NVIDIA Nemotron 3.5 Prompt Generator

You have access to four engineering specifications: [Document A — hardware architecture], [Document B — API reference], [Document C — performance benchmarks], [Document D — known issues log]. Synthesize these into a single executive summary that: (1) maps each hardware capability in Document A to its API endpoint in Document B; (2) flags any benchmark result in Document C that falls below the threshold implied by Document A's specifications; (3) cross-references every known issue in Document D against the relevant API endpoints in Document B. Output as a structured table with columns: Capability, API Endpoint, Benchmark Result, Status, Known Issues. Flag any gap where a Document A capability has no API endpoint.

I am sharing a product photograph of a consumer electronics device alongside its printed specification sheet. From the photograph, identify: (a) the physical port types visible on the device; (b) the approximate screen diagonal estimated from the chassis proportions; (c) any visible branding, model number, or regulatory marks. From the specification sheet, extract: (a) the declared port specifications; (b) the official screen size; (c) the regulatory certifications listed. Then compare both sets: flag every discrepancy where the photograph suggests a different configuration than the specification sheet states. Rate your confidence for each finding on a 1–5 scale.

I am providing a system architecture diagram [image] and a Python microservices codebase [attached files]. Perform a joint analysis: (1) from the diagram, extract the declared service dependencies — which services call which, and through what interface type (REST, gRPC, message queue); (2) from the codebase, extract the actual service calls and their interface implementations; (3) produce a gap analysis: list every declared dependency in the diagram that has no corresponding implementation in the code, every implementation that has no corresponding declaration in the diagram, and every interface type mismatch. Present results as a three-column table: Declared, Implemented, Status.

Analyze the attached research paper [PDF]. Perform the following steps in order: (1) Extract the paper's central claim as a single testable hypothesis. (2) Identify the methodology: what experiment or dataset was used, what the control conditions were, and what the measured outcome variable is. (3) List every quantitative result in the paper with its associated confidence interval or p-value. (4) Identify the three strongest supporting results and the two results that most weaken the central claim. (5) Compare the methodology against the stated limitations section: identify any limitation that is not adequately addressed by the experimental design. (6) Suggest one follow-up experiment that would resolve the largest remaining uncertainty. Cite every finding with paragraph and page number.

I am providing a video of a technical product demonstration [video file]. Perform the following analysis: (1) Produce a timestamped transcript of all spoken content. (2) At each product interaction shown on screen — button press, menu selection, output display — add a bracketed annotation to the transcript noting what was demonstrated and what the screen showed at that moment. (3) After the transcript, produce a feature matrix: list every product feature demonstrated, the timestamp where it was shown, whether it matched the product's official documentation, and a one-sentence assessment of how clearly it was demonstrated. Flag any feature shown in the video that contradicts documented specifications.

Review the attached contract [document]. Flag every clause that creates: (1) unlimited liability exposure — cite the clause text and explain why it is unlimited; (2) unilateral termination rights for the counterparty without cause — cite the clause text and the notice period; (3) automatic renewal provisions that could bind the party beyond the initial term — cite clause and renewal conditions; (4) intellectual property assignment clauses broader than work-for-hire scope — cite clause and explain what IP is transferred; (5) jurisdiction or governing law provisions that require dispute resolution in an inconvenient or unfavourable forum — cite clause and forum. For each flagged item, suggest a one-sentence redline that mitigates the risk. Format as a risk register with columns: Risk Type, Clause Reference, Clause Text, Risk Level (High/Medium/Low), Suggested Redline.

I am providing three consecutive annual reports [Documents 1, 2, 3] and a peer-group financial benchmark dataset [Document 4]. Perform the following: (1) Extract revenue, gross margin, operating margin, R&D spend, and capex from each annual report. (2) Calculate year-over-year change for each metric. (3) Compare each metric against the peer-group median in Document 4. (4) Flag any metric where: the YoY change exceeds ±20%; the value is more than 1.5 standard deviations from the peer median; or there is a directional inconsistency (e.g., revenue growing while gross margin declining sharply). (5) For each flag, provide a one-paragraph hypothesis for the underlying cause, citing specific sections of the annual reports that may explain it.

