Euro Digest - every morning, last night's European football on one screen
Translated by AIView original
Screenshots
Screenshot 1 of 5
What is Euro Digest?
A site that gives you last night's European football every morning at 7 AM as one "morning paper".
European matches end in the early hours of the Korean morning, and the fastest news comes out first in English articles. My son loves football and said he wanted to see European football news fast, so I built a site where it's already summarized in Korean when you wake up.
It covers the top 5 leagues (EPL, La Liga, Serie A, Bundesliga, Ligue 1), the Champions League, and info on Korean players and the national team. You can pick a day on the calendar and reopen that day's edition.
Euro Digest goes the other way: it publishes once a day, one neatly packaged issue. I designed it so you can catch up on all of last night's European football in 5 minutes on your commute or school run.
Key info
- Today's 5 lines : A Korean briefing written from yesterday's match results, standings and scorer data. Lines written by AI get an 'AI-written' label, and a code validator checks the scores and standings against the raw data.
- 45 main news items : Articles from overseas and domestic outlets like BBC, The Athletic, Di Marzio and Inter Football, with the original headlines and links to the source. The same story is grouped together so you can jump between "overseas coverage ↔ domestic coverage", and you can filter by team, player and category tags (transfers, injuries, managers, national team...). Turn on the 'Korean players only' filter and you'll see only Kim Min-jae and Lee Kang-in news.
- Side panel : Tonight's matches to watch, a mini league table and a transfer hot list are pinned on one side.
What makes it different from other football sites
1. It picks tonight's matches for you
It selects the 5 most worth-watching matches from those kicking off between 18:00 that day and 07:00 the next day (KST), using a score. +5 if a Korean player's club is playing, +4 for UCL, +4 for a clash between top 6 teams, +4 for a derby, +2 if the points gap is 3 or less... The selection criteria are public. The goal is not score updates but to give an answer to "what should I watch tonight?"
2. League standings "compared to yesterday"
It shows standings, schedules, results and top scorers for 6 competitions, and also marks how positions changed from the day before. You can see at a glance who moved up from last night's games.
3. Formations and team sheets per team
On each team page, it draws the most-used formation from recent lineups on a pitch. Starting and substitute lists, the last 5 matches' form (W·D·L) and a subscription to the next fixture are all on one screen, and Korean players are highlighted separately in the lineup. Once a week an AI writes a one-line team summary plus strengths and weaknesses (e.g. "fewest goals conceded in the league: 8", "23rd in Champions League scoring"), which is handy for quickly getting to know the opponent before a match.
How it was made
Biggest changes since the idea post
https://jarangter.co.kr/post/f86f3f20-3b8d-4ac1-a215-1827ef26c09f
When I first posted to the Ideas board, the plan was "summarize articles into 3 lines with AI". While building it, two things worried me: the copyright issue of summarizing and republishing other people's articles, and the risk that AI would make up facts not in the article.
So I changed direction. Not a single character of article text goes into the AI. News only gets the original headline and link as-is, and classification, tags and grouping of duplicates are all done by code. The AI writes "Today's 5 lines" only from structured data like match results, standings and scorers. It's not allowed to write facts that aren't in the input, and a code validator re-checks scores and standings and swaps any wrong line for a fixed sentence pattern. AI calls happen in just 3 places (daily briefing, weekly team one-liners, weekly Korean player summary). That cut costs, and on screen the "sentences AI wrote" and the "raw data facts" are clearly separated.
Serverless daily pipeline
It runs serverless on GitHub Actions + a GitHub Pages static site.
- Before dawn, cron-job.org wakes up Actions (with a separate backup schedule)
- Collects RSS from 14 outlet channels whose terms I checked, plus yesterday's match data (football-data.org, API-Football)
- URL normalization, dedup, grouping the same story and importance scoring pick the top 45 items, all in code
- Only the match data goes to Claude Haiku → zod schema validation → code validation → 1 retry → if it still fails, template sentences (publishing never stops)
- Publish validation gate: if it fails, nothing is committed and yesterday's site stays as is
- OG images → Next.js static build → search indexing → RSS and calendar → deploy
The publish time changes to 07:00/07:30 depending on European daylight saving time (in winter UCL matches end around 06:55 KST), and the code also determines the season.
Cost: target $3/month, actual about $0.2/month
- Built with
Comments 0
No comments yet
Be the first to comment.