Four long-running projects across crypto markets and macro. What they share is not a subject but a method: conclusions are published in falsifiable form, every figure carries a source grade, and the criteria are registered in advance and never revised afterwards.
Estimates for the next bull-market high of major coins, published as a range plus a skew direction — pre-registered, with falsification conditions, never revised after they trigger. Plus a Bitcoin cycle chart, daily net flow across the top 10 exchanges by on-chain reserves, and a six-dimension sentiment panel for nine assets.
Direct on-chain reads of the published BTC / ETH / Tron addresses of major exchanges, set against each venue's own proof-of-reserves page; aggregator caliber rules are stated, and every difference is flagged by one uniform reading with its verification method. Every number is reproducible with the repository scripts.
In what form does the debt get resolved after $40 trillion. It pins down the measures (total debt ≠ debt held by the public ≠ marketable), takes the mechanism apart (the decomposition of 1946-1980 financial repression, and which part its modern replica is missing), and pre-registers the criteria — six triggers with fixed adjudication sources, never revised after they fire.
Turning "is AI a bubble?" into checkable numbers: verified data on the 2000 dot-com bubble (31 months of staged declines, peak valuations, IPO-mania signatures, recovery base rates), a layered 2026 check-up of the AI trade, the three temperatures of one theme across A-shares, Hong Kong and the US, and a dashboard of nine leading indicators — scenario probabilities declared once and never revised, triggers read every edition.