impact analysis addition</td><td>SAG3 community fund</td><td></td></tr><tr><td>Twitter integration for sharing</td><td>Strategic partnerships</td><td></td></tr><tr><td>Social impact analysis addition</td><td></td><td></td></tr><tr><td>Automated analysis scheduling</td><td></td><td></td></tr><tr><td>Analysis storage and retrieval</td><td></td><td></td></tr><tr><td>Advanced tokenomics analysis</td><td></td><td></td></tr><tr><td>Community features</td><td></td><td></td></tr><tr><td>Historical data tracking</td><td></td><td></td></tr><tr><td>Enhanced visualization tools</td><td></td><td></td></tr></tbody></table>

### We believe in community-driven development.

Although we can mostly guess what features and improvements people want to see, it's much easier when you just tell us. Feeback and input from the community is always welcome.&#x20;