glass.trees

realtime multiagent coordination toy

theoretical premises and conflations (confabulations?)

• interaction as computation as coordination as game as dialogue
• interaction game as proceeding via ATTACKING (questioning) and DEFENDING (answering)
  • proofs and counterproofs
  • inherently dialogical
• interaction as the decomposition and recomposition of possibilities
  • fragmentation as the generation of new questions and answers from those existing
  • integration as resolution of questions (reduction, normalization, abstraction, compression?) and their corresponding answers

theoretical inspirations (poorly co-opted, half-digested spirits)

• reza negarestani: intelligence and spirit
• jonathan ginzburg: the interactive stance
• alain lecomte: figures of dialogue: a view from ludics

rules

• each workspace must begin with a question
• agent can either QUERY or RESPOND to a question
• QUERY means asking a question of the question
• RESPOND means providing a possible answer ("option") to the question
• agent can also QUERY any option
• agent can UPVOTE or DOWNVOTE any query or option
• RESOLVE means that the author of the question is satisfied with the information at hand
• anyone can UN-RESOLVE or REOPEN a resolved question

notes

• there is NO AUTOMATION of resolution - this is entirely up to the author of the question, and can be vetoed by any other agent
  • for ludics fans, this means there is no daimon move: 'winning' means no one vetoes your resolution
  • upvoting and downvoting are simply there to provide additional information on the other agents' preferences

bugs

• MAJOR: 401s due to tokens expiring should trigger a cache renewal of credentials

TODOs and upcoming features

• refactor graph component code, which is a mess
• each user can have many groups
• usernames
• nodes editable by respective author
• ability to start a workspace with an assertion (to be interrogated, vetted etc.)
• UI should be more HUD-like
• avatars
• emojis