Fix 4 dead neptune.ai blog links across course docs - #39
Fix 4 dead neptune.ai blog links across course docs#39julamarkova-vibe wants to merge 4 commits into
Conversation
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
… were dead (#44) The neptune.ai breakage in #39 was found by a reader, not by the repository. A course is mostly links to other people's work, and nothing here noticed when one of them rotted. A full lychee run over the README and every chapter found nine genuinely dead links besides the neptune ones: - three of this project's own MLOps Community articles, which moved to /blog/<slug> when that site restructured - Microsoft Learn's VS Code module, VS Code Live Share, Jupyter Book's cell hiding page, ZenML's starter guide, the OpenRAIL announcement, and Marquez, all moved or renamed - Censius, whose TLS handshake now fails outright; the two explainability definitions it backed now point at Molnar's Interpretable Machine Learning `check:links` is deliberately NOT part of `check`. The first run reported 63 failures against 706 unique URLs, and only 9 were real: 37 were HTTP 429 and 14 were bot-blocking 403s from hosts that serve a reader fine. A gate that cries wolf 85% of the time teaches people to ignore it. It runs monthly and on demand from links.yml instead, with a .lycheeignore for the hosts that refuse scanners and 429 accepted as proof a host is alive rather than evidence it is broken. The whole set now reports 0 errors.
|
Thank you — the diagnosis was correct and the report was genuinely useful. I verified it independently: every I went with pinned Internet Archive captures for all four rather than the glossary pages — which is the alternative you offered up front, and I appreciate you disclosing the affiliation without being asked. The reasoning: the surrounding prose in each chapter was written against those specific articles, so the archive preserves the actual reference. Each snapshot was verified through the CDX API as a real Your report also surfaced two larger problems sitting behind it:
Closing in favour of #43 and #44, but the credit for finding it is yours. Thanks for taking the time. |
neptune.ai was acquired by OpenAI in December 2025 and the entire blog now redirects to the acquisition announcement, so the neptune.ai/blog links in this repo no longer lead to the referenced articles.
This PR fixes 4 dead links:
Disclosure: I'm one of the authors of bestaiweb.ai. I've only used our pages where they genuinely cover the same ground as the dead article; the remaining fixes point to the Internet Archive (pinned to the last capture before the redirect). Happy to switch any of these to archive.org snapshots instead if you prefer.