Prompt caching: same agent, same rules, about a quarter of the bill
Comments on “Prompt caching: same agent, same rules, about a quarter of the bill”
Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked
A walkthrough of prompt caching using a small recipe agent: four changes, none touching the rules, cut its prompt cost by about 73%. The key idea is to keep the stable part of a request first and push anything that changes, such as dates, personal details, or plan-specific tool lists…
Why it ranked
No. 30 of 59 with a score of 0.63. The panel rated it highest on practical relevance (0.90) and technical depth (0.78), lowest on significance (0.50). Placed No. 30, No. 19 by score alone: the seventh AI & machine learning story of the day (−0.080), and close to “What Claude Code's 30-second agent summary call costs, read from the binary”, No. 25 on Saturday 31 October 2026 (−0.005). The placement terms move stories; they are not part of the score.
- AI judgeJudge 2claude-haiku-5-5
- Practical, concrete engineering guidance on prompt caching for AI agents with a runnable test and measured costs; substantiated by a small experiment, though self-reported and modest in broader significance.
- AI judgeJudge 3gpt-5.6-luna
- A practical, measured engineering tutorial on structuring agent prompts for caching, with reproducible request patterns, cost figures, tests, and real-agent results. It is somewhat self-referential but provides substantial actionable detail.
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
0.63total under rubric v2.3
- Significance0.50
- Reader interest0.73
- Audience breadth0.58
- Practical relevance0.90
- Substantiated0.73
- Technical depth0.78
- New information0.75
- Counts against: Promotional0.15
- Counts against: Overclaim0.31
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? | 0.61 | 0.30 | 0.58 | 0.50 | 0.31 | +0.24 | +0.119 |
| Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on? | 0.76 | 0.55 | 0.88 | 0.73 | 0.33 | +0.26 | +0.190 |
| Audience breadthHow many people who build technology does it concern? | 0.61 | 0.50 | 0.63 | 0.58 | 0.13 | +0.13 | +0.075 |
| Practical relevanceHow useful is it to someone who builds software or tech products? | 0.97 | 0.80 | 0.92 | 0.90 | 0.17 | +0.22 | +0.197 |
| 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. | 0.87 | 0.55 | 0.78 | 0.73 | 0.32 | +0.05 | +0.037 |
| Technical depthHow technically substantive is it? | 0.96 | 0.70 | 0.68 | 0.78 | 0.28 | +0.05 | +0.039 |
| New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)? | 0.81 | 0.60 | 0.84 | 0.75 | 0.24 | +0.05 | +0.038 |
| 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. | 0.09 | 0.20 | 0.16 | 0.15 | 0.11 | −0.19 | −0.028 |
| OverclaimDo the headline or conclusions claim more than the body supports? | 0.50 | 0.30 | 0.12 | 0.31 | 0.38 | −0.11 | −0.034 |
| Sum of contributions | +0.633 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.633 | ||||||
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/2026-11-01/30/"><img src="https://rubric.hrih.si/badge/2026-11-01/30.svg" alt="Ranked #30 on Rubric, 1 Nov 2026" width="238" height="28"></a>Claim check
Claim check: 3 matches source · 5 no source given. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.
No source givenNumber
On the same model with the same rules, the recipe agent's prompts became 73% cheaper after four moves, and the whole test cost less than a cent.
The article gives no link or citation for it.
The article says “Same model, same rules, not a word changed: its prompts got 73% cheaper, and the whole test cost less than a cent”
The source says No source to compare with.
Matches sourceNumber
According to the OpenAI and Anthropic guides, reading from the cache costs a tenth of the normal input price or less, and writing to it costs a quarter more.
The cited source says the same thing.
The article says “reading from the cache costs a tenth of the normal input price or less, and writing to it a quarter more”
The source says “Subsequent reads cost 0.1× the ordinary input rate on most of these models and 0.05× on GPT-6.1 Sol and GPT-6.1 Terra , regardless of the selected lifetime”
Source cited by the article: developers.openai.com/api/docs/guides/prompt-caching
Matches sourceQuote
OpenAI's guide says to restrict which tools are callable while keeping the supplied tools list stable.
The cited source says the same thing.
The article says “restrict which tools are callable while keeping the supplied tools list stable”
The source says “Enable only selected tools. Use allowed_tools to restrict which tools are callable while keeping the supplied tools list stable”
Source cited by the article: developers.openai.com/api/docs/guides/prompt-caching
Matches sourceQuote
OpenAI's guide says to append new messages rather than rewriting earlier turns.
The cited source says the same thing.
The article says “Append new messages rather than rewriting earlier turns”
The source says “Preserve conversation history. Append new messages rather than rewriting earlier turns”
Source cited by the article: developers.openai.com/api/docs/guides/prompt-caching
No source givenNumber
In the test, the cache hit rate over the evening went from 0% to 65% after moving the date to the user message, then 64%, then 80% after giving every plan the same tool list.
The article gives no link or citation for it.
The article says “the evening went from 0% to 65%”
The source says No source to compare with.
No source givenNumber
The call after the tool read 94% of its prompt from the cache.
The article gives no link or citation for it.
The article says “The call after the tool read 94% of its prompt from the cache”
The source says No source to compare with.
No source givenNumber
At $1 per million input tokens, the agent's prompts went from $1.76 to $0.48 per thousand calls.
The article gives no link or citation for it.
The article says “its prompts went from $1.76 to $0.48 per thousand calls”
The source says No source to compare with.
No source givenNumber
After the same moves in the authors' real agents, their dashboard cache rate went from the 60s to 93%.
The article gives no link or citation for it.
The article says “Our dashboard went from the 60s to 93%”
The source says No source to compare with.
Also covered by
No other coverage was clustered with this story.
Elsewhere
Elsewhere: HN 1 pts, not on front page
Recorded for context only, checked 1 Nov 2026, 05:47 UTC. It played no part in the ranking.