Prime Directives 2023-11-15 - Enhance Agent Capabilities; Construct the SOB

[daveshap]

I've been pleased to see the Collective becoming more self-sufficient and self-directed. This is a great model for how we'd like our swarm to eventually evolve.

Main Discussions

• The Cost of Tokens #97
• Agent Relation And Communication Architecture Theory Proposal #39
• Direct command sequencing to promote utilisation of documents #1
• Tool Builder - Interacting with the real world. Incredible behavior #62
• What about using Pub/Sub for AI Agent communications #32
• Automated Unit Testing #125

The overall themes that are emerging:

Agent Capabilities

The Collective is focusing on the abilities and limitations that we're finding with the current agent schemas as provided by OpenAI. Knowledge management is a limitation, however, we have made substantial progress on tool making and tool use. Please continue enhancing individual agent capabilities and resilience. Share your findings, document results.

SOB Steering and Self-Correction

There's been growing conversation around ethical steering and self correction. This is a meritorious conversation. Remember, the swarm's autonomy is defined by three categories:

1. Self-Direction
2. Self-Correction
3. Self-Improvement

To reflect this, I've created a new bounty: #149

New Deputy

Please welcome @GeorgiaPhillips1008 as the latest contributor. She's a core member of the ACE Framework team and has jumped into the HAAS project. Thanks for joining the Collective!

New Discussion Category - Assimilation Targets

https://github.com/daveshap/OpenAI_Agent_Swarm/discussions/categories/assimilation-targets

I noticed folks posting a lot of other projects, code, videos, and papers. The Assimilation Targets discussion category is now the designated area to share useful technologies, demonstrations, and other ideas that could be brought into the HAAS project.

Clarifying use of Issues vs "Issues and Bugs"

Please use the Issues tab here explicitly for BOUNTIES (as defined by the prime collaborators) or actual issues with the code in this repo. External issues, such as problems with OpenAI or other platforms, should be discussed here: https://github.com/daveshap/OpenAI_Agent_Swarm/discussions/categories/issues-bugs