Deepscient

Use when a campaign is running and work must be dispatched, retried or throttled — when a shard sits in running with no heartbeat, when a lease expired and the result was duplicated, when retries storm the source APIs, when budget or concurrency must be cut mid-campaign, or when someone wants to fix a stuck task by editing state directly.

Use when a research goal is too large for one agent turn and must be split into schedulable work — when planning a campaign across S0–S11, when deciding how fine a shard should be (literature, statistics, figures, writing), when setting gates and vote counts so parallel results can be trusted, or when a batch plan produced duplicates, idle workers or unusable outputs.

Use when several agents or votes produced different answers and the result must be merged into one defensible statement — when running the three-vote rule, when weighting preprints against peer-reviewed evidence, when escalating an unresolved disagreement, or when writing the evidence table and gap list that a campaign report must carry.

Use when a research question is still a question rather than a plan — when you know the field but not which study design the field accepts, how many samples it demands, which measurements count as evidence, or which software is standard; also use when a review came back saying the design cannot answer the question, when you are switching fields or advising a student in an unfamiliar one, when two supervisors disagree about method, or when you need to know what reviewers in that field attack first before you commit months to a protocol.

Use when turning an approved plan into an actual run — when the design and sample size are settled and someone needs to produce data, execute code, or reproduce a result; when a run must be recordable (inputs, outputs, environment, seed) so that six months later the numbers in the paper can be traced back; when a result cannot be reproduced and the difference has to be explained rather than waved away; also use when deciding where raw data, analysis code and generated artifacts belong.

Use when a figure has to survive peer review and print — when choosing a figure type for a comparison, a distribution, a paired design or a survival curve; when a journal demands specific column widths, minimum type size, vector formats or resolutions; when error bars must state what they are; when a figure came back from review as "unreadable" or "insufficient resolution"; also use when deciding whether a bar chart is the wrong choice.

Use when you want to turn a fast-moving implementation from GitHub — a new optimizer, a training trick, a scaling recipe — into a publishable contribution, and you need to find which of them is actually worth reading, extract what it really claims, test whether the field has already filled that slot, and decide what evidence a band-3/4/5 venue would demand; also use when a repo reports a number but no mechanism and you need to know what is missing before writing.

Use when you have a DOI/PMID/arXiv id and need the actual PDF or JATS full text, when a download returns HTML instead of a PDF, or when building the methods section and you need page-level quotes; covers the nine-stage fallback chain, how to verify a real PDF, and what provenance to record.

Use when you need to know how a journal actually behaves rather than what its aims-and-scope page claims — its real subfield mix over the last two years, first-round review duration and comment difficulty from authors, Chinese-author review speed and accept-rate wording; also use when a manuscript was rejected for "scope" and you need evidence about which journals genuinely publish that kind of work.

Use when choosing where to submit, when a paper was desk-rejected without review, when deciding between a top journal and a solid sub-journal, or when planning the submission chain; covers the three-tier strategy, why scope beats prestige, format red lines, and how to use transfer and portable-review routes.

Use when starting a literature review, checking whether an idea is already published, or updating a topic search; covers query construction, multi-source coverage, deduplication, inclusion criteria, and what must be recorded so the search is reproducible.

Use when a manuscript draft needs to become a submittable file — you have the content (from science_paper_plan or your own draft) and now need it in the target journal's actual layout; also use when a submission was desk-rejected for formatting (word count over, too many figures, uncited references, wrong reference style), when the same paper must be retargeted from one journal to another without rewriting, or when a submission system demands DOCX or JATS rather than PDF.

Use when deciding who should do a piece of research work, when a task has no owner or two owners, when a role's authority is unclear (who may approve a design, who may veto an analysis), or when a disagreement must be escalated; covers the 13-role RACI table across S0–S11, the six authority types, the veto bypass to the PI, and the escalation path from shard worker to PI.

Use immediately after receiving reviewer reports, when a revision decision says major revision, or when you disagree with a reviewer; covers the numbered-response structure, how to split fixable from arguable points, why unprincipled compliance backfires, and how to escalate a genuine disagreement to the editor.

Use when preparing a submission, when a reviewer asks for CONSORT/PRISMA/STROBE/ARRIVE/TRIPOD items, when the journal demands data/code availability statements, or when an audit reports blocking findings; covers which standard applies to which design and what each one actually requires.

Use before reporting any experiment result, when a result cannot be re-run, when writing the methods section, or when preparing a reproducibility package; covers run provenance, environment capture, seed handling, artifact hashing, and the clean-room re-run check.

Use when choosing which retrieval source to query, when a search returns suspiciously few results, when OpenAlex budget runs low, or when a source is rate-limited or blocked; covers per-source capability, quotas, fallback order, and how to tell "source broke" from "query too narrow".

Use when you have data and must decide which test to run — when someone asks "该用什么检验", when a reviewer says the analysis does not match the design, when assumptions are violated and you need the honest fallback instead of the convenient one, when the effect size or correction method is missing from the results, or when a p-value is being asked to carry a claim it cannot carry; also use before data collection to size the study, and when interpreting a result that is significant but tiny.

Use when a statistical result has to leave the analysis tool and enter a manuscript — when writing the statistics paragraph, the table notes, or the figure caption that carries a p-value; when someone asks how to report a test, an effect size, or a confidence interval; when a result is significant but the assumption check failed; when a chi-square had Yates correction or not and the value differs between tools; also use when a reviewer asks for the numbers behind a claim.

Use when someone is starting research and does not know the order of operations — a first-year student asking "从哪开始", "这一步要做什么", "我这样算不算做完了"; when a project feels stuck and it is unclear which prerequisite is actually missing; when explaining what a complete study requires end to end; when deciding whether the next move is reading, planning, collecting, analysing or writing. Also use before writing a Methods section, to prove every step actually happened.

Use when drafting or restructuring a manuscript for a high-impact journal, when a paper was rejected for "limited conceptual advance" or "not of broad interest", or when the abstract and title need to carry the claim; covers the C-C-C structure, the five-sentence abstract, figure-first ordering, and the Introduction funnel.