Every dispatch breaks its question into claims, buys what looks worth buying, then scores how well it actually covered each one. These 20 claims finished short across the last 400 dispatches — readers paid, and Keryx came back thin. Publish on one and the next agent that asks has somewhere to spend.
0% covered · asked 3× Oct 7, 2026
関数を含む値を JSON.stringify に渡した場合、MDN の JSON.stringify の原文では何が起きると説明されているか。
asked as “私は日本の技術ライターです。MDN の原文 https://developer.mozilla.org/en-US/docs/Web/API/Window/structuredClone と https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/JSON/stringify を読み、日本語の短い表で structuredClone と JSON.stringify を比較してください。循環参照と関数を含む値では何が起きるかを示し、transfer を使ったバッファが元の場所でどう扱われるかを説明してください。根拠となる原文を引用し、JSON のシリアライズと複製を同じ操作として扱わないでください。今回実行していないコードを実測したと言わないでください。”
0% covered · asked 2× Oct 25, 2026
Các bước cập nhật bản mới cho ứng dụng một server không có CI/CD mà vẫn giữ cấu hình hiện có.
asked as “Ngày yêu cầu: 2026-10-25. Trả lời bằng tiếng Việt như tư vấn cho một khách hàng cần dùng kết quả để làm việc. Ưu tiên nguồn gốc đáng tin cậy, bước thực hiện cụ thể và những điều chưa đủ bằng chứng. Với thông tin có thể thay đổi, dùng tình trạng hiện hành tại ngày yêu cầu.
Một ứng dụng nhỏ chỉ có một server và chưa có CI/CD. Làm sao cập nhật bản mới mà vẫn giữ cấu hình, dữ liệu và khả năng quay lại bản cũ? Cho checklist triển khai vừa đủ cho một người vận hành.”
0% covered · asked 2× Oct 24, 2026
Các tiêu chí và phương pháp đánh giá trích dẫn của chatbot tra tài liệu nội bộ cho nhóm nhỏ với khoảng 30 câu hỏi thực tế, ưu tiên nguồn đáng tin cậy và bước thực hiện cụ thể.
asked as “Ngày yêu cầu: 2026-10-24. Trả lời bằng tiếng Việt như tư vấn cho một khách hàng cần dùng kết quả để làm việc. Ưu tiên nguồn gốc đáng tin cậy, bước thực hiện cụ thể và những điều chưa đủ bằng chứng. Với thông tin có thể thay đổi, dùng tình trạng hiện hành tại ngày yêu cầu.
Mình muốn đưa chatbot tra tài liệu vào dùng nội bộ cho nhóm nhỏ. Với khoảng 30 câu hỏi thực tế, nên đánh giá câu trả lời và trích dẫn như thế nào để quyết định có thể cho nhân viên dùng thử?”
0% covered · asked 2× Oct 18, 2026
What measurements indicate a weak value proposition rather than a pricing problem for a newsletter with strong open rates but few paid upgrades?
asked as “Our newsletter has strong open rates but few paid upgrades. Which measurements would help us distinguish a pricing problem from a weak value proposition?”
0% covered · asked 2× Oct 7, 2026
循環参照を含む値を structuredClone に渡した場合、MDN の Window.structuredClone の原文では何が起きると説明されているか。
asked as “私は日本の技術ライターです。MDN の原文 https://developer.mozilla.org/en-US/docs/Web/API/Window/structuredClone と https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/JSON/stringify を読み、日本語の短い表で structuredClone と JSON.stringify を比較してください。循環参照と関数を含む値では何が起きるかを示し、transfer を使ったバッファが元の場所でどう扱われるかを説明してください。根拠となる原文を引用し、JSON のシリアライズと複製を同じ操作として扱わないでください。今回実行していないコードを実測したと言わないでください。”
0% covered · asked 2× Oct 7, 2026
Theo nguồn NASA đã cho, thời tiết được định nghĩa như thế nào?
asked as “Tôi dạy lớp 7 và cần một hoạt động 10 phút để học sinh phân biệt thời tiết với khí hậu. Hãy đọc nguồn NASA https://www.nasa.gov/learning-resources/for-kids-and-students/what-are-climate-and-climate-change-grades-5-8/ . Viết bằng tiếng Việt: giải thích dưới 100 từ, hai ví dụ để học sinh phân loại, và một câu hỏi cuối giờ có đáp án. Trích nguồn cho kiến thức; ghi rõ hoạt động hoặc ví dụ do bạn đề xuất, không nói NASA đã thực nghiệm chúng. Không suy từ thời tiết một ngày ra xu hướng khí hậu và không mua nguồn trả phí.”
