Extract upcoming scheduled events per tracked asset and estimate impact
Builds a forward-looking calendar of scheduled events for your tracked assets. It reads recent news articles and uses a language model to extract upcoming events (token unlocks, protocol upgrades, governance votes, regulatory hearings, ETF decision windows, mainnet launches), each tagged with a date, type, description, and confidence, then keeps only those tied to assets you follow. Use it to know what is coming and when. It matters because scheduled events mark when volatility is more likely to be elevated, for reasons already on a calendar.
Extracting structured events from unstructured text - and the right division of labour between a model and code.
WHAT IT DOES. Reads recent news articles and uses a language model to extract upcoming events - token unlocks, protocol upgrades, governance votes, regulatory hearings, ETF decision windows, mainnet launches - each tagged with a date, type, description and confidence.
WHY A MODEL IS THE CORRECT INSTRUMENT HERE, and this is the point worth generalising. The task is to read a sentence such as 'the upgrade is expected to ship in the second half of next month, following the governance vote' and produce a date and a type. That is COMPREHENSION: it needs resolution of relative dates, of hedged language, of which entity the event belongs to. A keyword list cannot do it, and no amount of extending one would. Compare the sentiment analyses in this catalog, which score text by word membership and inherit a permanent ceiling as a result: the honest test is whether a human would need to READ the sentence, and here they plainly would.
WHAT CODE STILL DOES. It checks each date it can parse against the day of the run, so a past event is not listed as upcoming, while a date in words it cannot place, or no date at all, is kept. It merges one event reported by several outlets and enforces the output schema. Structure stays deterministic; only the reading is inferred.
THE CONFIDENCE FIELD IS DOING REAL WORK. An extracted event has two independent uncertainties: whether the article said it, and whether it will happen. A model can speak to the first with some reliability and cannot speak to the second at all - a protocol upgrade 'scheduled' for a date slips constantly, and a regulatory decision window is a range with a history of extensions. Read confidence as extraction confidence, not as probability of occurrence.
THE SOURCE LIMIT. Events are extracted from NEWS, so the calendar contains what has been written about. Scheduled events with formal calendars - an unlock schedule in vesting documentation, an on-chain governance deadline, a published regulatory docket - are knowable directly and precisely, and reading them from journalism is a lossy path to information that exists in structured form. That is the same 'use code for structure' point arriving from the other direction.
WHY A FORWARD CALENDAR IS USEFUL AT ALL. Under semi-strong efficiency an anticipated event should already be substantially priced. The value is not surprise - it is knowing WHEN volatility is likely to be elevated, and when a position's risk changes for reasons already on a calendar.
The event taxonomy and the confidence scale are our own convention.
Published reports are readable without an account. The newest is Ether (ETH) and XRP extend a month-long slide as Bitcoin falls below $75,000: trend, momentum and news (Nov 9, 2026). Create an account to run research on the stocks and crypto you choose.