In 1957 the CIA asked IBM to invent Google Images, and IBM said yes.

This story starts with a scene straight out of a spy movie and I swear nearly every word is true.

In the mid 1950s, a CIA spook comes to IBM with a request: “Hey Blue suited dudes, make us a machine that can store millions of images that are searchable by keywords. In other words a data lake of images – just like Google Images, except Google is still 40 years in the future.”

The IBMers didn’t even blink (or ask what a Google was). They replied, Hell yeah!” (or something similar), and the resulting product (code named Walnut) was delivered to the CIA in 1961.

It kinda worked like this:

The Walnut has 100 large cylindrical carousels called document stores.
Each document store held it’s data in in 200 small boxes that IBM referred to as cells.
Each cell contained 50 strips of photographic film
Each strip of film contained 99 photographs arranged in a 3 by 33 grid.

In total, each document store contained images of 990,000 documents, and up to 100 document stores could be used in a single Walnut system, for a total storage of 99,000,000 pages

Users would look up keywords stored on a separate system that used an IBM 1405, identifying individual documents to be retrieved. The index produced punched cards that were inserted into the Walnut. The Walnut system then retrieved the documents, copied them onto a blank film strip and developed them. The output could be viewed using a photo-negative viewer, or developed for full-sized printouts.

So to get this right, our CIA Operative searches for photos of villains using keywords and gets a list of images on punched cards. They then feed them into the Walnut, which then spits out a negative they can use to develop photos.

Imagine the James Bond scene: “Walnut, show me all images of Blofeld!” (followed by a 10 minute wait involving punched cards and photo development).

This is just like Siri, but with punched cards! (now hang on, which one is the real Blofeld?):

IBM tried to sell a commercialised version of this but got no takers. It seemed no one wanted to buy a product called Walnut (although it was now known as the IBM 1350).

However IBM realised that if they stored actual data on those negatives, rather than images, they might have a truly giant online read-only file system.

So they then productised this idea in response to demand from the Atomic Energy Commission to store 1 terabit of information to support super computer simulations. Yes you read that right – 1 whole terabit… that’s 128 GB! At that time no single system could address that much data.

The resulting product however was mildly insane and bizarrely huge and was an incredible mix of pneumatics and robotics combined with an electron gun and automatic photographic development. The product looked nothing like this:

It actually looked like this (a little bit larger, but how could you not love those blue covers?):

The Lawrence Livermore National Laboratory also loved it, using it for over 10 years as a (for that time) enormous online archive of read-only data (albeit with a response time in minutes). But despite the enormous amount of data you could potentially store in such a beast, it proved as equally impossible to sell commercially as it’s photo storing cousin.

The Computer History Museum in Mountain View has the only remaining example of an IBM 1360 (from Lawrence Livermore National Lab) shown in the following video. Note the tour guide claims it was retired because all the Service Reps who knew how to fix it had retired, but the truth is IBM no longer wanted to make spare parts for it and the client could not find any one to take over.