Ivan Sutherland Sketchpad Demo 1963 [video] | Hacker News Reader

Ivan Sutherland Sketchpad Demo 1963 [video]

youtube.com

101 pointsfs_software14 comments

The first demonstrator here is Timothy E. Johnson, not Ivan Sutherland. Johnson did the "Sketchpad III" work extending Sketchpad into 3D which is demonstrated from 11:30 https://www.youtube.com/watch?v=6orsmFndx_o&t=690s . Then Larry Roberts appears to demonstrate his own work on hidden-line removal, which is shown from 14:42 https://youtu.be/6orsmFndx_o?t=882 .

See https://tx-2.github.io/videos#sketchpad for more video, including appearances by Sutherland himself. https://tx-2.github.io/ is an active effort to get Sketchpad running again on a new TX-2 emulator.

At 5:38 he describes "gravity fields" that snap your cursor to lines and endpoints - letting you "be sloppy while drawing and get a precision drawing at the same time."

Every design tool today (Figma, Illustrator, CAD) still uses this exact UX pattern. Sutherland nailed it 62 years ago with a light pen and an oscilloscope.

Super excited about this kind of stuff. Sutherland, mother of all demos, Bret Victor, live fish, etc.

However, going this route for real likely means multi-decade research and iteration.

Demos are quick to make. Generalizing and turning it into real reliable software seems tremendously hard, and beyond just a shift in mindset.

Fortunatelly, we now have vibe coding, so anyone can experience first-hand the frustration of trying to just shift your mindset and immediately reaching a metric ton of limitations in the very first iterations. It's a humbling experience that I recommend to anyone (go ahead and change the world with precision UI, just try it).

you'll like this then https://www.youtube.com/watch?v=oKg1hTOQXoY

Sutherland figured out how graphic interaction ought to work, with the computer recognizing near points and connecting them. What we now call "snap". He had the key idea of CAD - you can draw with more accuracy if the computer helps.

That demo is running on the MIT TX-0, a transistorized version of Whirlwind and the predecessor of the PDP-1. It was somewhat obsolete at that point, so projects like this could get time on it.

He brushes over the zoom out, which I think was pretty impressive for a computer of this time. There is a lot of redrawing/recalculating going on there. Would be impressive on a 80s microcomputer.

No, rendering to a vector display (hardware whose primitive operations are points and lines) is almost free for the kind of drawings he was rendering. Zoom is just one linear transformation on each point in the model, no different from panning the view.

I mean, the panning was pretty impressive at the time as well. In Sutherland's 1994 "Bay Area Computer History Perspectives" talk https://www.youtube.com/watch?v=k_cevebLO-A&list=PLKTTWvMgeg... he points out flickering or 'jumping' due to the processor burden of rescaling in a number of places in the Sketchpad demo video, including at 42:30 https://www.youtube.com/watch?v=k_cevebLO-A&list=PLKTTWvMgeg... and at 50:10 https://www.youtube.com/watch?v=k_cevebLO-A&list=PLKTTWvMgeg... .

Sutherland started programming on the TX-0, which was widely accessible on the MIT campus, but Sketchpad was definitely done on the big gun, the TX-2, which was still inside Lincoln Laboratories. (Sutherland's uncle-in-law https://en.wikipedia.org/wiki/Ivan_Getting helped get him into Lincoln Labs. See https://www.computerhistory.org/collections/catalog/10273819... .) There's an active TX-2 emulation project at the moment https://tx-2.github.io/ , which has the primary goal of getting Sketchpad running.

Constraints as well, you can hear him talking about them at 8:20, this is fundamental to CAD programs.

The company I worked for in the early 80’s was in the market for a CAD system, and they all had constraint systems in the UI, and showed them in the demos.

CAD people got this a long time ago, but, unfortunately, HTML people never did. Layout is a 2D constraint problem, but HTML/CSS approaches it as a procedural problem, partly because early browsers were so compute and bandwidth constrained.

CAD is great at parametric design, which gets significantly more complicated when the viewport changes shape.

Taking a procedural approach instead of solving constraints is necessary for folks to stay sane when they're trying to get the same website to work on phones and desktops.

"it's real nightmare material" (11:18)

Reply on news.ycombinator.com