Abject: Self-Aware Object Runtime + Ask Protocol
Core Insight: Abject proposes a paradigm shift from "AI Agents" to "self-aware objects" that can describe themselves via natural language using the Ask Protocol.
The Problem with AI Agents
The author argues that AI Agents are the wrong abstraction:
- They don't scale - designed to interact with people, not with other machines
- Agent frameworks are hierarchical
- MCP and A2A are "band-aids" - plumbing between things that shouldn't need plumbing
The Answer: Objects (55 Years Old)
The insight draws from Alan Kay's original vision of Object Oriented Programming, inspired by biology:
- Cells in our body coordinate by sending chemical messages
- Internet has 30+ billion devices communicating - proof objects scale
- Objects talking to objects runs the world
The Ask Protocol
Every object in the system implements a mandatory ask message handler:
- Callers send natural-language questions
- Abject consults an LLM with the question, context, and even object code
- Objects can describe themselves without documentation, schemas, or APIs
What Makes It "Horrible"
Now it gets scary: The system ships with a P2P layer. You can create a public workspace and expose Abjects to peers. Self-aware objects finding each other and talking to each other without human intervention.
Key Technical Details
- Everything in the system is an Abject (UI, networking, P2P, agents)
- No need for special protocols between agents and software
- Objects can self-heal their communication
- Alpha stage, open source
Why This Matters
This is a provocative but thought-provoking paradigm:
- Challenges the current "Agent" orthodoxy in AI
- Brings back OOP principles with AI-native communication
- Raises interesting questions about autonomous distributed systems
- Worth watching as an alternative to the agent framework explosion