Team Running 100-200 Agents Daily Hits Disk Space Bottleneck

A team running 100-200 agents daily found the scaling bottleneck has shifted from tokens and compute to disk space, with power predicted as the next constraint. A workaround using APFS clones to share pnpm dependencies helps but agents' use of /tmp limits its effectiveness.

2026-09-12 ~ 2026-09-12 · 3 related posts

• Running 100-200 agents daily: disk space is now the bottleneck, not compute — vincent_koc · 2026-09-12
• Thread part 2: scaling agents beyond 100-200 daily is blocked by scratch space — vincent_koc · 2026-09-12
• Thread part 3: APFS clones for pnpm deps as a workaround for agent disk hunger — vincent_koc · 2026-09-12