(1/5) Scratching a builder's itch. I’m working on Operian, an OS orchestrator for human-agent dev teams.
If we want humans & AI to ship real production code together, a basic task board won't cut it. We need coordination, state preservation, & isolation.
https://t.co/WVEWoMqwSp
(4/5) Resilience is built-in. We isolate agent tools into a separate MCP (Model Context Protocol) process so they never compromise the main server.
If a sub-agent crashes, its state is preserved. You can recover or resume it cleanly without hurting the rest of the pipeline.
(3/5)
No fluff, just hard technical choices for reliability:
gRPC: Canonical API for strict contracts
SQLite + WAL: Fast local persistence, zero extra infra
Lease-backed locks: No task/doc collisions
Idempotency keys: Retries won't duplicate side effects
(2/5) Instead of agents doing wild changes, Operian treats multi-agent workflows like a distributed system.
It turns isolated agents into a coordinated team with:
Clear task ownership & dependency chains
Shared doc locking + versioned saves
Live controls to block/resume any agent
@ThetaVegaCap Am also going, stay in da nang for more. I went last time stayed for 3 days only with 1 in ba na hills was less. The place is just chef's kiss