A multi-agent orchestration system that runs AI agents across several machines around the clock. OpenClaw, Codex, Python, cron, and SSH coordinate distributed work, validation, and machine sync.
An AI agent fleet designed and operated solo by Raehyeon (Neo) Jeong. Autonomous agents on several machines split collection, processing, and validation work. cron and launchd start scheduled jobs, while SSH and shared storage carry state between machines. Agents handle work that needs model judgment; deterministic code handles repeatable transformations and checks. A real browser verifies the final state before unattended work is accepted.
An AI agent fleet is a setup for multiple specialized agents distributed across processes or machines, with shared scheduling, state, and validation rules. This fleet splits collection, processing, and verification work across several machines and runs it around the clock.
cron and launchd start scheduled jobs and assign role-specific work. SSH synchronization and shared storage carry state between machines. Deterministic pipelines and a real browser then verify the outputs.
Model judgment and deterministic checks are separate. Agents handle unstructured decisions; scripts validate formats and state; a real browser checks the final screen. The system does not rely on one model response as proof that a task succeeded.
Several machines can process recurring collection, transformation, and validation work in parallel. The founder can spend more time on product decisions and exceptions. This fleet supports real data, content, and operations workflows used by the teams Raehyeon works with.
The stack includes OpenClaw, GPT-5.5 Codex, Python, launchd and cron, SSH, and Supabase. Agent execution, scheduling, machine synchronization, shared state, and result validation stay in separate layers.
No. It also handles data collection, transformation, validation, real-browser checks, machine synchronization, and unattended backend operations.
</>openclaw · gpt-5.5 codex · multi-machine · cron