Using some of the early royalty prepayments from Personal Software, Bob and I were able to make the move to start being a "real" company to finish the coding. We sublet some space from another tiny software company, John Strayhorn's Renaissance Computing, in Central Square, Cambridge, Massachusetts, down the road from MIT. It was in the basement. Here is a picture of the windows up in the top of the walls:

Software Arts offices in the basement in Central Square, Cambridge, Massachusetts, mid-1979

We borrowed money from a bank and bought a Prime minicomputer on which to run our development tools to finish the first release. The Prime had a good version of PL/I in which to write the tools. Bob wrote a macro assembler and linker for the 6502 microprocessor, and I wrote a simple visual editor. Since we were below ground, sometimes when it rained the drains would back up and the floor would get wet. Here's a picture of me standing in front of the Prime about to push away the water:

Dan in front of Prime minicomputer, fall 1979
Photo courtesy of Bob Frankston

We first shipped about five copies of version 1.35 to some early customers in the late summer of 1979. I hand typed the labels. The first "real" release, version 1.37, shipped in mid-October 1979.

Entry in my journal: First complete VisiCalc in a package was October 19, 1979. I received a copy the next day as I recall.

Here is what the first VisiCalc packaging looked like:

First VisiCalc packaging: "VisiCalc, Personal Software Inc."

It was a brown vinyl binder holding the manual, diskette (5 1/4" diskette), reference card, and registration card. The reference card was written by me. The production and volume printing of the card was done by my father, Baruch Bricklin. The diskette duplication and all other production of the product was done by Personal Software. The manual first shipped with VisiCalc was written by Dan Fylstra (the 3rd in a series of attempts: first by me, then a freelance writer, and then Dan Fylstra until they got one acceptable to Personal Software). It included a mixture of examples, including sales / cost projections to teach the basics, home budgeting to show more advanced commands and techniques, and scientific calculations to demonstrate the built-in functions.

The Reference Card looked like this on the first two (of 10) panels:

VisiCalc refcard panels 1 and 2
VisiCalc Reference Card, 6/79 V.1.35, shipped with first release, panels 1 and 2

A scan of the entire Reference Card is reproduced here on this web site. The original 10 panels (5 on a side, folded accordian-style) is presented as 5 scans, each 60K-80K bytes in size. The Reference Card serves as the specification of that original version of VisiCalc. The last panel, which has a picture of the screen with the parts labeled, is probably the most interesting:

VisiCalc refcard panels 9 and 10
"A VISICALC Screen" from panels 9 and 10

To get a feel for what VisiCalc was like to use, you can try a copy of the early IBM PC VisiCalc from 2 years later in 1981. It still runs on today's PCs under MSDOS in Windows. That version of VisiCalc was only 27K bytes in size. The original Apple II version from 1979 could run on a 32K byte Apple II  -- those 32K bytes included the screen memory and the OS, as well as the VisiCalc program and the space for data.

There is a plaque on the wall in room 108 in Aldrich Hall at the Harvard Business School commemorating the invention of VisiCalc. It includes a copy of a sketch I made as I designed the reference card:

Plaque in Aldrich 108 at Harvard Business School

Peter Jennings, who was one of Dan Fylstra's partners at Personal Software/VisiCorp and responsible for their MicroChess program that helped fund the early advance royalty payments for VisiCalc, has posted some of his recollections of the early days of VisiCalc on his web site.