Commence with the End in Mind

None of these three finished artefacts should be produced here - a summary of text you hold, a sonnet, and a report restricted to your own five data points are all things Perplexity is built against. What it does uniquely well is interrogate the target before you build it: whether the interview's claims are still current, what the sonnet form actually requires according to scholarship, and whether the market-leadership framing your brief assumes is a claim the world supports at all.

• Interrogate the target before building it elsewhere.
• Never let it near a report restricted to your own data.
• Use it to check whether a claim your brief assumes is supportable.
• Academic focus for form, scholarship and evidence.
• Ask whether your source material has been superseded.

Summarising a transcript you already hold is not a Perplexity job - it will search alongside your text and colour the summary with things the scientist never said. Do the five bullets in another tool. What Perplexity can do, and the others cannot, is check the interview against the current state of the field before you circulate a summary of it.

I have an interview with a leading climate scientist covering rising sea levels, carbon capture technology, and policy recommendations. Before I summarise and circulate it, research these with sources and dates:

1. What is the current state of published evidence and projection on sea level rise? Use Academic focus. Cite the major assessments and give the date of each, so I can tell whether an interview from some months ago is still current.
2. What is the published evidence on carbon capture technology at scale - what is actually operating, at what capacity, at what cost? Cite it, and separate demonstrated performance from announced capacity. This is the area where interview claims date fastest.
3. Where is there genuine scientific disagreement on these topics, as opposed to manufactured controversy? Cite the disagreement in the literature. If my scientist is stating a contested position as settled, I need to know.
4. What climate policy has actually changed recently that a policy-recommendation section might now be behind? Cite with dates.
5. What is specifically documented about sea level rise projections for Indian coastal cities - Mumbai, Chennai, Visakhapatnam? Cite the sources. Indian readers will ask, and a general interview may not cover it.

Cite everything with dates. Flag anything you can only source to a single place.

Then summarise the transcript in another tool, using this as the check on what to flag as possibly dated rather than as material to add.

Do not write the sonnet here. Perplexity is an answer engine with citations - it will not hold iambic pentameter, and a sonnet with footnotes is not a sonnet. Write it in Claude, ChatGPT or Gemini.

There is a genuinely useful question to ask here first, though, and it is the one most people skip: what the form actually requires, from sources rather than from a model's impression of it.

1. What are the formal rules of the Shakespearean sonnet as authoritative sources state them - rhyme scheme, metre, the function of the volta, the role of the closing couplet? Cite scholarly or reference sources, not writing blogs. Use Academic focus.
2. Where in Shakespeare's own sonnets does he depart from strict iambic pentameter, and what do scholars say about those departures? Cite the criticism. I want to know how much licence the form actually allows before I hold a poem to a rule Shakespeare himself broke.
3. What early modern English vocabulary existed for optical and visual instruments - the perspective glass, the mirror, the eye as instrument? Cite period or scholarly sources. This is the practical problem: I need period-plausible words for a camera, and inventing them is how the poem goes wrong.
4. How have poets and critics handled the sonnet written about modern technology? Cite examples of the form used for contemporary subjects, and what reviewers said about whether it works.
5. What do critics identify as the tells of a bad pastiche of early modern English? Cite the criticism - I would rather know the failure modes in advance.

Cite everything. Then take answers 1, 3 and 5 into a writing tool as the brief, which is a much better brief than "act as Shakespeare".

The report itself should not be written here - but this example is worth pausing on, because Perplexity would actively damage it. Give it a company name and figures and it will search, find organisations with similar names, and fold their details into your report. A fabricated fact with a citation attached is harder to catch than one without. Write the report in another tool, on your data only.

Where Perplexity genuinely helps is the question your data cannot answer. You have 50,000 units against one competitor's 45,000, and the brief wants a report about market leadership. The data does not support that claim. Find out what would:

1. What is actually reported about the market this company competes in - total size, number of players, the largest participants' shares? Cite what you find with dates. If I cannot establish total market size, "market leadership" is not a claim I can make at all, and that is the finding.
2. How do business publications and financial journalists define and qualify "market leader"? Is there a standard threshold, or is it rhetorical? Cite style guides or journalism sources if any address it.
3. What are the rules in India on unsubstantiated superiority claims in published material - ASCI guidance, or anything from SEBI if this is a listed company's reported figures? Cite the rule.
4. What do credible sources say about how quarter-over-quarter growth should be presented to avoid misleading readers - seasonality, single-quarter comparisons, the base effect? A 20% rise against one prior quarter can be seasonal rather than structural. Cite what you find.
5. Is 20% quarter-over-quarter growth notable in this sector, or ordinary? Cite comparable reported figures. Without that, "strong growth" is an adjective rather than an assessment.

Cite everything, with dates, and flag single-source claims. Then write the 150-word report elsewhere, from my five data points and nothing else, using answer 1 to decide whether the leadership framing is available to me at all.