Hey @mohi-devhub — really like what you're building here. The framing of "learn what AI writes, not just accept it" lands perfectly.
I've been thinking about the adjacent problem: even when you understand AI-generated code, it still becomes invisible in the codebase over time. No ownership, no discoverability, no context graph. Nate B Jones wrote about this as "dark code" — AI output that works but nobody can navigate or maintain.
AntiVibe tackles the understanding gap. I've been building Kernal — an MCP-based knowledge graph that tackles the discoverability + ownership gap. It auto-extracts entities and relationships from conversations, meetings, and code context into a navigable graph.
I think there's a natural integration point:
- AntiVibe generates a deep dive → explains what code does and why
- Kernal ingests the deep dive → maps it into a knowledge graph (who wrote it, what it connects to, what patterns it uses, which decisions led to it)
- Result: AI-generated code becomes searchable, navigable, and owned — not just explained once and forgotten
Concretely, this could look like:
- AntiVibe's output markdown gets structured with entities (functions, patterns, decisions) that Kernal can extract
- A Kernal MCP tool call after each deep dive to persist the knowledge graph
- Queries like "show me all code that uses JWT auth pattern" or "who decided to use refresh token rotation and why" become possible
Would love to hear your thoughts. Happy to demo Kernal if you're curious — it runs as a local MCP server with SQLite, zero cloud dependency.
— Cesar (@pintomatic)
Hey @mohi-devhub — really like what you're building here. The framing of "learn what AI writes, not just accept it" lands perfectly.
I've been thinking about the adjacent problem: even when you understand AI-generated code, it still becomes invisible in the codebase over time. No ownership, no discoverability, no context graph. Nate B Jones wrote about this as "dark code" — AI output that works but nobody can navigate or maintain.
AntiVibe tackles the understanding gap. I've been building Kernal — an MCP-based knowledge graph that tackles the discoverability + ownership gap. It auto-extracts entities and relationships from conversations, meetings, and code context into a navigable graph.
I think there's a natural integration point:
Concretely, this could look like:
Would love to hear your thoughts. Happy to demo Kernal if you're curious — it runs as a local MCP server with SQLite, zero cloud dependency.
— Cesar (@pintomatic)