Usenet-Rewind is a new archive and search service from Erie Data Systems. It contains 1,039,260,173 public messages dating back to 1981. The service emerged after Google stopped accepting new Usenet posts in Google Groups in 2024, leaving an old archive that remained online but was difficult to search.

Ernie Smith tested Usenet-Rewind with the phrase “Eternal September” and found the earliest post cited by Wikipedia. The archive also surfaced a September 5, 1995 message in which Matt Levine promoted his Top 40 mailing list on rec.music.misc. Searches found Michael Wolff’s 1994 promotions for Net Guide and records of the spam-related mass cancellations that affected his company’s posts later that year.

The collection covers text messages and discussions, but it does not appear to include usable binaries. Smith found software-related groups such as alt.ibm.warez and references to removed uuencoded data. Users should not expect to retrieve 35-year-old warez from the service.

Erie Data Systems founder Craig J. Stadler spent about 15 years working on search-engine architecture with Apache Solr. He says he spent a year gathering dumps, backups and donated archives, then built an NNTP crawler to fill gaps with newer public Usenet data. Usenet-Rewind offers limited free searches for new accounts. A subscription for serious use costs $9.99 per month, while a higher plan with API access costs $39.99 per month.

The archive also preserves material from mid-1990s celebrity chats held through AOL and IRC. Smith located exchanges involving Iggy Pop, Mike Judge, Chris O’Donnell, Marilyn Manson and John Popper. These transcripts survived because someone copied them into Usenet, where a modern search system can now expose them again.

That access raises privacy questions. Many Usenet users assumed their posts would fade from view after a short time. Usenet-Rewind supports DMCA takedown requests, but removing a page from its index does not erase the original Usenet content. Stadler says the data was not collected for large language model training, is not known to be legally suitable for that use, and will remain focused on traditional tokenized search.

The archive is useful for historical research, sentiment analysis and investigations into online culture. It also makes old jokes, arguments, career experiments and personal mistakes easy to locate decades later. Usenet-Rewind turns a neglected historical record into a practical research database, with the privacy consequences attached.