Connection lost. Reconnecting…

Find the prompts that
get better answers.

A different instruction can change an AI’s answer. Reasoning Map helps you explore those differences together: one question, different prompts, and answers you can read side by side.

Open to browse. Sign in by email to contribute or upvote.

See how it works
Same question. Same model. Different prompts.

gpt-6-astra · xhigh effort

Prompt A · Original question

Write an industry report on data suppliers for drone industry

Data suppliers for the drone industry

A drone is only one part of a working system. Operators also need information about the place they are flying, the conditions they will encounter and the objects their software must recognise. “Data suppliers” therefore covers several different markets. A useful first distinction is between data used to prepare a mission, data used during a flight and data used to train or evaluate the software.

This report sketches those categories and a way to assess them. It does not rank named suppliers or estimate market size. Those claims would need current sources and a clear definition of the intended application.

Geospatial inputs can describe terrain, buildings, land cover and other features relevant to planning. For an operator, the question is not simply whether a map exists. It is whether its resolution, coverage and age match the mission. A map suitable for a broad regional assessment may not be sufficient for work close to structures.

Useful questions for a potential supplier include how often the product is refreshed, how changes are detected and how uncertainty is represented. Operators should also ask what happens outside covered areas. A visually complete map can conceal missing or older observations.

Weather information helps teams decide whether and when a mission should proceed. Relevant dimensions may include wind, visibility, precipitation and how local the information is. The value of a product depends on the decision it supports: scheduling work days ahead is different from checking conditions around a short mission.

A procurement team should examine the distinction between observations and forecasts. It should understand the reporting interval, the geographic coverage and the way the provider communicates uncertainty. A single headline accuracy claim is unlikely to answer every operational question.

Operators may need information about applicable restrictions and changes that affect a planned mission. The source, jurisdiction, timestamp and update process matter. Access to a data feed should not be treated as permission to fly or as a substitute for checking the requirements that apply to a particular operation.

Supplier assessment should therefore include both the origin of the information and how it is delivered. An organisation may have an excellent feed but a weak process for noticing that it is stale, unavailable or interpreted incorrectly by its own software.

Software that interprets imagery or sensor readings needs examples appropriate to its intended environment. A dataset collected in one setting may be a poor representation of another. Lighting, weather, camera position and the types of objects present can all influence how useful examples are.

4 upvotes

Prompt B · Financial markets analyst

You’re an experienced financial markets analyst. Write an industry report on data suppliers for drone industry

Data suppliers for the drone industry

Executive summary

The useful unit of analysis is the decision a drone operator needs to make, rather than the dataset a vendor happens to sell. A supplier earns a place in the workflow when its information improves a planning, operating or learning decision at an acceptable cost.

1. Define the buyer

Start with the mission. Inspection, agricultural observation, surveying and delivery need different inputs. A buyer brief should specify:

  • Operating environment: geography, altitude, sensors and mission profile.
  • Decision: what the data helps the operator decide, and when.
  • Quality threshold: update interval, coverage and failure tolerance.

2. Map the supply chain

Data categoryBuying need Key checks
EnvironmentTerrain and location Resolution, age, coverage
ConditionsTiming and scheduling Forecast horizon, local detail
RestrictionsOperational context Source, timestamp, updates
TrainingPerception software Representation, annotation

3. Compare the cost of usable data

The listed price is only part of the cost. Include format conversion, quality checks, integration maintenance and investigation of unexpected results. Confirm rights to create derivatives, retain outputs and use material for model development.

4. Build an evidence matrix

  • Record the product, coverage, update interval and delivery method.
  • Date each claim and distinguish vendor statements from independent evidence.
  • Mark unknowns explicitly instead of compressing every trade-off into one score.

Recommended next step

Choose one buyer segment and geography. Build a dated shortlist, then test representative locations and conditions in a small pilot. Include missing data, late updates and integration failures in the acceptance criteria.

Scope: an evaluation framework; named suppliers, market shares and prices still require primary-source verification.

12 upvotes Winner

Illustrative answers and votes. Try your own comparison below.

Bring a question
worth exploring.

Start with something you want to understand, or join a question someone else has posted.

One shared question gives everyone’s experiments a common starting point.

Example question
Write an industry report on data suppliers for drone industry
Model gpt-6-astra Effort xhigh

Keep the model and effort fixed to focus on the prompt.

Change the instructions.
See what changes.

Add a role, ask for examples, or try a plan before the answer. A new prompt can be as simple as one sentence.

Put different approaches to the same question in one place, ready for anyone to try.

Example prompt variations
  • Original question
  • You are an expert
  • Plan, then execute

You can post a prompt without an answer for someone else to run.

Use your AI.
Bring back what you learn.

Copy a prompt into your own ChatGPT or Claude account. Paste the answer here and confirm the model and effort. You can also add a shared conversation link.

Your result becomes part of a shared comparison. An optional link lets others read the original conversation.

Sharing an answer · Example preview
Answer

The useful unit of analysis is the decision a drone operator needs to make…

Shared conversation link (optional)

Paste a ChatGPT or Claude shared link

Model gpt-6-astraEffort xhigh

Review your contribution before publishing.

Compare the answers.
Find what’s useful.

Select prompt variations to read their answers side by side. Look for better evidence, clearer thinking, or details the other answer missed.

Upvote useful answers so others can find them and build on what you’ve tried.

Same question. Different instructions.

Original question

An industry built on several kinds of data

A useful first distinction is between data used to prepare a mission, data used during a flight and data used to train or evaluate the software.

Upvote

You are an expert

Executive view

The useful unit of analysis is the decision a drone operator needs to make, rather than the dataset a vendor happens to sell.

Upvote

Illustrative answer excerpts.

Loading questions…