World Model · snapshot of
A weekly
self-correcting
future world model
It places technological trends, industry signals, and first-principles reasoning onto the same map, describing the future with falsifiable scenarios rather than opinions. Every Saturday, an autonomous programme revises each judgement against that week's evidence, and publicly registers what would prove it wrong.
Top three strongest beliefs right now
Scenarios
Seven future scenarios
Scenarios are the core units of this model: what a local part of the world will look like after multiple trends converge. Each carries a belief strength that is continuously revised by evidence, along with a clearly stated "what would change the model's view". Click any scenario to view.
Trends
Trends driving the world
Each trend is an independent, directional structural judgement. It only describes its current strength and trajectory, without explaining causality. Causality is left for the map in the next section. Click any trend to view its first-principles mechanism, cross-domain analogies, and confirmation / falsification signals.
Couplings & Chains
How they connect
One trend pushes another through specific mechanisms, with time lags, strength, and reverse feedback in between. Connecting these wires yields a map that propagates evidence on its own. The background shading is a contour field calculated from the belief strength of each node, where the denser the lines, the more concentrated the belief. Hover to view neighbours, click a node to read deeply, or follow a causal chain below.
Chains
Five end-to-end causal chains
Chains answer "how far along the entire path has been travelled". Progress is not probability, but the progress of end-to-end realisation.
Snapshot week ·
What moved that week
Revised up
Revised down
Judgements adjudicated by reality
Every judgement has pre-written "if X happens, it means we are right / wrong". When X truly happens, it is recorded here, neither deleted nor modified.
New hypotheses under observation
Candidates not yet in the map. They need to find enough independent evidence within four weeks, otherwise they will be abandoned.
Method
How this model works
It does not predict a certain future, but simultaneously maintains multiple possible futures, and continuously asks: which is becoming more probable, why, and what signals would falsify it.
Probability is belief strength, not frequency
The world will only run once, so "0.76" is not "76% will happen", but the model's current belief strength that the proposition is true. It does not pursue absolute accuracy, but requires comparability: a belief strength of 0.76 must truly be stronger than 0.62.
The weekly revision magnitude is restricted to a very small range, unless a pre-registered event truly happens and triggers an adjudication.
What would prove it wrong
Every judgement has registered falsification conditions, divided into three categories: Class A will be adjudicated by a specific event within a foreseeable time; Class B is a tripwire, which once triggered only weakens and does not strengthen; Class C is a long-cycle open proposition, continuously increasing or decreasing belief strength according to evidence.
How it absorbs the world
Every Saturday morning, an autonomously running programme reads the industry news, research, and signals of the week, finds out which evidence hits existing judgements and which structural trends are not yet in the map, and then revises them one by one. Every modification is written into an append-only evolution log, therefore the full history of any belief can be reconstructed week by week.
What it does not do
It does not look at stock prices, holds no stances, and gives no investment advice. It is isolated in data from any market-perspective systems, and only deduces the future starting from structural changes of the world. Calibration only measures "whether confidence tiers correlate with hit rates", and does not measure profit and loss.