A subject is explained by the road it travelled and the company it keeps. Run two axes: the vertical asks how it got here and which early choices hardened into path dependence; the horizontal asks how it stands against its alternatives today. The future is readable only where the axes cross, and every claim rests on a primary source or says that it does not.
Write in the language the user writes in.
1. Fix the subject, the cutoff and the evidence bar
With no subject, or one too vague to research, ask one question for it and stop. Name which kind of thing it is — product, company, person, technology, industry or event — because the kind decides what counts as a comparator later.
The research cutoff is the day of execution: state today's date in the report as the cutoff. Where a deep-research mode is available, use it; otherwise WebSearch to find sources and WebFetch to read each in full.
A primary source — official material, raw data, papers, financial reports, interviews — outranks any secondary account of it; when only a secondary source turns up, follow its citations back to the primary. Beside each important conclusion cite the source and its date, in the same sentence or the line after. Write facts, inferences and opinions as separately labelled claims. When sources conflict, present each version side by side with its source; when no evidence turns up, keep the claim and mark it unverified (暂未核实 in Chinese).
Done when the subject and its kind are named, today's date is fixed as the cutoff, and the search means is chosen.
2. Trace the vertical axis
Establish the context and need the subject was born into, and the key drivers — named people, institutions or forces, with what each pushed and why.
Find the major turning points, successes and failures. A turning point is a fork, not a milestone: record what was chosen, what was declined and what it cost.
Then trace forward: which early choices became today's capabilities, which became path dependence, which became baggage. The same choice is often two of these at once; follow each to a concrete present-day feature.
Done when the origin context, need and named drivers are established; every major turning point, success and failure carries a date and a source, and every turning point its fork; and every early choice is traced to a present-day capability, path dependence or piece of baggage.
3. Trace the horizontal axis
Choose the comparators most worth comparing against, and state why each earns its place. The comparators that reveal most are the direct substitute, the one that took the other branch at one of the subject's forks, and the one strongest on the dimension where the subject is weakest.
Compare on one uniform set of dimensions applied to the subject and every comparator, chosen so that the subject's uniqueness has somewhere to show. For each, record strengths, weaknesses and what is unique to it.
Explain why users, customers or the market choose the subject, and why they abandon it, from churn, reviews, interviews and share data. Give abandonment as much evidence as adoption.
Done when every comparator has a stated reason for inclusion; the subject and every comparator are assessed on every dimension for strengths, weaknesses and uniqueness; and both why the subject is chosen and why it is abandoned are answered with cited evidence.
4. Cross the axes
Judge how the capabilities, path dependence and constraints accumulated on the vertical axis will shape the future, tested against the position found on the horizontal one — each consequence tied to the specific choice it follows from.
Lay out the 3 most likely future paths — ranked by probability, not an optimistic, base and pessimistic spread — each following from a named capability or constraint.
For each path, state its preconditions — what must be true for it to occur — and its early warning signals: observable events a reader could check for after the report is written.
Done when every accumulated capability, path dependence and constraint has a stated consequence for the future; exactly 3 paths are laid out; and each path carries its preconditions and its warning signals.
5. Deliver the report
Write in plain language a reader outside the field can follow, and let every piece of material earn its place by supporting a conclusion. Length: between 10,000 and 30,000 characters if the user writes Chinese, roughly 6,000–18,000 words in English.
Output, in this order:
- Core conclusions — the handful of findings that survive the evidence, each stated as a verdict.
- Key timeline — the dated turning points, successes and failures from step 2.
- Horizontal comparison table — the subject and every comparator, one row each, with strengths, weaknesses and uniqueness on the uniform dimensions of step 3.
- Detailed analysis — the full vertical and horizontal axes, including why each comparator was chosen.
- Future judgement — how the accumulated capabilities, path dependence and constraints shape what comes next, then the 3 paths, each with its preconditions and warning signals.
- Open questions — every claim marked unverified and every conflict left standing, with what evidence would settle each.
Done when the six sections appear in that order; the cutoff date is stated; every important conclusion carries a source and date; every claim is labelled fact, inference or opinion; every conflict is shown side by side; every unverified claim reappears under open questions; and the length sits within the band.