China governs as an engineered system. We read its outputs the same way.
Forward Indicators is an independent think tank at the intersection of geopolitics, AI, technology, and business.
We publish analysis, essays, and six active research streams — each a public, dated, falsifiable inquiry into systems most institutions prefer to treat as a black box.
Can a model predict state behavior?
Six research streams run the same experiment on six different corpora: a calibrated model trained on the published record of a state apparatus — or the firms and capital flows around it — scored against what actually happens. The instrument is the same. What changes is the question we point it at.
Almost nothing in AI and tech policy comes out of nowhere. Almost everything was on a timetable.
Most institutions analyze AI policy, tech policy, and geopolitical risk as if each one were a black box. They read the press release when an export control drops; they infer intent from the rhetoric; they call the consequences a surprise. That frame is convenient for the people selling it — and a methodological dead end. It treats every sanction as an act of will, every restriction as a shock. It rewards access over evidence. It produces narratives that are unfalsifiable by design.
Forward Indicators reads the specifications. Doctrine, regulation, procurement schedules, subsidy cascades, talent restrictions, sectoral controls — these are the source trail of modern AI and tech power, published and dated. Read them as code, and the surprises become legible.
Read the full manifestoFour commitments we make to our readers.
Predictions are dated.
Every claim about future state behavior is logged with a date, a confidence, and a falsification criterion. The log is append-only. No retroactive edits.
The prior is public.
System prompts, few-shot examples, feature definitions — versioned and viewable. You can read why the model says what it says.
Baselines stay shipped.
Random, last-observed, and feature-only logistic regression run alongside the LLM model. Lift over baseline is shown on every scoreboard.
Calibration over accuracy.
A 70% prediction must be right 70% of the time, not 90%, not 50%. Reliability diagrams are the headline chart, not point accuracy.
We work with the institutions building this future.
Forward Indicators collaborates with universities, research institutions, governments, foundations, and businesses on commissioned analysis, custom research programs, and structured policy work.
If your team needs a public, dated, calibrated read on a policy or technology question — or wants to fund or co-author a program — write in. We work with a small number of partners at a time.
Inquire about a collaboration