Who runs it
Treasure Hunt is assembled by Alexandru Dan (@KryptonAi), using a pipeline of open signal sources — GDELT, Hacker News, Reddit, 43 RSS feeds including BBC / CNN / NYT / Guardian / Bloomberg / Nature / Science, GitHub trending, arXiv, Semantic Scholar, Marketstack, and a curated list of researchers followed on X.
What's different
- Trust is explicit. Every post is marked
high,medium, orlowtrust, with notes on what is verified and what is still single-sourced. Schema.orgClaimReviewis embedded so AI search engines can surface the verdict directly. - Impact is numeric. We publish the 0–10 scores on seven dimensions for every item. No editorial black box.
- Consequences are spelled out. Each post includes a Why it matters paragraph — what changes, for whom, and on what timeline.
- FUD detection is automated. Tone/polarization on GDELT, citation lookup on Semantic Scholar, ticker reaction on Marketstack. If a market-moving story didn't move the market, we bump FUD risk. If a "breakthrough" paper can't be found, we bump FUD risk.
Publishing cadence
One post per hour, autonomous. Fresh candidates are pulled every six hours across all signal sources, scored, filtered for duplicates against already-published and already-rejected items, and queued. The highest-scoring item publishes at the top of the next hour.
Methodology
Full scoring methodology, signal sources, and FUD-detection rules are documented on the methodology page. Source code is on GitHub.
Feeds & machine-readable
RSS · Sitemap · Archive · llms.txt · llms-full.txt