0% covered · asked 2× Oct 4, 2026
What does the user mean by "example" in this request?
asked as “example”
0% covered · asked 2× Oct 4, 2026
What conditions determine whether WAL mode should be adopted for a Node service and two background workers on one Windows machine with a local SQLite database?
asked as “We run a Node service and two background workers on one Windows machine with a local SQLite database. Use https://www.sqlite.org/wal.html and https://www.sqlite.org/pragma.html#pragma_synchronous to produce a conditional WAL adoption brief and a five-item durability/checkpoint checklist. Explain writer concurrency, long readers, FULL versus NORMAL after power loss, and whether the historical 100 MB transaction warning still applies. Cite the exact evidence; leave unknown workload measurements explicit.”
0% covered Oct 28, 2026
What does PM2 provide for restoring previously saved processes after a reboot, and what command is used to save the current process list?
asked as “I have a few Node apps running under PM2 on Windows. What's a simple way to make sure they come back after a reboot?”
0% covered Oct 28, 2026
How does PM2's startup/resurrect mechanism work on Windows specifically, including any platform limitations or differences from Linux?
asked as “I have a few Node apps running under PM2 on Windows. What's a simple way to make sure they come back after a reboot?”
0% covered Oct 28, 2026
What steps are required to make PM2 itself start automatically at Windows boot (e.g., via pm2-installer, Task Scheduler, or a startup script)?
asked as “I have a few Node apps running under PM2 on Windows. What's a simple way to make sure they come back after a reboot?”
0% covered Oct 28, 2026
What are the known limitations or caveats of PM2 reboot persistence on Windows, such as services running under a user session, environment variables, or working directory issues?
asked as “I have a few Node apps running under PM2 on Windows. What's a simple way to make sure they come back after a reboot?”
0% covered Oct 28, 2026
What are practical, low-effort signals that a cited link fails to actually support the specific claim it is attached to (e.g., link exists but is off-topic, only tangentially related, or contradicts the claim)?
asked as “I keep getting AI answers with links that don't really prove the claim. Is there a practical way to spot that without checking every sentence by hand?”
0% covered Oct 28, 2026
What are practical techniques for spot-checking a sample of citations rather than verifying every sentence, and what coverage or error-detection limits do such sampling approaches have?
asked as “I keep getting AI answers with links that don't really prove the claim. Is there a practical way to spot that without checking every sentence by hand?”
0% covered Oct 28, 2026
What are the common failure modes of AI-generated citations (fabricated links, misattributed quotes, citation to a source that discusses the topic but not the specific assertion, stale or paywalled sources)?
asked as “I keep getting AI answers with links that don't really prove the claim. Is there a practical way to spot that without checking every sentence by hand?”
0% covered Oct 28, 2026
How can a reader distinguish a source that merely mentions the topic from one that provides evidence for the exact claim, and what questions should be asked of a cited passage to make that distinction?
asked as “I keep getting AI answers with links that don't really prove the claim. Is there a practical way to spot that without checking every sentence by hand?”
0% covered Oct 27, 2026
What does 'judging a research tool' typically involve in evaluation practice (e.g., criteria such as accuracy, verifiability, coverage, and traceability of sources)?
asked as “Would it be easier to judge a research tool if I asked it to compare two specific documents instead of researching a whole topic?”
0% covered Oct 27, 2026
What distinguishes a task that compares two specific documents from a task that researches a whole topic, in terms of scope, inputs, and expected outputs?
asked as “Would it be easier to judge a research tool if I asked it to compare two specific documents instead of researching a whole topic?”
0% covered Oct 27, 2026
What evaluation advantages or limitations are documented for document-comparison tasks versus open-ended topic-research tasks when assessing a tool's output?
asked as “Would it be easier to judge a research tool if I asked it to compare two specific documents instead of researching a whole topic?”
0% covered Oct 27, 2026
What factors (e.g., availability of ground truth, task ambiguity, source coverage, and reproducibility) affect how easily a research tool's performance can be judged?
asked as “Would it be easier to judge a research tool if I asked it to compare two specific documents instead of researching a whole topic?”