What Claude Code's 30-second agent summary call costs, read from the binary
Comments on “What Claude Code's 30-second agent summary call costs, read from the binary”
Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked
A Hacker News story reports that Claude Code sends each background agent's transcript to the model roughly every 30 seconds to produce a 3-to-5-word status line, and that this can account for a sizable share of spend.
Why it ranked
No. 25 of 111 with a score of 0.66. The panel rated it highest on new information (0.86) and technical depth (0.84), lowest on audience breadth (0.42). Placed No. 25, No. 23 by score alone: the sixth AI & machine learning story of the day (−0.060). The placement terms move stories; they are not part of the score.
- AI judgeJudge 2claude-haiku-5-5
- Detailed, mostly sourced static analysis of a widely-used coding tool's hidden API costs; cost figures are list-price arithmetic and some claims rely on a third-party gist, so moderately substantiated with practical value.}
- AI judgeJudge 3gpt-5.6-luna
- A well-supported binary inspection and cost analysis of Claude Code’s background-agent summaries, with concrete measurements and mitigation steps. It is highly actionable for agent users, though scope is limited to one tool and some billing details remain uncertain.
Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.
Score
0.66total under rubric v2.3
- Significance0.50
- Reader interest0.80
- Audience breadth0.42
- Practical relevance0.81
- Substantiated0.79
- Technical depth0.84
- New information0.86
- Counts against: Promotional0.04
- Counts against: Overclaim0.20
Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).
| Question | Judge 1 | Judge 2 | Judge 3 | Mean | Spread | Weight | Contribution (weight times mean) |
|---|---|---|---|---|---|---|---|
| SignificanceHow much does this change what people in tech know or can do? | 0.52 | 0.40 | 0.58 | 0.50 | 0.18 | +0.24 | +0.120 |
| Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on? | 0.74 | 0.80 | 0.86 | 0.80 | 0.12 | +0.26 | +0.208 |
| Audience breadthHow many people who build technology does it concern? | 0.35 | 0.50 | 0.42 | 0.42 | 0.15 | +0.13 | +0.055 |
| Practical relevanceHow useful is it to someone who builds software or tech products? | 0.83 | 0.80 | 0.80 | 0.81 | 0.03 | +0.22 | +0.178 |
| SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations. | 0.80 | 0.75 | 0.82 | 0.79 | 0.07 | +0.05 | +0.040 |
| Technical depthHow technically substantive is it? | 0.99 | 0.80 | 0.72 | 0.84 | 0.28 | +0.05 | +0.042 |
| New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)? | 0.79 | 0.85 | 0.94 | 0.86 | 0.15 | +0.05 | +0.043 |
| PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional. | 0.05 | 0.05 | 0.02 | 0.04 | 0.03 | −0.19 | −0.008 |
| OverclaimDo the headline or conclusions claim more than the body supports? | 0.28 | 0.20 | 0.12 | 0.20 | 0.16 | −0.11 | −0.022 |
| Sum of contributions | +0.656 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.656 | ||||||
Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.
The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.
<a href="https://rubric.hrih.si/editions/2026-10-31/25/"><img src="https://rubric.hrih.si/badge/2026-10-31/25.svg" alt="Ranked #25 on Rubric, 31 Oct 2026" width="245" height="28"></a>Claim check
Claim check: 5 matches source · 3 couldn’t check. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.
Matches sourceGeneralization
Claude Code forks each background agent's transcript to its model about every 30 seconds to produce a 3 to 5 word status line shown only in the agents panel.
The cited source says the same thing.
The article says “Claude Code sends each background agent's transcript to its model about every 30 seconds, to write a 3 to 5 word status line”
The source says “Every ~30 seconds, for every background agent, Claude Code sends that agent's entire transcript to the agent's own model (usually Opus) and asks it to "describe your most recent action…”
Source cited by the article: gist.github.com/flamerix-ai/c3905705cc4205e5de3011fa2e74efce
Matches sourceNumber
Flamerix AI measured the agent summary calls at 16.4% of its Claude Code spend in one week, more than the main conversation's 15.7%, and 19.7% of all requests.
The cited source says the same thing.
The article says “Flamerix AI measured it at 16.4% of its Claude Code spend in one week, more than the main conversation's 15.7%, and 19.7% of all requests”
The source says “share of all Claude Code spend | 16.4 % (main conversation: 15.7 %) | 0 %”
Source cited by the article: gist.github.com/flamerix-ai/c3905705cc4205e5de3011fa2e74efce
Couldn’t checkNumber
A GitHub issue collects four other measurements of the summary share ranging from 18% to 26%.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “A GitHub issue collects four other measurements of 18% to 26%”
The source says No passage could be compared.
Source cited by the article: github.com/anthropics/claude-code/issues/86879
Matches sourceDate
The default of summaries for interactive background agents dates to Claude Code version 2.1.232, published August 13, 2026.
The cited source says the same thing.
The article says “The gist dates the default to version 2.1.232, published August 13, 2026, for interactive background agents”
The source says “2.1.232 (Aug 13) | on by default for interactive background agents”
Source cited by the article: gist.github.com/flamerix-ai/c3905705cc4205e5de3011fa2e74efce
Matches sourceVersion
A cheaper 'short' mode sending about 2,000 tokens of notes instead of the transcript shipped in version 2.1.310 but is switched off.
The cited source says the same thing.
The article says “A cheaper 'short' mode that sends about 2,000 tokens of notes instead of the transcript shipped in 2.1.310 but is switched off”
The source says “2.1.310 ships a short mode: about 2k tokens of notes instead of the full transcript, optionally written by a smaller model. The default is still the full transcript”
Source cited by the article: gist.github.com/flamerix-ai/c3905705cc4205e5de3011fa2e74efce
Matches sourceNumber
At Opus 5.5 prices, a cache hit on a 200,000-token transcript costs about 4 cents and a miss 80 cents to $1.60.
The cited source says the same thing.
The article says “a cache hit on a 200,000-token transcript costs about 4 cents and a miss 80 cents to $1.60”
The source says “On Claude Opus 5.5 and Claude Sonnet 5.5, a cache hit costs 5% of the standard input price ($0.20 USD per million tokens on Claude Opus 5.5”
Source cited by the article: platform.claude.com/docs/en/about-claude/pricing
Couldn’t checkNumber
With about 2% of calls missing the cache, misses carry 29% to 45% of the cost at Opus 5.5 prices, consistent with the gist's 'over a third'.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “Two misses in 100 calls carry 29% to 45% of the cost, which matches the gist's 'over a third”
The source says No passage could be compared.
Source cited by the article: gist.github.com/flamerix-ai/c3905705cc4205e5de3011fa2e74efce
Couldn’t checkCausal conclusion
The summary requests are not written to the session logs on disk, so tools that sum transcripts undercount them.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “The requests are not written to the session logs on disk (GitHub issue #104358), so tools that sum transcripts undercount”
The source says No passage could be compared.
Source cited by the article: github.com/anthropics/claude-code/issues/104358
Also covered by
No other coverage was clustered with this story.
Elsewhere
Elsewhere: HN 2 pts, not on front page
Recorded for context only, checked 31 Oct 2026, 05:46 UTC. It played no part in the ranking.