Verified against Perplexity Comet · 2026-08-13
Lay two competing accounts of the same story side by side without letting one quietly win by default
A short, direct comparison prompt for when two sources tell meaningfully different versions of the same event — built to keep both accounts intact and visible rather than blending them into whichever one Comet found more persuasive.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Compare these two accounts of the same event side by side. Do not blend them into one combined narrative — I need to see exactly where they overlap and exactly where they diverge, kept clearly separate. EVENT a public dispute between two former co-founders over who originated a product idea ACCOUNT A founder A's blog post timeline of events ACCOUNT B founder B's public statement responding to the blog post WHAT I'M TRYING TO FIGURE OUT whether our company should continue publicly crediting founder A as the sole originator in our own materials Build a two-column comparison covering: what happened (the core sequence of events, and where the two accounts agree versus diverge on it), who is credited or blamed, what evidence or sourcing each account offers for its version (a named witness, a document, an anonymous source, no sourcing given), and what each account leaves out that the other one includes. Where the two accounts simply can't both be accurate, say so plainly rather than implying a middle-ground version that neither source actually supports. Then answer whether our company should continue publicly crediting founder A as the sole originator in our own materials directly, stating which account your answer leans on and why — and naming the specific piece of missing information that, if it surfaced, would most change which account looks more credible.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
The explicit instruction not to blend the two accounts matters because a language model's default behavior when comparing two versions of a story is to produce a synthesized middle account — a version that sounds balanced but that neither original source actually supports, which is a worse outcome than honestly reporting two accounts that disagree, since the synthesized middle can misrepresent both. Keeping the comparison in a strict two-column structure, rather than a flowing paragraph, prevents the model from silently resolving a divergence by choosing which detail to lead with — a structural constraint does work here that an instruction alone can't, because a table format makes an unaddressed cell visible in a way a paragraph doesn't. Asking specifically what evidence or sourcing each account offers — a named witness versus an anonymous source versus no sourcing at all — matters because the two accounts of a personal or reputational dispute often differ far more in evidentiary weight than they do in the plausibility of the story itself, and a plain narrative comparison tends to treat both accounts as equally weighted by default simply because both are being described in the same register. Requiring the final answer to name the specific missing piece of information that would flip the comparison is the mechanism that keeps the whole exercise honest about its own uncertainty — it's an explicit acknowledgment that the comparison, however careful, is being made with an incomplete picture, and states exactly what would close that gap rather than implying the comparison is more settled than it is.
What you get back
Agreement: both accounts agree the product idea was discussed at a specific meeting in early 2024. Divergence: Account A says the idea was fully specced by founder A before the meeting; Account B says it emerged collaboratively during the meeting itself. Sourcing: Account A offers a dated internal document; Account B offers no documentary evidence, only its own recollection. Given the documentary evidence gap, this leans toward Account A's version being more defensible for now — though the specific missing piece that would change this is whether founder B can produce any dated notes or messages predating the meeting.
Verified against
Perplexity Comet 2026 browser agent · 2026-08-13
Changelog
- 2026-08-13 — Initial publish, verified against Perplexity Comet 2026 browser agent.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
