AI helps construct and interrogate models but does not silently define reality. It can translate a qualitative description into candidate variables and causal relationships, search literature for plausible parameter ranges, propose competing model structures and explain the behaviour of a simulation. Separate critic agents look for omitted actors, impossible conservation assumptions and conclusions that depend on one fragile parameter. The system keeps proposed structure, accepted structure and measured evidence visually distinct. Users can ask imaginative questions—what would have to become true for this region to exist?—and receive a chain of model changes rather than a fictional answer.
PROJECT 014 · Research platform
Atlas of the Possible
“Can mathematics reveal futures that imagination alone would never find?”
MOVE TO EXPLORETHE IDEA IN ONE LINEgap
01 / PREMISE
The beginning
Atlas of the Possible is a visual instrument for exploring complex systems as landscapes rather than forecasts. When I think about a city, a company, an energy network or a social problem, I do not want one predicted future. I want to see the shape of possibility: which regions are reachable, which constraints create cliffs, which small intervention opens a new path and which apparently attractive future collapses when one assumption changes. The Atlas would combine mathematics, simulation and generative AI to let people travel through possible worlds while keeping every world connected to explicit rules.
Most decision tools force complexity into a spreadsheet, dashboard or linear scenario. These formats show outputs but hide topology: decision-makers cannot see that two outcomes are separated by a fragile threshold, that several strategies converge on the same failure or that a neglected variable creates an entirely new region of possibility. Forecasting then rewards a single confident line through a space nobody has mapped. Pure simulation can generate millions of outcomes but leave humans unable to understand them. Pure storytelling is understandable but may ignore physical, financial or causal constraints. We need an interface between rigorous model space and human imagination.
02 / THE PRODUCT
What it could become
The Atlas is a navigable world-model studio. A team describes a system—its actors, resources, relationships, rules, uncertainties and goals—using data, diagrams and ordinary language. The platform builds an explicit computational model, runs large families of scenarios and projects the results into an explorable visual landscape. Peaks may represent resilient outcomes, valleys may reveal lock-in, bridges may show narrow transition paths and fog indicates insufficient knowledge. A person can move a constraint, introduce an intervention or ask an AI guide why two regions differ. Every visual feature opens back into equations, assumptions and source data. The product is simultaneously map, simulator, conversation object and experiment generator.
For whom
- Scientists and researchers exploring high-dimensional systems
- Strategy teams making decisions under structural uncertainty
- Cities and public institutions testing policy combinations
- Venture builders searching for newly possible companies
- Educators helping people develop systems intuition
Core capabilities
- Natural-language and diagram-to-model translation
- Agent-based, causal and dynamical-system simulation
- High-dimensional possibility-space mapping
- Interactive sensitivity, threshold and path analysis
- AI explanation tied to equations, assumptions and evidence
The project becomes meaningful only when a new technical possibility is translated into a clear human advantage, an experience people can understand, and a system capable of earning trust over time.
03 / FOUNDATIONS
Intelligence and mathematics
The heart of the project is the geometry of possibility. Dynamical systems describe how states evolve; causal graphs distinguish intervention from observation; agent-based models represent heterogeneous behaviour; constraint programming eliminates impossible worlds; Monte Carlo and uncertainty quantification populate the reachable space. Manifold learning and topology reveal the lower-dimensional structures hidden inside millions of simulations. Bifurcation analysis finds thresholds where behaviour changes qualitatively. Multi-objective Pareto surfaces show futures that trade one value against another without collapsing them into a single score. Path planning then asks not only which state is desirable, but whether a plausible sequence of actions can reach it.
04 / VENTURE LOGIC
How it might live
I see the Atlas first as a research platform and later as a company around consequential exploration. Universities and public-interest researchers could use an open core for published models. Enterprises, cities and venture studios would pay for secure collaborative workspaces, computational scale, domain-specific model libraries and expert-supported modelling programmes. A marketplace could allow researchers to publish inspectable system modules without turning models into opaque consulting reports. The strongest defensibility would come from the interaction language, evaluation methods and accumulated library of verified model components—not from keeping mathematics secret.
For me, a venture is more than an interesting technology. It needs a narrow first user, a repeated problem, a distribution path, a credible advantage and a reason to improve as more people use it. I would test those conditions before deciding whether this idea should become a company, a product, an open technology or an ongoing research programme.
05 / DESIGN PRINCIPLES
Rules for making it real
- 01
A map of possibilities is not a prediction.
- 02
Every visual shape must open back into assumptions and mathematics.
- 03
Impossible worlds should be excluded explicitly, not aesthetically.
- 04
Show uncertainty as territory, not a footnote.
- 05
Keep competing models alive when evidence cannot choose between them.
- 06
The best output is often the next experiment, not the preferred scenario.
06 / ROADMAP
From question to company
- 01Frame
Choose one bounded system whose behaviour is complex but measurable.
- 02Prototype
Build an end-to-end model-to-landscape prototype with transparent equations.
- 03Prove
Test whether domain experts can discover non-obvious thresholds and intervention paths.
- 04Build
Add competing model structures and explicit uncertainty geography.
- 05Advance
Develop collaborative annotation, versioning and experiment generation.
- 06Advance
Expand through verified domain modules rather than one universal world model.
07 / HONEST QUESTIONS
What could go wrong
Serious imagination includes the possibility that an idea should change radically—or should not exist. These are the tensions the project would need to resolve:
- Beautiful landscapes creating more confidence than the underlying model deserves.
- Dimensionality reduction hiding rare but catastrophic outcomes.
- Users treating generated causal structure as discovered truth.
- Political or organisational values being smuggled into technical constraints.
- Computational expense favouring elaborate models over useful simple ones.
- The metaphor of navigation suggesting control over systems that remain fundamentally unpredictable.