Human-authored judgement
AI accelerates clustering and correlation. Every strategic conclusion — every Mill Alpha — is written and signed by an analyst. No exceptions.
RESEARCH & UPDATES
An intelligence house that hides its methodology is asking to be taken on faith. We publish ours — the metrics, the validation discipline, and the record, including the misses. Credibility compounds; imagination does not.
METHODOLOGY
Every reading in the HKM line is built from four disclosed metrics: Sentiment (direction and intensity of the conversation), Dominant Emotion (the specific feeling carrying it), Volume Score (share of conversation, normalised against the post pool), and Growth Velocity (the week-on-week burst factor). Read together, they separate a passing spike from a narrative that is genuinely forming.
AI accelerates clustering and correlation. Every strategic conclusion — every Mill Alpha — is written and signed by an analyst. No exceptions.
Where we publish probability readings, we publish how well calibrated similar readings have historically been — hits and misses alike.
If the data says the relationship is weak, the report says so. A null result is still a decision input — and the reason clients trust the positive ones.
FLAGSHIP PUBLICATIONS
Five to eight globally resonant themes, read through the Conviction Gap lens — where public sentiment and capital-backed probability diverge. Executive summary free; full note for institutional subscribers. Methodology demonstrated in the open, validation record included.
Request the Current IssueA recurring public reading of Hong Kong's commercial mood, built on the four-metric methodology across the city's platforms and forums — designed as a reference artefact for operators, media, and institutions. Methodology note to be published at launch.
Join the Notification ListSAMPLE BRIEFS
A redacted standing brief: trend table, sentiment clusters, hook phrases, and a signed Mill Alpha.
Share of Narrative radar, parity gaps, and audience migration — anonymised from live category work.
The free executive layer of the current monthly issue.
Samples are released on request to qualified counterparties. Client work is confidential; samples are redacted or synthetic.
Request AccessALGO TRADING LAB RESEARCH
Lab notes are published as they conclude: walk-forward validation records of strategies that survived, documented null results of those that did not, and regime studies underpinning both. Where probability readings are published, their historical calibration is published alongside — hits and misses together.
Request Lab Research Access