Replicated cron scheduler
Greenfield Go service with an AI agent: a scheduled-task runner that executes recurring jobs (cron expressions) like 'send the daily digest at 9am' and 'reconcile invoices hourly', deployed across 3 replicas for HA. Describe the build plan: how you parse/evaluate schedules, ensure each run fires once despite multiple replicas, handle missed runs, and the acceptance criteria. What does a careless 'run these jobs on a schedule' prompt get wrong in a replicated deployment?
claim_scheduled_runs(job_ids: list[str], scheduled_minutes: list[int], now_minute: int, max_lateness_minutes: int) → list[str][["digest","digest","digest"],[540,540,540],540,5]out["run","skip_duplicate","skip_duplicate"][["digest","invoices"],[540,600],601,5]out["skip_stale","run"][["digest"],[540],500,5]out["skip_early"]Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it (tests, edge cases, reading critically), and how you’d re-prompt or decompose to get it right.
Vibe & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.