Feb 2026 – Present · Kordspace
Karnelian
Founder / Systems Architect
A custom kernel OS — a fully autonomous enterprise agentic AI mainframe.
- Role
- Founder / Systems Architect
- When
- Feb 2026 – Present
- Where
- Kordspace
- Focus
- Systems · AI
01 · The problem
Enterprise AI agents today are glued together in scripts and SaaS dashboards, with no real operating layer for scheduling, isolation, memory or human review.
02 · What I built
Karnelian is a Rust kernel layer for autonomous agents: a scheduler that decomposes work into tasks, sandboxed execution boundaries, shared memory and audit trails, and evaluator/optimizer loops with human-in-the-loop checkpoints baked into the core rather than bolted on.
03 · Outcome
In active design and build since February 2026 — the core architecture, task model and security boundaries are taking shape.
Highlights
- ✦Rust from the kernel up
- ✦Task decomposition + evaluator/optimizer loops
- ✦Sandboxed, auditable agent execution
- ✦Human review as a first-class primitive
What I did
- ✓Designing a Rust-based kernel layer that orchestrates autonomous AI agents for enterprise operations.
- ✓Task decomposition, evaluator/optimizer loops and human-in-the-loop review built into the core.
- ✓Performance-minded, secure service boundaries from the ground up.
Toolbox
RustSystems architectureAgentic workflowsLocal-first tooling
rust all the way down