Just like learning Lisp will make you rethink programming languages, and learning Erlang will make you understand the true power of concurrency: learning Smalltalk will make you understand what "object oriented" actually means.
I haven't used Squeak since college but I'm glad it was part of the cirriculum.
Btw, almost all of Javascript's good parts come from Smalltalk.
For me, Smalltalk's "wow" moment was the concept of a persistently running image, with the developer's job being to mould and manipulate it into the shape s/he wants.
Ive always "dreamed" of having something like the ST image but backed by persistent virtual memory, not simply RAM.
(Crudely) mmap the ST image to a 100GB file and just change pages. Let the OS flush the pages back and forth.
Maybe full boat, heap sweeping GCs would be Bad as it pages the entirety of the heap in and out. But we have (had) lots of RAM these days. We have generational GCs that leave idle stuff alone. Its no doubt impractical, but I think it would be neat to have the entirety of my historical email in the global "mailbox" array, and I can build indexes off it as I wished. But the mail isn't "on disk", it's all marshalled up as first class objects.
I don't know enough about it, no doubt it "won't work", but the idea of simply mapping the entire, large, ST VM heap to a persistent backing store, just be interesting I would think.
Dangerous too, as there is a demarcation between "running image" and "saved image". But, still think it could be interesting.
I'm also interested in this and made decent progress on a git-style content addressed image where everything is identified by hash and residency determines what's in memory. For example, there's a filesystem on the image that works this way and you can page over the "cold" non-resident bytes for very large files that you would not want to hold in memory.
You don't persist the state of the full image continuously to the host disk, but any given state (new programs, new runtime state, filesystem) can be saved via a delta, not a full rewrite, due to this representation.
So this is a little different than what you are talking about, but I'd say it's possible.
It's not widely known, but GNU Emacs works like this.
When you launch the Emacs editor, you're loading an Emacs Lisp image that was populated at build time by running Lisp code, with the resident Lisp definitions "dumped" to make an image file.
The image file used to actually be the `emacs` executable you'd run, using a clever but non-portable mechanism called `unexec` to make an executable.
But as of version 27.1 (in 2020), the image file is separate from the executable for portability reasons.
The only mainstream-ish language where that still happens? R. Too bad, there's a lot more use cases for this sort of thing now - versioning, anything non-persistent agents touch, collaboration, auditable enterprise LOB. It's 2026, why are we (or our agents) still writing serialization code? Even if the AI's write the boilerplate, the state management fragility is often a tax/risk.
I wish this didn’t go out of fashion as much as it did. An environment that you perturb makes more sense in a lot of ways than a “tear it down and restart” model.
I think the gnarlier part is that it reveals how much (and how error prone) initialization code is. When objects start glued together in the right structure a lot becomes magically easier to deal with.
I'm considering this for my language but it's a lot more complex to implement (especially with today's ecosystems) and I keep wondering if it's worth it.
Lisp, Forth, Erlang, Smalltalk, and Rebol, so easily forgotten.
All languages that multiply your insight into the nature of computer programming. A bit sad that many just stop at the first of these, and many more don’t even venture past their Pythons and Javascripts.
Lisp is not "forgotten". It's quite alive, kicking and running [big] circles.
Nubank (the largest digital bank in the world) using Clojure as their primary driver grew from ~12M customers in 2019 to 131M in 2025 - almost 1000% growth in less than a decade. JPMorgan that has about 80M customers took around 150 years and dozens of acquisitions.
Apple, Walmart, Netflix, Cisco, Amazon - they all use Clojure. Google and Grammarly use Common Lisp.
Check the GitHub language stats and be surprised to see Emacs Lisp among the visible top. There's so much Elisp on GitHub alone - it's just effing crazy. That fact alone at least should be surprising. Remember? This is a not a general-purpose language - it's made for one and only thing, tis a damn configuration language for a text editor.
And that's completely natural. Compare it with woodworking.
Yes, some people would dive deep into the craft and create works of art, exquisite furniture and tools they need to make it even better.
But those are the few. For the most of woodworkers this is just something they somehow learned to do on a level needed to be able to get a job and make living.
I find that lamentation a bit funny, because the list in practice is never really complete. For instance, you forgot the mighty Prolog. Surely, tcl belongs on there too.
The more of them you explore, the more insight you get. The list does not have to be complete. It just needs to list some languages, that are truly different in how they work and what the ideas behind them are.
From what I've seen... I think if I were writing business applications, I'd reach for Elixir... if I were writing something lower level like communication software, I'd reach for Erlang.
I have written a fair amount of Elixir (and targeting business applications) and I think Elixir's expressiveness lends itself well to making more abstract models of the business processes and data seen in solving business problems. However, writing Elixir I find you end up reading a lot of Erlang and in certain problem spaces I could imagine the appeal over Elixir. I could absolutely see where if I were writing more technically oriented programs.... programs closer to the hardware in some ways or programs designed to communicate with other programs (protocols, etc.) I could see Erlang as being a more clear and direct language in these circumstances.
All of this is very hand-wavy, but it's my impression.
For me, Elixir misses the mark. If you're an unc with a background in Ruby or that era of webdev, I can see the appeal. It's a familiar set dressing. I don't have that background, I much prefer the syntactic structure of Erlang's psuedo-horn clauses. Both for similar familiarity reasons (I knew Prolog before I touched Erlang), as well as a general preference for the simplicity.
I didn't imply otherwise, nor is the story different on the Erlang-Prolog axis (actually it's even more pronounced there.) I literally called it "set dressing". Where I think you go off the rails is that you're conflating surface-level things with irrelevance. Which isn't really true, especially in matters of taste.
I find this to be a very strange thing to get defensive about. It's not exactly a subtle part of Elixir's history. It was pretty well documented to be born out of the creator's dissatisfaction with concurrency in Ruby, and simultaneously, for a more Ruby-like language on the BEAM. That doesn't mean it's literally Ruby-on-the-BEAM (which was originally attempted, and didn't work out), however mistaking that as cause to dismiss the link between the two and the carry-over appeal for developers is an unsound overcorrection.
While I appreciate guest languages for bringing new ideas to specific language platforms, at the end of the day I rather use those that are developed alonside the platform, hence Erlang instead of Elixir, at least if I would be ever doing something with it again.
Last time I touched Erlang, "Learn you some Erlang for great good!" was fresh out of the printers.
I loved Erlang. I want to use it, those who can master mnesia natively are a powerful wizard/ess. I've only touched the surface and seen it's raw power.
I've just not had time to learn further. As swapping between languages when trying to complete my original Tcl project is tricky. I'm now poking around with Crystal.
A rookie example of erlang and converting strings to base16. I posted in haste a while ago on Tcl's IRC but fun. Never did implement the teddy bear colour.
> almost all of Javascript's good parts come from Smalltalk
Is this underselling the role of Self in JS’s prototype-based object system, or saying that a prototype-based object system is not one of the good parts?
And some months ago I was flirting with the idea of "Smalltalk and Erlang had a baby" and have Smalltalk syntax to an actor environment ran by a Erlang VM style. Plus some insanely cool ideas about a causal debugger for distributed bugs.
The problem is that as beautiful as programing in that sound, we're not even coding anymore, AIs will take the fun of it.
Although, thinking long term and for understendabilitymaxxin and reliabilitymaxxin it might still be worth having something like that.
Since Smalltalk can be so close to english and Erlang VM so reliable, humans and LLMs should thrive on it.
I was flirting with the idea ... Plus some insanely cool ideas ...
The problem is that as beautiful as programing in that sound, we're not even coding anymore, AIs will take the fun of it.
We've come to this. The mere existence of AI is discouraging people from pursuing creative ideas. This is a catastrophe! It's not just taking the fun out of life, it's destroying alternative futures we might have built.
There is this advice for dealing with tyrants and would-be tyrants: "Do not obey in advance." In your case that means pursuing and implementing your exciting ideas.
... it might still be worth having something like that.
> we're not even coding anymore, AIs will take the fun of it.
Or... Writing a compiler from a large language test suite is already an LLM sweet spot. Writing code which is atypically pretty and readable isn't... but there's at least a possibility of that changing someday.
Imagine a future Lisp-ish culture of "write code to communicate with humans - prioritize readability - machine execution is incidental" and "don't merely solve a problem - write a language for the problem space (and keep it cleanly updated)"... but moving at LLM-catalyzed speed. So much fun.
Metalworking, lathes, precision tooling, are all way more valuable skills and tools today than they ever were before the industrial revolution.
I suspect that over time a good grasp of fundamentals and great tooling will only get more valuable in software as well as automation expands, in whichever form.
Automation expands the demand for skills. It can never diminish it. Required variety indicates that complexity doesn't go away with automation, and at some level someone will have to understand and make decisions about complex issues. You can't automate all the way down. Or, put another way, everything is already automated all the way down. The expansion of automation as we perceive it is merely the growth of complexity over new dimensions.
Or put a third way: automation means two different things: constraint and liberty. The confusion arising from not discerning which one is in force makes it seems like one can grow unbounded. Liberty requires constraint and vice versa.
I didn't understand programming at all until I read about (not even tried) Smalltalk. And it's because of Smalltalk I write Ruby today.
My favourite thing about it is the lack of "reserved words" - just a handful of punctuation marks. It's incredibly pure - almost all of the syntax is "send this message to this object".
I learned about Smalltalk from Ruby, because DHH recommended a book about Smalltalk back in the day. I really enjoy Smalltalk-style OOP, but at the same time it can be so different to Java enterprise OOP with all its patterns.
Nowadays I write a lot of TypeScript and prefer having a type system (especially for agentic coding), but Ruby is one of my favorite languages to write myself :)
I would give a lot for a version of Squeak which could instantiate standard GUI controls and easily compile to a stand-alone binary or HTML page w/ included JavaScript.
That was going to be my guess, too. Looking through some syllabi over the years it looks like they used Scala for it for a while and Python, more recently.
>Btw, almost all of Javascript's good parts come from Smalltalk.
Like what, exactly? I'm genuinely curious. There is no root object or message passing in JS, and the OO aspects are only incidentally bolted on (classes don't actually exist, prototypal inheritance was an afterthought), named params are newer sugar, and polymorphism is a struggle due to all of the above. I've always thought of JS as what Crockford himself said: "a Lisp in C's clothing".
Ignore the syntax for a moment. Minus Javascript, HTML + CSS is a green screen terminal with better fonts and image support. Message passing (fetch, XMLHttpRequest, polling, streaming, etc...) is how you build something that feels like a real tool out a system where the sandbox gets in the way of everything you want to hold locally.
The main thing you’ll learn is that actually the static and inflexible abstractions of an operating system do a lot to protect the system integrity and prevent everything from blowing up. Many smalltalk VMs rely on recovery features for a reason
recovery features? how do they work? and what smalltalk runtime do you have in mind? I still have a very vague understanding on how changes could be applied to a production running images, or if there is one and preferable way to do that.
Congrats to the Squeak 6.1 team! I was an early contributor to Squeak from when Alan Kay's group was at Apple (I see that SameGame, the first game implemented in Morphic is still in the image), and still enjoy following the progress of Squeak and all of its spinoffs.
one thing I love about Smalltalk is being able to inspect the code as it's running. Especially from the GUI. You can just be like Oh I wonder where the code for this button is from, inspect it, and it takes you right to the code. It's a shame we can't have this level of introspection without negative performance implications.
We could in Java Swing (and Visual basic, but lets not talk about that) as well. Well, not runtime, but we could design applications and just right click on the objects in the IDE and go directly to the click (or double click or right click) handlers.
Then again, not having to disable this for runtime is a feature if you ask me.
Yes, current web applications are prettier and easier to upgrade but we lost a lot on the way wrt developer experience and ux.
What are the best books/papers/blog posts to learn about Morphic's architecture? I'm not a user of any Smalltalk implementation but I'd like to learn more as the Smalltalk approach to UI is very interesting.
After Self, the short lineage is Self (~1992, Smith & Maloney) -> Squeak port (Maloney & Ingalls, Etoys/Scratch 1) -> two independent JavaScript reimplementations, plus Squeak-in-the-browser via SqueakJS.
I dug into this same question a decade ago and posted a longer treatment -- my email exchange with Alan Kay on MVC vs Morphic vs watchers, plus replies on Self's "soup of objects" UI. Good philosophical complement to the architectural links below:
Self Morphic is not a classical class hierarchy. The handbook describes parallel traits objects (shared behavior) and prototypes (structure), with instances delegating via parent* slots and "copy-down" differential prototypes. Bottom-up: morph copy, tweak, factor shared behavior into a traits object. Worth reading section 7.3 alongside pjmlp's links.
The Squeak port is single-inheritance Smalltalk classes -- closer to what most people mean by "OOP UI toolkit." Etoys is Morphic. Full Squeak Morphic in a browser, no install:
In JavaScript there are two Morphic implementations people often conflate. Same family, not the same code.
Dan Ingalls's Lively Kernel / Lively Web (~2008, Sun) is the full live system: modular core/lively/morphic/ (Halos, Serialization, Scrubbing, Connectors, constraints), plus IDE, parts bin, world persistence. JSConf 2012 demo:
Jens Monig's morphic.js (~2010, BYOB4/Snap!) is a separate codebase: single-file Canvas kernel (~13k lines), World/Hand/stepping/dirty rects, template peel-off, ScrollFrame inertial pan. Jens names Ingalls's LK as the gold standard but says it is not a direct port -- though fullCopy() was ported almost literally from Squeak, comments included. Snap bundled it from day one (2013-03-16). Standalone extract:
On multiple inheritance: neither JS port uses it, and Self Morphic didn't really either. LK cheats with an explicit Trait composition layer (Object.subclass plus Trait(...) mixins). Snap cheats with Squeak-style copying (fullCopy, isTemplate peel-off). The live feel comes from copyable morph trees and shared behavior, not MI.
If you want one more readable architecture tour beyond sin-ack's intro, the Self handbook chapter 7 is still the root document; everything else is commentary on it.
I vouched as well, but it's [flagged][dead] not just [dead] which is usually the result of user moderation, not the mods or auto moderation (which typically will just render as [dead] in either case). Which was odd because it was [flagged][dead] by the time it was 4 minutes old (maybe earlier, that's just when I saw it).
After Self, the short lineage is Self (~1992, Smith & Maloney) -> Squeak port (Maloney & Ingalls, Etoys/Scratch 1) -> two independent JavaScript reimplementations, plus Squeak-in-the-browser via SqueakJS.
I dug into this same question a decade ago and posted a longer treatment -- my email exchange with Alan Kay on MVC vs Morphic vs watchers, plus replies on Self's "soup of objects" UI. Good philosophical complement to the architectural links below:
Self Morphic is not a classical class hierarchy. The handbook describes parallel traits objects (shared behavior) and prototypes (structure), with instances delegating via parent* slots and "copy-down" differential prototypes. Bottom-up: morph copy, tweak, factor shared behavior into a traits object. Worth reading section 7.3 alongside pjmlp's links.
The Squeak port is single-inheritance Smalltalk classes -- closer to what most people mean by "OOP UI toolkit." Etoys is Morphic. Full Squeak Morphic in a browser, no install:
In JavaScript there are two Morphic implementations people often conflate. Same family, not the same code.
Dan Ingalls's Lively Kernel / Lively Web (~2008, Sun) is the full live system: modular core/lively/morphic/ (Halos, Serialization, Scrubbing, Connectors, constraints), plus IDE, parts bin, world persistence. JSConf 2012 demo:
Jens Monig's morphic.js (~2010, BYOB4/Snap!) is a separate codebase: single-file Canvas kernel (~13k lines), World/Hand/stepping/dirty rects, template peel-off, ScrollFrame inertial pan. Jens names Ingalls's LK as the gold standard but says it is not a direct port -- though fullCopy() was ported almost literally from Squeak, comments included. Snap bundled it from day one (2013-03-16). Standalone extract:
On multiple inheritance: neither JS port uses it, and Self Morphic didn't really either. LK cheats with an explicit Trait composition layer (Object.subclass plus Trait(...) mixins). Snap cheats with Squeak-style copying (fullCopy, isTemplate peel-off). The live feel comes from copyable morph trees and shared behavior, not MI.
If you want one more readable architecture tour beyond sin-ack's intro, the Self handbook chapter 7 is still the root document; everything else is commentary on it.
I just downloaded on Windows 11. My antivirus (Symantec Endpoint Protection) did not like it and erased the executable. I reinstalled and got it working.
Question? Does this work with Etoys? If so, how? I tried following various directions online without success. Thanks!
Help/Useful Expressions has some EToys instructions.
The Squeakland images stopped getting upograded when Squeeak went 64-bit (I think).
You can get a "halo" on any object, click the Inspector icon (the eye) and create an etoys script from there.
Also, under Extras/Themes and Colors there is a Set Etoys Mode option. You can try that but, it didn't work last night.
Lastly, files.squeak.org/etoys has a 6.0beta image you can download. Drop the image/changes into SqueakJS [1] (that is the only way to run a 32bit image).
And after all these years, still no fixes for high-DPI displays. The UI is still pixely and slow. Perhaps I should take Fable and try myself, this was the only bug I cared about and it is not solved after 10 years at least.
Same, I've been tracking this and Pharo for high-DPI fixes. Pharo is working on a new graphics stack that is intended to address this as well, but until then no high-DPI support.
Does Glamorous Toolkit work for you? It's pharo with a different rust based GUI stack, and some batteries included for making custom inspector views and such. I don't have a high DPI display so I'm not completely confident, but my understanding is that their rust based GUI works fine with high DPI.
I see some movement on high-DPI in the latest release. What are the current gaps? I'm not familiar with the project, just curious.
> Improves high-DPI support for buttons, scroll panes, sliders, menus, multi-selection lists, trees, drop-shadows, the scratch pad, the keyboard exerciser, and others.
The UI has been a huge impediment in my getting into Squeak development. Every couple years I decide I will try it again, and then see that very little has changed.
It's a shame. Smalltalk/Squeak/Pharo was supposed to be the future, but it feels like an IDE stuck in the 90's.
I haven't used Squeak since college but I'm glad it was part of the cirriculum.
Btw, almost all of Javascript's good parts come from Smalltalk.
reply