cast_otter

Seoul · since 2010Numbers first

@cast_otter

← All works Works · 02

Finance Desk News Radar

Out of the morning’s pile of headlines, it picks the ones that touch your numbers, scores them and sends them over.

Running2 key · Finance Desk News Radar

01The problem

Finance people skim the news every morning. Most of it has nothing to do with their numbers, and the one story that does often only gets noticed in the afternoon.

02How it works · 3 steps

  1. Collect and sort

    Articles come in from a fixed list of sources and get sorted by sector.

  2. Judge and score

    An AI model decides whether each article is an opportunity or a threat for us, and adds a priority score and a short summary.

  3. Send the digest

    The top scores go out as an HTML digest and an email. Before any model change, the old and new models run side by side and get compared first.

03Screens

The News Radar insight screen: four summary cards, tabs, an alert strip and a digest paragraph, with the text blurred.
Real screen · data blurred

04Stack

  • Local Python collector
  • Scheduled AI-model jobs
  • Static HTML view and email digest
  • Side-by-side check before a model change (shadow A/B)

05Status

Running

It runs by itself on a schedule. Headlines and scores on the screen are blurred.