6uclz1 Tech News Digest
Workflow
Use the bundled scanner before browsing individual articles:
python3 scripts/scan_tech_feeds.py --sources references/default_sources.json --days 3 --top 20 --detail-candidates 5 --format markdown
Use python instead of python3 when that is the available interpreter. Add --format json when you need structured data for additional processing.
Source Strategy
- Prefer RSS, Atom, and RDF feeds from
references/default_sources.json. - Do not bulk-fetch article pages or scrape HTML directly during the broad scan.
- Read only the feed-provided title, URL, published time, source, language, and short summary/description.
- Inspect full article pages only for the highest-value
detail_candidateitems, and only as needed to answer the user. - If a feed fails, keep going and mention important gaps in the final digest.
Interest Tuning
Use keyword flags to adapt the scan without editing the default source file:
python3 scripts/scan_tech_feeds.py --include-keyword rust --include-keyword security --exclude-keyword sponsored
Use repeated flags or comma-separated values. Treat include keywords as score boosts, not hard filters; treat exclude keywords as filters.
If network calls are slow, lower the per-feed timeout:
python3 scripts/scan_tech_feeds.py --timeout 6 --top 10
Digest Format
After scanning, produce a concise digest with these sections:
- Top signals: 3-7 cross-community or high-impact themes with source links.
- English communities: notable items from Hacker News, Lobsters, Reddit, DEV, or similar English-language sources.
- Japanese communities: notable items from Hatena Bookmark IT, Qiita, Zenn, JSer.info, Publickey, or similar Japanese-language sources.
- Worth deeper read: the few articles most worth opening and reading closely, with why they matter.
- Sources checked: feeds checked, feed failures, and the time window.
Keep summaries short. Clearly distinguish feed-derived summaries from conclusions made after reading a selected article.