Anthropic’s @ClaudeDevs account posted on X at 19:29 UTC on 4 November that “Background subagents in Claude Code now use less of your usage limit”. The post says a running subagent updates its status line every 30 seconds, and that each update used to re-read the subagent’s whole conversation. “Now Haiku 5.5 writes it from short notes instead, so each update costs about 99% less”, it says.
A status line every 30 seconds is 120 updates an hour for each running subagent, by our arithmetic. At about 1% of the old cost each, those 120 updates would cost roughly what 1.2 of the old updates did, if each old update cost the same. That is illustrative arithmetic only: Anthropic gave no token counts.
Business Pill · Usage-Based Pricing
An evergreen Business Engineer explainer on usage-based pricing. It teaches the general idea only and does not discuss Anthropic or Claude Code.
The key insight: Anthropic says each status update from a running Claude Code background subagent now costs about 99% less, because Haiku 5.5 writes it from short notes instead of re-reading the whole conversation; at one update every 30 seconds, that is 120 updates an hour per subagent, by our arithmetic.
Subagents are the helper agents Claude Code starts for side tasks. Its documentation says each subagent “sends its own requests, which count toward the same usage limits as your main conversation”, and that subagents Claude spawns from the main conversation already start in the background by default.
The change targets the status line, not the subagent’s work itself. Before, each 30-second update re-read the subagent’s whole conversation, according to the post. Now a small model, Haiku 5.5, writes the update from short notes. The post does not say which Claude Code version carries the change or whether it applies on every provider.
Each icon stands for the cost of 10 status updates under the old method. Interval and the about-99% per-update saving: @ClaudeDevs (Anthropic) on X, 4 November 2026. The 120 updates an hour and the ~1.2 old-update equivalents are our illustrative arithmetic; Anthropic gave no token counts.
What Was Posted on X
Background subagents in Claude Code now use less of your usage limit.
A running subagent updates its status line every 30 seconds, and each update used to re-read its whole conversation. Now Haiku 5.5 writes it from short notes instead, so each update costs about 99% less.
Anthropic’s @ClaudeDevs post, 19:29 UTC on 4 November. It says a running background subagent updates its status line every 30 seconds, that each update used to re-read its whole conversation, and that Haiku 5.5 now writes it from short notes, so each update costs about 99% less.
What the Changelog Shows
Anthropic’s public Claude Code changelog lists earlier fixes to subagent progress summaries. Version 2.1.128 lists “Fixed sub-agent progress summaries missing the prompt cache”, with a roughly threefold cut in cache creation, by its own note. The latest version we read, 2.1.318, lists a fix for the “Now using usage credits” notice “reappearing about every 30 seconds while a background subagent is running”.
The same changelog lists Haiku 5.5 at $0.10 per million input tokens and $0.50 per million output tokens on the Anthropic API, for prompts up to 100,000 tokens.
Why the Usage Limit Matters
Claude Code’s cost guide, under the heading on why usage climbs in a long session, says every subagent “sends its own requests on top of the main conversation’s”. Its advice for spending less on them is to “choose a smaller model for a subagent”. The post says a smaller model, Haiku 5.5, now writes the status line.
As we read the post, the saving sits in a task users never asked for directly: a status line that refreshed 120 times an hour, by our arithmetic, and re-read the whole conversation each time.
As we read the docs, subagents draw on the same usage limit as the main conversation, so a cheaper status line leaves more of that limit for the work itself; Anthropic gave no token counts, so the size of that share is not known.
@ClaudeDevs (Anthropic) on X
“Now Haiku 5.5 writes it from short notes instead, so each update costs about 99% less”
Three Implications
FOR CLAUDE CODE USERS The post says background subagents now use less of the usage limit; the docs say subagent requests count toward the same limits as the main conversation.
FOR HEAVY SUBAGENT USERS Five background subagents running for an hour would produce 600 status updates, by our illustrative arithmetic, each now about 99% cheaper, according to the post.
FOR READERS OF THE NUMBERS Anthropic gave no token counts and did not say how much of a subagent’s usage the status line took before, so the total saving per session is not known.
As we read it, this change is one small line item on that cost side: Anthropic says a background task that ran 120 times an hour, by our arithmetic, now costs about 99% less per run. The analysis does not discuss this change. For the full argument, read The Subsidized AGI Economy on businessengineer.ai.
Anthropic gave no token counts, so the size of the saving per session is not known; it depends on how long each subagent runs and how large its conversation grows. The post says each status update costs about 99% less; it does not say how much of a subagent’s total usage the status line took before. Our 120-an-hour figure is arithmetic from the stated 30-second interval.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Claude Code’s background subagents refresh their status every 30 seconds, 120 times an hour by our arithmetic. Anthropic says each of those updates now costs about 99% less, because Haiku 5.5 writes it from short notes instead of re-reading the whole conversation.
A note on sourcing. The change, its 30-second interval and the about-99% per-update saving are from the @ClaudeDevs post on X of 4 November 2026; version details are from Anthropic’s public Claude Code changelog and cost details from the Claude Code documentation, read the same day. Anthropic gave no token counts; the hourly figures are our illustrative arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
