Time series · Change point detection · LLM
BetaInflection
Find the moments a metric changed, and the news that explains why.
Loading data…
About this tool
Pick a time series, such as a stock price, COVID cases, or attention to a topic. Inflection detects where its behaviour shifted, searches the news around each date, and writes a cited summary of what happened.
Dashboards show that a number moved. The follow-up question is always why. At Uber I built change point and anomaly detection into an org analytics product so leaders could see when their metrics shifted and what drove it. Inflection applies the same idea to public data anyone can explore.
How it works
- 01
Choose a series
Start from an example, or enter any stock ticker or Wikipedia article. Prices come from Yahoo Finance, pageviews from Wikimedia, and COVID data from Our World in Data.
- 02
Detect the shifts
Dynamic programming finds the best possible segmentation for every number of shifts at once, so the slider updates instantly. Choose between changes in trend and shifts in level, on a linear or log scale.
- 03
Find the context
For each shift, Claude searches news coverage from the surrounding days. Only dates and numbers leave the browser; the server writes the prompt.
- 04
Summarize with citations
The model writes a short explanation for each shift and an overall narrative of the series. Source links come from the search results it cited, and it is told to say so when no clear cause turns up.
Next project
Org Health Explorer