About, privacy, and use
What this site is, how to use it, how it handles personal data, and how to request a correction or removal.
What this site is
SWE Digest is an independent, automated software engineering digest. Scheduled runs collect public engineering news from primary sources, Hacker News, Reddit, arXiv, and engineering blogs, then publish dated summaries that link back to the original sources. Digests update five times a day, at 01:30, 05:30, 09:30, 13:30, and 17:30 UTC. It is a personal editorial project. It is not affiliated with, endorsed by, or speaking for any company, vendor, or the people and projects it covers.
How it works
The model can't change anything outside its call:
- Python collects candidates from the watchlist before the model starts.
- The model runs with every built-in tool removed, including the shell, file access, and web search. Its only tool,
fetch_url, reads an https page, and the run record logs each call. - The model answers in JSON. Python checks each story and refuses a story that cites a URL the run never saw or repeats a published story. Python writes every file.
- A separate job with the write token checks the changed files against a path allowlist before it commits them. A run can propose changes to the watchlist and the reading profile only through a pull request.
Every run leaves a public run record. It lists the pages the model fetched and the stories it added or refused, with the reason for each refusal.
| Runs | 155 |
|---|---|
| Candidates collected | 47,548 |
| Pages the model fetched to verify | 1,684 |
| Stories published | 713 (1.5% of candidates) |
| Stories the validator refused | 28 |
| Model time per run, median | 57 s |
| Model cost per day | $5.97 |
The stats page charts these numbers over time.
To run your own digest with your own profile, see Run your own.
No warranty and no reliance
The digest is provided for general information only, "as is" and without warranty of any kind. Summaries are generated automatically and may contain errors, omissions, or items that were accurate when written but have since changed. Nothing here is professional, security, legal, or financial advice.
Security and outage items are time-sensitive and may be incomplete or wrong. Always verify against the linked primary source (the vendor advisory, release note, or status page) before acting, and follow your own vendor's official guidance for patching and incident response. Do not rely on this site as a sole source for any operational decision.
Licensing and reuse
- Source code in the repository is licensed under the MIT License (see the
LICENSEfile). - Original digest text and site copy are licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). Reuse with attribution and a link back.
- Linked sources, quoted titles, and other third-party material remain the property of their respective owners and are referenced under applicable quotation provisions.
JSON API
Every build publishes the digest as static JSON under /api/. Start at /api/index.json, which links to the rest.
| File | Holds |
|---|---|
index.json | every day, newest first, with its story count and links |
digests/<date>.json | one day's stories with every field |
stories/<n>.json | all stories, newest first, 100 a page, with prev and next |
search.json | each story's date, title, section, category, blurb, thread, and URL |
threads.json | every follow-up thread, open or closed, and its stories |
The site is static, so the server answers no queries. To search, download search.json and filter it. Text fields are plain text, with code marked by backticks. Treat them as untrusted input. Each file carries a version, which changes only when a field changes meaning or goes away.
Privacy
This site runs no analytics, sets no cookies, embeds no third-party trackers, and has no public accounts or comment system. Reading a page sends no data to the maintainer. The site is hosted on GitHub Pages, and the host may record standard request logs (such as IP address and user agent) under GitHub's own privacy policy; the maintainer does not collect or process those logs.
A few readers the maintainer invites can enroll a browser to leave feedback on stories. That feedback goes to a Cloudflare Worker the maintainer runs. Every page loads the script for it, but in a browser that isn't enrolled the script sends nothing.
Everyone else interacts through the public GitHub repository. Anything you submit there is public.
People and projects covered
The digest reports on public, engineering-relevant work and public statements by named practitioners and organizations: releases, talks, posts, advisories, and similar. It links the primary source and attributes statements only to verified accounts. It does not publish private contact details, private personal information, or special-category data, and it does not track private behavior.
This processing is carried out for editorial publication under the Swiss Federal Act on Data Protection (FADP).
Corrections and removal
If you are mentioned and want a correction, added context, removal, or exclusion from future coverage, open a correction or removal request and identify the specific page or item. Use that template rather than a blank issue: it keeps the request open for a human. Issues are public and are monitored. Reasonable requests are honored promptly, and asking to be excluded does not require a reason.