The Most Dangerous Forecast Is the One That Comes True: Three Scenarios in Gemini Notebook
How to make Gemini Notebook build three futures you can verify, instead of one confident guess
A wrong forecast rarely destroys an organization. A right one does it more often.
It sounds like a paradox, but it isn’t. Once a forecast comes true, it earns the status of truth. It becomes the only permitted future. Budgets, plans and promises get built on it. Then the world turns, quietly and without warning. The organization carries on as before, because last time it was right.
The problem is not a mistaken estimate. The problem is the monopoly of a single estimate.
This thought came from Mohamed El-Erian’s piece in The New York Times on September 17, 2026. His thesis is uncomfortable: the US economy can be read three ways at once. As stable, as unstable and as unknown. None of these readings is wrong. Each holds a piece of reality.
El-Erian writes about the economy, but the logic is universal. It applies to publishing, products, teams and careers. And Gemini Notebook, formerly NotebookLM, is an almost ideal tool for turning it into practice.
Three Futures, Not Three Moods
I call the first scenario STABLE. Shocks are temporary; the system absorbs them and returns to its old pattern. The danger is not crisis but complacency. Reserves drain quietly.
The second is UNSTABLE. One weakness feeds another. Costs squeeze margins, margins squeeze investment, investment squeezes quality. Consequences outgrow their causes.
The third is UNKNOWN. Not the numbers change, but the rules. The biggest mistake is applying an old model to a new reality.
These are not optimistic, base and pessimistic cases. Three moods of one model look like three futures but are one future in three shades. Real scenarios differ by mechanism: by what drives the change, not by how strong it is.
The test is simple. If all three scenarios lead to the same decision, you have not built three scenarios. You have built one, three times.
The Notebook Starts With a Decision, Not a Topic
Most people open a notebook and throw in everything they find on the topic. That is the first step toward a single forecast.
Scenario work starts with a short note of four fields: what we are deciding, over what horizon, under which constraints, and when we will reassess. That note is the anchor.
Then come sources, by role rather than volume. One carries facts. Another makes the case for stability, a third for instability, a fourth for changing rules. The fifth is a counterweight, a serious disagreement with what you already believe. Three documents repeating one interpretation are not three pieces of evidence. They are one scenario with an amplifier.
This is where Gemini Notebook shows its greatest value. It answers from your sources, with citations. Every scenario ties to evidence, and every piece of evidence can be checked with one click. A scenario without citations is a red flag.
Before the first analysis, ask the notebook how biased it is:
For each source, state, with a citation, whether it primarily supports
the STABLE, UNSTABLE or UNKNOWN scenario, or is neutral. Then state
which scenario is underrepresented and which is overrepresented.
Do not invent data.If one scenario is overrepresented, the notebook will favor it however cleverly you ask. Fix the sources first, then the prompts.
Get Mihailo Zoin’s stories in your inbox
Join Medium for free to get updates from this writer.
The main prompt asks, for each scenario, for the assumption, evidence for and against, early signals and a trigger for changing the decision. With three prohibitions: do not pick the most likely scenario without evidence, do not assign percentages, do not invent data. A percentage without a basis is not precision. It is decoration.
Then comes the test most people skip:
For each pair of scenarios, name at least one evidential difference:
a data point from the sources that would look different in one than
in the other. If there is none, say so openly.The model happily writes three nice stories. This prompt makes it admit when they are the same story.
When the Analysis Becomes One Document
Scenario work used to be scattered. Analysis in chat, a table in one output, a mind map in another. Only the author’s head held it together.
The newest Studio panel capability in Gemini Notebook changes that. Under Reports there is an Interactive format, a report with embedded Studio content. The default template, Learning Overview, summarizes key concepts. But the template has a pencil, and a pencil means the report follows your rules.
Into the template I write a structure: the decision, facts separated from assumptions, three scenarios with evidence for and against, the matrix, robust moves and an uncertainty register. I also ask for embedding suggestions: a mind map, a data table, a signal-board infographic and a team quiz. I choose them with the Add button.
The AI Usage indicator shows what a generation will consume, and Generate later postpones a large report. According to the official help, the feature is web-only for now, and the report should be reviewed because it may contain inaccuracies.
That is not a footnote. An interactive report looks convincing even when it is wrong. Polished form amplifies the authority of content. So citations matter more here than in chat, not less.
Order matters: mind map first to check structure, data table second to check evidence, interactive report last.
A Scenario Without a Trigger Is Just an Essay
You can have three perfect scenarios and never change a decision. That happens when nobody says in advance what they would need to see to change course.
So every scenario gets a trigger: a threshold, a data source and a person with authority to act. The notebook then becomes a living system. Every new source raises one question: which scenario does it strengthen, and which does it weaken? On the reassessment date, I do not edit the old report. I create a new one and compare them. The difference is the most valuable information.
An example from publishing. In the stable scenario, readers buy professional books in the familiar pattern. In the unstable one, sales weaken as price, attention and free AI content converge. In the unknown one, the function of the book itself changes: it becomes a source, a course, a protocol, a working tool for AI. The first two call for better sales. The third calls for a different product.
One rule stays with me from all this. The less known the future, the more reversible the decision should be, unless waiting costs more than being wrong.
El-Erian does not claim to know which economy will prove real. That is his strength. He does not ask what will happen, but which futures we have to survive.
That question belongs in every notebook in Gemini Notebook: “Which different futures do we have to survive, and what evidence will show us that one of them is emerging right now?”
A forecast gives you an answer. Scenarios give you attention. When the rules change faster than the data, attention is worth more.
The complete guide, with all prompts and templates, is in the Gemini Notebook Mastery collection.
