Graphics editor without toolbar
On breaking a 50-year-old UX tradition and a surprising effect
daniel@squaresight.io July 2026
Beginnings of human–computer interaction
Early computers were seen as pure calculating machines, filled entire rooms, and direct interaction with the user was not possible. Instead, it was often the case that input had to be carefully planned and punched onto cards. A technician collected cards from multiple users and fed them into the computer's reader in batches. After a waiting period, which could last hours or sometimes even days, the output was produced in printed form.
Sketchpad, in the early 1960s, was downright revolutionary. It enabled direct, real-time interaction with graphical objects on the screen, allowing vector drawings to be created iteratively.
Sketchpad was operated with both hands. With the light pen in one hand, a position on the screen could be marked. With the other hand, command keys could be pressed.
Although the command keyboard appears in front of the screen on the table in many photos of Sketchpad, in a video from the Computer History Museum [1] it is arranged to the left of the screen. This resembles later toolbars. However, the keys in Sketchpad were only used for commands, and the operating concept did not include switching between tools.
Augmentation Research Center
Sketchpad showed that a computer can be far more than a calculating machine. Also in the early 1960s, and independently of Sketchpad, Doug Engelbart founded the Augmentation Research Center to find new ways for computers to support people and expand their capacity for thought.
In 1968, Doug Engelbart presented NLS (oN-Line System) as the result of this work. NLS enabled the editing of structured documents made up of text and graphics.
As with Sketchpad, both hands were used to operate NLS, with the light pen replaced by a three-button mouse. In addition to the regular keyboard, there was a second keyboard with five keys for the other hand, also known as a chord keyset.
Xerox PARC
As impressive as NLS's capabilities were, the learning effort required to use it was just as steep. However, Doug Engelbart considered this necessary in order to unlock its full potential. There were also researchers, though, who saw ease of use as the foundation for the future success of personal computers.
Starting in the early 1970s, Xerox PARC (Palo Alto Research Center) offered researchers a lab for this purpose, where they could develop new interaction concepts. Some employees of the Augmentation Research Center subsequently moved to PARC.
Starting in 1973, the Xerox Alto, an in-house computer, was available. Similar to Engelbart's NLS, it relied on a three-button mouse and a chord keyset alongside a regular keyboard for input. The Xerox Star followed in 1981 as another computer, where the trend toward simplification was also visible in the input devices: the mouse now had only two buttons, and the chord keyset was dropped.
Several graphics editors were developed at PARC. In SuperPaint, users could switch from the drawing canvas to a full-screen tool palette [3]. In Markup, a palette could instead be displayed at the mouse pointer [4a]. Shortly after Markup, Draw emerged starting around 1975 [5], featuring a toolbar next to the drawing canvas. However, the video from the Computer History Museum [4b] shows a later version from 1980, so it is unclear whether the toolbar was included from the very beginning.
Smalltalk-76's BitRect toolbar can be dated somewhat more precisely. Smalltalk was one of the most important projects at PARC and is an object-oriented, interactive programming environment. BitRect is an interactive object within it.
“This was something a drawing program that anticipated MacPaint by quite a few years. I guess this was done right about 1976. … Here we have a menu of tools that you can use.”
In addition to graphics editors, text editors were also developed at PARC. Particularly notable is Gypsy from 1975. It was the first to use a modeless interface, in which Larry Tesler played a major role. He saw the reduction of modes as an important criterion for ease of use and strongly advocated for software to become modeless.
Fun fact: Larry Tesler was considered one of the strongest advocates of modeless interfaces. He had a license plate made for his car reading NOMODES and used the web address www.nomodes.com.
Apple
Some Apple employees received a tour of Xerox PARC in 1979 and were impressed by the possibilities of graphical user interfaces. With Xerox's permission, they were able to draw inspiration from what they saw and further developed many of the ideas.
But it wasn't just ideas that made their way to Apple. Just as researchers had once moved from the Augmentation Research Center to PARC, some now moved from PARC to Apple to work on the Lisa and Macintosh projects. Among them was Larry Tesler, who joined in 1980 and, in a leading position, advocated for a modeless design across all applications.
“The Lisa software engineers internalized modeless editing principles and found clever ways to make their applications modeless. My job was to manage them and make evidence-based decisions about the design of the interface.”
A modeless interaction concept that had already been successfully implemented for a text editor at Xerox PARC was now to be introduced for a graphics editor as well. The same approach as with the text editor was followed: the user first makes a selection and then executes a command.
“The Lisa editor was modeless in that you would stretch out a rectangle of marching ants [animated dashed selection outline], and then say ‘Put an oval in it’ …. You'd get no feedback that you were making an oval. And when you thought about making a curved line, that became ridiculous, stretching out a rectangle that was going like this and now saying ‘put a curved line in it,’ that was a joke.”
Bill Atkinson was on Larry Tesler's team and shared his passion for simple interaction concepts. He recognized that this approach for a modeless graphics editor was too cumbersome and that there should be a drawing tool. Although he, too, preferred a modeless concept, he saw holding a pen or an eraser in hand as a natural mode and proposed a deliberate exception for it.
“I was arguing that we needed to be in a mode for the creation stroke, like you grab a crayon and you draw with it and every stroke you make a crayon mark. You pick it up. Instead, a marker and every stroke is a marker or an eraser and these are really modes. You're in eraser mode.”
Unlike at Xerox PARC, where researchers could develop solutions without time pressure, Apple faced the imminent market launch of the Lisa and Macintosh computers. To make the most of the limited time, Bill Atkinson often built software at night, which was then tested by others during the day. He proposed a toolbar to solve the problem with the drawing app, and created a test program for it.
“So, I had to make a little mock up. I realized, okay, if there's going to be a mode for the tool that you're in, that mode has to be in front of your face and visible, so that at least it isn't a hidden mode. If you select a tool from a menu, it's not really in front of you like a palette, and that was the origin of a tool palette; I wanted to put it out there on the screen. You can see at any moment at a glance, what tool am I in? Is this stroke going to erase some stuff or going to draw with marker or et cetera?”
For the tests, users with no computer experience were deliberately included as well, in order to assess how self-explanatory an interface was. Bill Atkinson's toolbar was successful in these tests, and he was able to win over his team with it. As a result, all three graphics editors under development at Apple at the time included a toolbar: LisaDraw, MacDraw, and MacPaint. The Lisa computer launched with LisaDraw in 1983, and the Macintosh followed a year later with MacDraw and MacPaint.
The trend toward simplification was also visible in the Lisa and Macintosh mouse, which had only one button. This was later jokingly described by a co-inventor of the three-button mouse and colleague of Doug Engelbart as one of the greatest mistakes in product history
[10]. It wasn't until 2005 that Apple reintroduced the option for a second button with the Mighty Mouse, followed by touch gestures with the Magic Mouse in 2009. Multi-touch gestures, such as those on MacBook trackpads, are no longer self-explanatory, but they enable more effective use, very much in keeping with Doug Engelbart's philosophy.
Other companies, such as Adobe, also adopted a toolbar after MacPaint, with the number of tools continuing to increase over time. In Photoshop 1.0, released in 1990, all the tools still fit in the toolbar. Later, there was no longer enough space, and the toolbar gained menus. Even Larry Tesler seems to have made peace with the toolbar:
“As for graphics editors, I agree with the common wisdom that modes can be good when they support a metaphor like picking up a brush, and when feedback identifying the current mode is displayed where the user is almost certain to be looking.”
It can certainly be assumed, however, that Larry Tesler would have greatly appreciated a modeless graphics editor. But what could a new approach to this look like? Or is this simply not solvable in any meaningful way?