Implementing Grassroots Logic Programs with Multiagent Transition Systems and AI
summary
Claims a deterministic multiagent semantics madGLP implements maGLP by replacing shared variable pairs with local pairs linked by global links and proving correctness via disjoint substitution commutativity that follows from the single-occurrence invariant (one reader, one writer per variable). Tools: formal multiagent transition systems, proofs of safety/liveness, and AI coding (Claude) that generates a Dart GLP VM and bytecode compiler for a workstation implementation and is evolving a smartphone multiagent build; the Dart implementation runs and uncovered type-system bugs fixed by refining the specs.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:07New concurrent logic language
- 2:31Grassroots platforms vision
- 5:44Grassroots formal definition
- 8:14Human science AI programming
- 10:45Abstraction cascade with AI
- 11:48Grassroots logic basics
- 15:06Concurrent logic power
- 16:52Type system for runtime bugs
- 17:36Semantics to code path
- 20:48Multi-agent implementation
- 23:13Current progress summary
- 25:25Q&A
related talks
41mPatterns & Practices for building Multi-Agent Systems
32mEvent-Driven & Multi-Agent Architectures for Gen AI Workflows
56mHerding LLMs: Predictable Outcomes from Unpredictable Agents
31mCompositional Verification of Stigmergic Collective Systems
01h 24mBuilding Multiagent Systems: An Ontology-Driven Approach with SPADE
28mThe Put That Sh*t On Agent: Building Multi-Agent AI Systems That Actually Talk to Each Other*
30mLive Demo – Develop and Deploy a Trustworthy Multi-Agent System
56mDesign Patterns for Multi Agent Systems