I am providing five photographs of the same industrial facility taken at six-month intervals over two years [Images 1–5]. Perform a change-detection analysis: (1) For each consecutive image pair (1→2, 2→3, 3→4, 4→5), list every structural or equipment change visible — additions, removals, modifications. (2) Classify each change as: infrastructure expansion, equipment upgrade, maintenance activity, or degradation indicator. (3) Identify any change that follows a pattern across multiple intervals (e.g., seasonal cycle, progressive degradation). (4) Produce a timeline of the facility's evolution with the confidence level for each identified change. (5) Flag any change that appears inconsistent with normal facility operation and may warrant inspection.

Process the attached dataset of 500 customer feedback records [CSV file with columns: id, text, rating, date]. For each record: (1) Classify the primary issue type into one of: Product Quality, Delivery, Customer Service, Pricing, UX/UI, Other. (2) Identify the sentiment polarity: Positive, Neutral, Negative, Mixed. (3) Extract the specific product feature or process step mentioned, if any. (4) Flag records where the star rating contradicts the text sentiment (e.g., 5-star rating with negative language). After processing all records: produce a summary table of issue type frequency, average rating per issue type, and the top three most-mentioned specific features. Export as a structured JSON array with one object per record plus a summary key at the end.

[For research/educational use only] I am providing a de-identified chest X-ray image. Perform the following analysis steps: (1) Describe the overall image quality — exposure level, patient positioning, any motion artifact. (2) Describe the cardiac silhouette — size relative to thoracic width, border sharpness. (3) Describe the lung fields — any opacity, consolidation, or lucency in each zone (upper, middle, lower, left/right). (4) Describe the costophrenic angles and diaphragm contour. (5) Note any visible bone abnormalities. (6) Provide a structured summary of findings in the format used for radiology reporting, clearly marked as AI-assisted analysis for research purposes only, not a clinical diagnosis.

I am providing a legacy codebase [attached files] and a target architecture diagram [image]. Plan the migration: (1) From the codebase, extract all external dependencies with their current versions. (2) From the architecture diagram, identify every service and data store in the target state. (3) Map each legacy component to its target-state equivalent. (4) Identify dependencies with no equivalent in the target architecture — these are migration blockers. (5) Sequence the migration into phases: each phase should be independently deployable without breaking the current system. (6) For each phase, list the specific files to be changed, the new dependencies introduced, and the integration tests that must pass before the phase is complete. Output as a Gantt-style phase plan.

I am providing a contract originally written in German [Document 1], its official English translation [Document 2], and a clause-level commentary in French from legal counsel [Document 3]. Perform the following: (1) Identify every clause in Document 1 that does not have a corresponding clause in Document 2 — cite the clause number and provide a direct translation. (2) Identify every place where the English translation in Document 2 uses a term with a different legal meaning than the German original — cite the term, its German source, and the translation risk. (3) Cross-reference the French commentary in Document 3 against both documents: list every concern raised in the commentary that corresponds to a specific clause, and whether that concern is present in both the German and English versions.

I am providing a supplier directory [Document 1], a geographic risk index [Document 2], and three years of supplier performance records [Document 3]. Build a supply chain risk matrix: (1) For each tier-1 supplier in Document 1, look up their country of origin in Document 2 and extract their performance metrics from Document 3. (2) Score each supplier on three dimensions: geopolitical risk (from Document 2), delivery reliability (from Document 3), and quality incident rate (from Document 3). (3) Calculate a composite risk score. (4) Flag any supplier where a single dimension score is critically high regardless of composite score. (5) Identify geographic concentration risk: any country supplying more than 20% of a critical component category. Output as a ranked risk register.

I am providing a complex scientific diagram from a materials science paper [image — a phase diagram or crystal structure visualization]. Perform a structured interpretation: (1) Identify the type of diagram and what variables are represented on each axis. (2) Identify every phase region or structure visible and label it using standard materials science nomenclature. (3) Identify any critical points, phase boundaries, or invariant points. (4) Explain in plain language what a materials engineer would do with this diagram in practice — what does it tell them about how to process this material? (5) Note any ambiguity in the diagram labelling or any region where the boundaries are not clearly defined. Cite each observation by region of the diagram (e.g., 'upper-left quadrant', 'transition zone at approximately X temperature').

