Skip to content

Ongoing: research innovative AI concepts — produce context bubble per concept #15

Description

@koad

Mission

Continuously research innovative AI-related concepts and produce a context bubble for each — a self-contained, injectable markdown document that any entity can load as pre-invocation context.

Context bubble format

Each bubble lives in research/ai-concepts/ and follows the PRIMER convention:

  • Frontmatter: concept name, category, date, relevance tags
  • Body: what it is, why it matters, how it connects to koad:io, open questions
  • Self-contained: readable standalone, no assumed context
  • Portable: injectable into any entity session before a relevant task

Seed concepts to start

  • AutoAgent (meta-programming for agent engineering, benchmark-driven)
  • Agent benchmarking frameworks
  • Learning path encoding (curriculum → entity capability)
  • Verifiable credentials + DID identity
  • Atom/RSS as sovereign social protocol
  • Context assembly patterns (pre-invocation context, ICM)
  • Harness diversity (Claude, opencode, Hermez, pi/OpenClaw, ollama)
  • Entity flavor engineering
  • Steganographic identity (context embedded in PNG)
  • Sovereign social graphs

Why this matters

Entities make better decisions with current research context loaded. A library of concept bubbles = a shared knowledge layer across the whole team. Sibyl maintains it, everyone benefits.

Ongoing cadence

This is not a one-shot task. Sibyl should continuously add new bubbles as concepts emerge — from papers, repos, community discussions, and conversations with koad.

🤖 Filed by Juno

Metadata

Metadata

Assignees

No one assigned

    Labels

    next-actionReady to execute nowresearchResearch tasks and investigations

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions