I haven’t been there, but I was reading about Dynamic Land just yesterday (via Dominic Cummings’ blog), and I’ve read some of Bret Victor’s writings. I approve of the ideas tremendously, but it’s not clear to me that in practice the work has provided any more of an advance in “visual programming” than other efforts in this area. Beyond the decades-old WIMP (ETA: and spreadsheets) interface, none of these, it seems to me, ever make more than toy demos. I have never seen them scale up to real power tools that someone would use to accomplish something. Ideas like these have been around long enough that toys and dreams will no longer do.
There are lathes that can make all of their own parts. Could Dynamic Land create Dynamic Land? What would such a system look like if it could?
I agree that it only makes toy demos, but it definitely goes beyond WIMP. It’s a simulation of the sort of interface one might expect in a future where every surface is a screen—it’s a janky, extremely-low-fidelity simulation, which holds together barely well enough to serve the purpose, but it does serve, and it’s an interesting way to try out this interaction style.
A surprisingly large part of Dynamic Land is actually (in some sense) self-hosting. There is a small core/kernel that is extrinsic, including e.g. the parser for the language. But a lot of low-level functionality is indeed implemented inside the interpreter (as I recall, they use a sort of pseudo-Lua, crossed with Smalltalk-like agent stuff—my vague recollection is that there’s something like a Lua interpreter with a custom preprocessor underlying it.)
I haven’t been there, but I was reading about Dynamic Land just yesterday (via Dominic Cummings’ blog), and I’ve read some of Bret Victor’s writings. I approve of the ideas tremendously, but it’s not clear to me that in practice the work has provided any more of an advance in “visual programming” than other efforts in this area. Beyond the decades-old WIMP (ETA: and spreadsheets) interface, none of these, it seems to me, ever make more than toy demos. I have never seen them scale up to real power tools that someone would use to accomplish something. Ideas like these have been around long enough that toys and dreams will no longer do.
There are lathes that can make all of their own parts. Could Dynamic Land create Dynamic Land? What would such a system look like if it could?
I agree that it only makes toy demos, but it definitely goes beyond WIMP. It’s a simulation of the sort of interface one might expect in a future where every surface is a screen—it’s a janky, extremely-low-fidelity simulation, which holds together barely well enough to serve the purpose, but it does serve, and it’s an interesting way to try out this interaction style.
A surprisingly large part of Dynamic Land is actually (in some sense) self-hosting. There is a small core/kernel that is extrinsic, including e.g. the parser for the language. But a lot of low-level functionality is indeed implemented inside the interpreter (as I recall, they use a sort of pseudo-Lua, crossed with Smalltalk-like agent stuff—my vague recollection is that there’s something like a Lua interpreter with a custom preprocessor underlying it.)