Anthropic's Claude Code has changed how background subagents write their status updates, reducing the usage cost of each update by about 99%, according to an announcement from the official Claude developer team.
Claude Code · Background subagents
Status updates for background subagents now cost about 99% less per refresh
Instead of re-reading the full conversation every 30 seconds, a lighter model now writes each status line from short notes, cutting the usage cost of each update by roughly 99%.
99%
Reduction in usage cost per status update
30s
Interval at which the status line refreshes
Haiku 5.5
Smaller, faster model that now writes the status line
Relative cost per status update
Previous approach indexed to 100; the reported 99% reduction leaves about 1.
100
Previous: full history re-read
1
Now: short notes via Haiku 5.5
How each status refresh works
Subagent runs in the background
→
Short notes capture progress
→
Haiku 5.5 writes status line
→
Status shown to user
↺ Repeats every 30 seconds, so the savings compound across each active subagent.
What improves
Lower overhead for each progress refresh
More usage quota left for main work
Most relevant for several subagents or long-open sessions
What remains unclear
No before-and-after usage figures for typical workloads
Real savings depend on subagent count and how long they run
How status notes are maintained and how accurate they are versus the earlier approach is not specified
Independent verification is not yet available
Takeaway
Routing routine status reporting to a lighter model is one way vendors are making long-running background tasks affordable.
Users should watch for further detail on how the feature behaves across plans and workloads.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...