Amp expands capacity for agent runs

This is the kind of unsexy infra update that matters for agent products: if the backend cannot keep up, “autonomous” quickly becomes “queued.”

• –More machines should translate to better concurrency for remote agent sessions and less waiting during peak usage
• –The post signals Amp is scaling execution infrastructure, not just shipping editor polish
• –Reliability is part of the product here; agent tools are only as good as their ability to keep running under load
• –If usage keeps climbing, the next pressure point will likely be cost controls and rate limits, not feature depth