I am providing the transcript of a quarterly earnings call [document]. Perform the following extraction and analysis: (1) List every forward-looking statement made by management, with the exact quote and the page/timestamp. (2) For each forward-looking statement, extract: the implied timeline, the specific metric or outcome projected, and any caveat or qualifier used. (3) Compare each projection against the previous quarter's projections (if mentioned in the transcript). (4) Identify every question from analysts that management did not directly answer — provide the question and a characterisation of the deflection strategy used. (5) Extract every mention of competitive dynamics, market conditions, or regulatory environment. (6) Produce a sentiment timeline: chart whether the tone of management language shifted from the opening statement to the Q&A, citing specific language shifts.

I am providing screenshots of a mobile application's onboarding flow [Images 1–6, one per screen]. Perform an accessibility audit against WCAG 2.2 AA standards: (1) For each screen, check and report: contrast ratio of text against background (estimate from the screenshot); presence and quality of touch target sizing (minimum 44×44 pt); whether interactive elements are distinguishable from non-interactive ones by means other than colour alone; whether any information is conveyed by colour alone without a secondary indicator. (2) List every WCAG 2.2 AA criterion that appears to be violated, with the specific screen number and element. (3) For each violation, suggest the minimal design change that would resolve it. (4) Estimate the effort to fix each violation as Low/Medium/High. Present as an audit report with a pass/fail summary table per screen.

I am providing five patent documents [Documents 1–5] in the same technical domain. Perform a landscape analysis: (1) For each patent, extract: the independent claims (the broadest protections), the priority date, the assignee, and the primary technical problem solved. (2) Map the claim scope of all five patents: identify where claims overlap and where there are potential freedom-to-operate gaps. (3) Identify the prior art cited in common across multiple patents — these are the foundational references in this technology space. (4) Flag any dependent claims in Patents 2–5 that appear to be designed specifically to work around the independent claims of Patent 1. (5) Produce a brief competitive landscape summary: which assignee appears to hold the strongest foundational IP position based on these five documents, and why.

I am providing a product listing image [image] and 50 customer reviews from the product's listing page [text document]. Perform the following cross-validation: (1) From the image, list every visible product feature, material quality indicator, and dimension estimate. (2) From the reviews, extract every mention of a specific product feature, material quality observation, or dimension-related comment. (3) Produce a cross-validation matrix: for each feature visible in the image, count the number of reviews that confirm it, the number that contradict it, and the number that are silent on it. (4) Identify any feature where the image implies a higher quality than the review consensus supports. (5) Flag any review that describes a product significantly different from what the image shows — these may indicate listing mismatches, quality control issues, or fraudulent reviews.

I am providing press releases, earnings transcripts, and product announcements from three competitors [Documents 1, 2, 3 — approximately 50 pages each]. Produce a competitive intelligence brief: (1) Extract every product launch or update announced in the last 12 months, with launch date and key capability. (2) Extract every pricing change or new pricing tier. (3) Extract every partnership, acquisition, or distribution agreement. (4) Extract every customer win, case study, or reference customer mentioned. (5) Identify the three most significant strategic moves by each competitor and assess the strategic intent behind each. (6) Identify any white space in the competitive landscape — product categories, customer segments, or geographies that none of the three competitors appear to be targeting aggressively. Produce as an executive brief with a three-company competitive positioning matrix.

I am providing a system incident report [Document 1 — timeline of events with log extracts], an architecture diagram of the affected system [image], and post-incident comments from three engineering teams [Document 2]. Perform root cause analysis: (1) Reconstruct the incident timeline from Document 1, identifying the exact moment of first failure and each subsequent cascading failure. (2) Map each failure event to the corresponding system component in the architecture diagram. (3) Identify whether each failure was a direct cause, a contributing cause, or a consequence. (4) Synthesise the post-incident team comments in Document 2: identify where teams agree, where they disagree, and what each team may be omitting (the absence of blame is itself a signal). (5) Produce a five-whys root cause analysis. (6) List the three highest-impact preventive actions ranked by implementation effort.