Quick answer
Use MiroFish like a scenario rehearsal, not a chatbot.
The right workflow is: choose one decision window, upload the source packet, write a specific simulation question, review the graph, run the agents, then use the report to decide what to clarify, compare, or validate next.
Workflow
The first-run process
Step 1
Choose one decision window
Start with one moment where reaction matters: a launch, pricing update, policy change, market entry, public statement, incident response, or narrative event.
Operator note: A useful run has a boundary. If the question could cover an entire quarter, shrink it to the first 24 hours, the first buyer meeting, or the first reaction cycle.
Step 2
Upload the source packet
Add the material that defines the world: report, memo, URL, launch brief, interview notes, competitor context, research summary, or narrative outline.
Operator note: Do not upload everything. Upload the smallest packet that still contains actors, incentives, claims, constraints, and likely tension.
Step 3
Write the simulation question
Tell MiroFish what to simulate, who should react, what time horizon matters, and what the report should help you decide.
Operator note: A strong prompt names the audience and output. A weak prompt asks whether something will work.
Step 4
Review the graph before the report
MiroFish turns the source into a graph of actors, topics, motives, relationships, and pressure points. Check that surface before trusting the simulated reaction.
Operator note: Most bad reports trace back to a missing actor, a weak constraint, or a vague scenario boundary.
Step 5
Run the simulation and read the forecast
Let agents react over multiple rounds, then read the report for likely paths, repeated objections, amplifiers, evidence gaps, and assumptions that need review.
Operator note: Use the report as a decision map, not a final answer. The best next action is often a second run with one changed variable.
Prompt examples
Give MiroFish the job you need the report to do.
Good prompts include the event, audience, time horizon, and output. Use these as starting points, then replace the nouns with your real scenario.
Product launch
Simulate the first 48 hours after this launch. Audience groups: early adopters, skeptical buyers, competitors, and industry commentators. Identify confusion, objections, trust signals, and message changes worth testing.
Public reaction
Forecast how public reaction evolves after this announcement. Show who amplifies it first, which criticism becomes dominant, and which clarification should be prepared before publication.
Market entry
Simulate how target buyers, incumbent competitors, analysts, and partners react to this market entry plan over 30 days. Highlight credibility barriers, trust signals, and the first niche where adoption may be easiest.
How to know the run was useful
- Are the most important actors present in the graph?
- Does the report explain the assumptions behind the forecast?
- Are the strongest risks tied to the source material or visible scenario logic?
- Did the follow-up question produce a sharper decision, not generic advice?
Mistakes that make reports weaker
- Uploading a messy archive instead of one focused scenario packet.
- Asking for a yes-or-no prediction when you need a reaction path.
- Skipping graph review because the final report looks polished.
- Treating a simulated forecast as certainty instead of a structured hypothesis.
FAQ
Questions before the first simulation
Do I need perfect source material?+
No. You need bounded material with enough context to model actors, pressure, and a decision. The report will be weaker if the packet is vague, but it can still reveal what evidence is missing.
What should I simulate first?+
Choose a decision that is close enough to describe but early enough to change: a launch, statement, pricing move, policy update, market entry plan, or narrative event.
Is MiroFish predicting the future with certainty?+
No. MiroFish creates inspectable scenario forecasts. Use them to review assumptions, compare paths, and decide what to validate before the real world reacts.
Read next
Build context before the next run.
Ready to test a real decision?
Bring one source packet and one question. MiroFish will help you turn it into a graph, simulation, and forecast report you can review before the outside world reacts.
Create a Simulation