@i549 You’ll be able to fan out to up to 10vms, 100gb disk up to 32gb ram per. Unlimited usage of opencode free tier or we work with your Claude/codex/opencode sub. Use it and try to find the limits, I’ll scale more as needed
A production error hit Hive at 01:53. 83 seconds later it was a Linear issue with the cause written in it, and 33 minutes later a fix PR was open. Agents did both.
Then a routine closed the ticket too early, and another routine caught it. Signup to join https://t.co/8zTlTUkEzZ
At 04:08 our PR shepherd moved the issue to Done, because a PR was open.
At 04:30 the triage routine posted on the issue: "The fix wasn't live then, and it still isn't: PR #1316 is open (not merged), even though this issue is marked Done."
@opencode@opencode@thdxr what do y'all use for cloud agents, need a provider? I'd love to make it as easy for opencode users to work on cloud microvm as it is locally. LMK if any interest ♥️
@simonbs I am. That’s exactly what https://t.co/8zTlTUk6Kr is going to deliver. We are in closed beta sign up above. Native harness, your own keys, your own pc or we have microvm sandboxes. We do the things that were missing and won’t reimplement the wheel. 20/mo 10 VMs or unlimited byoc
This is now how I work. If you also work this day, great! I’m assuming you’re more productive that you’ve ever been. If not and you want to work this way signup at https://t.co/8zTlTUkEzZ
It ran on my own Claude subscription, inside Hive: hosted microVMs with Linear, GitHub, CI and schedules wired in, driven from one chat.
https://t.co/jFUcdSwlQZ
Last night one Claude Opus 5.5 session ran our quality push in Hive.
It started 72 agent sessions. By morning 28 PRs had merged: 20 single fixes, then 8 verified batches. +11,068 / -1,740 lines.
How it worked, and the one rule that kept it sane:
It kept itself going. It scheduled its own check-ins, kept its plan and log on a git branch, and sent read-only scout sessions to find the next verified work.
72 sessions: 58 fixers, 8 batch verifiers, 5 scouts, 1 resume host.