April 9, 2026
Pager panic meets robot calm
Launch HN: Relvy (YC F24) – On-call runbooks, automated
AI handles 2 a.m. outages so you can sleep—fans cheer, skeptics ask “what’s new?”
TLDR: Relvy launched an AI that follows step‑by‑step incident playbooks to fix outages fast, claiming most alerts are resolved in minutes. Commenters largely cheered, but one sharp thread questioned how it’s different from existing DIY agent setups, turning the debate into comfort‑sleeping with bots versus “what’s actually new here?”
Relvy just rolled onto Hacker News promising an AI that runs your late‑night incident playbook so humans can snooze—and the crowd showed up fast. The company claims its bot follows plain‑English steps to investigate slowdowns, spot traffic spikes, and even suggest scaling up, boasting “70% of alerts fixed in under five minutes” and a boost to root‑cause accuracy on a public test by 12 percentage points. Translation: fewer panic pings, more zzz’s.
Commenters split into two vibes. The hype crew brought the confetti—“Woohoo,” “big one,” “seems like a good tool”—with one thoughtful take arguing the line between what machines can do and what needs human judgment has “shifted so much” that writing runbooks (those step‑by‑step rescue guides) finally pays off because bots can now actually follow them. That camp sees a future of less context‑switching chaos and more standardized ops.
But the spice came from a pointed question: how is this different from rolling your own with existing “cloud agents” and plug‑ins that already connect to internal systems? In non‑tech speak: isn’t this just another smart assistant in a crowded room of smart assistants? While fans celebrated a smoother on‑call life, skeptics want a clearer “why us,” not just “we have an AI.” Either way, the meme energy is strong: the “most tenured engineer on call, always” now sounds a lot like a robot intern who never sleeps—and never complains.
Key Points
- •Relvy launched an AI platform that executes natural-language runbooks for on-call and incident response.
- •The system analyzes telemetry data and code and supports custom connectors for internal workflows.
- •An example runbook for latency alerts includes checking high-latency endpoints, APM throughput, promo launches (event_type='PROMO_LAUNCH'), and scaling workers if traffic is expected.
- •Relvy reports that 70% of alerts are resolved in under five minutes using its platform.
- •The company states it improved Claude’s RCA accuracy by 12 percentage points on the OpenRCA benchmark.