The HARVEST computer (a.k.a. IBM 7950) was a one-of-a-kind cryptanalytic machine designed and built by IBM and the NSA in the late 1950s and early 1960s. In 2010, the NSA declassified a number of documents about HARVEST, revealing more of the machine's design and the planning process that led to it.

I focus more on HARVEST and its technical aspects, particularly its architecture and programming. However, some documents here corroborate other IBM Stretch historical work by Eric Smith and the notes from a Stretch/HARVEST Reunion in 2002.

If you are looking for more details on Stretch, Mark Smotherman's page is a great start.

Please direct questions and comments to eric@fpgacpu.ca


The documents are arranged chronologically, and named using the following scheme:


where FILE denotes where this document came from (usually a NARA File Unit Title), and TYPE is either JPEG or TIFF to denote the format of the scanned image. Any unknown or irrelevant values are set to zero. These are not all the available HARVEST documents from NARA. See the File Request Process below for a link to the relevant NARA page.

Reading documents
These documents are PDFs containing 300 dpi 8-bit greyscale JPEG images, perfectly crisp for reading or printing, but some fine JPEG artifacts around the text might make them unsuitable for OCR or enlargement.
Archival documents
These are PDFs containing the original scans as 600 dpi 8-bit greyscale TIFF images with Deflate (Zip) compression. They are 10x larger than the JPEG versions.
Other documents
These are other documents which supplement the above NARA files. They are of various origins and qualities.


File Request Process
How to order archive files from NARA.