I love, love, love flipping through old magazines. Look, an ad for a commercial Emacs! A C compiler for just $495! A port of vi to MS-DOS for $149! A sort command for just $135! A PC card with a 68000 coprocessor for heaven knows why!
The good old days were fun for their sense of everything-is-new adventure, but there's an awful lot I don't miss.
If you listen to RMS’ talks, arguably this is a feature, not a bug.
Most people pick up on the idea of freedom as in liberty with OSS: that you should be able to amend software at will. Strangely there are limits in his mind (at one talk I attended he insisted he didn’t care about being able to amend the software in his microwave, so didn’t care if it was Free or not).
However when asked about how programmers should make a living, his economic argument is that we should be paid for our hours, not for our software. In his telling, the right economic model is not the author who writes something once and is paid forever by everyone who wants to consume that “art”, but more like the trading crafter who makes and sells things on an on-going basis.
In support of that model, the “devaluation” of software, has meant we have a planet running on it that would not have been possible if every library and application on ever machine had cost $100-$500 each. The advances in scientific and medical research powered by that software driven World would not have been achieved yet, but neither would the damage caused by social media and adtech.
I’m not sure which side of the fence I sit on. I’ve had a good career being paid to write software because of its growing influence in the World, which likely would have stalled if OSS didn’t exist. But sure, I like the idea of spending a year writing something and living the rest of my life off the proceeds, like most people would.
The return to 90's style licenses kind of makes the point of everyone realizing someone has to put into the money, capitalism doesn't work with pull requests to upstream.
I subscribe to Microcenter's emails and the one with the miscellaneous lesser priced accessories always completely loads in Gmail and it gives me a little bit of that feeling of flipping through an old computer shopper. I think those are all being scanned too and uploaded to the archive I think https://archive.org/details/computer-shopper-may-1996-images...
For the young'uns among us, imaging that you didn't have access to the Internet, but you still needed to research and build a PC from parts. Now imagine someone mails you a 896 page (!!!) magazine describing basically every part you could ever actually need or want, along with a gazillion vendors for each.
It was wonderful. It's like having the entire industry in your hands. If it's for sale, chances are Computer Shopper has it.
The magazine was nearly 100% ads and I could spend a long time doing nothing but consuming ads. Nonetheless, I never felt annoyed by them like I do with animated and pop-up ads.
Indeed. The ads were the point of Computer Shopper. The reason it worked so well is the advertisers knew their most aggressive price competitors would also have ads, so it would be pointless to advertise unless you knew you could compete. While companies always prefer to avoid such side-by-side comparisons, the number of readers of Computer Shopper was simply too large to resist for many.
I recently have also been thinking about Jef Raskin’s book The Humane Interface. It feels increasingly relevant to now.
Raskin was deeply concerned with how humans think in vague, associative, creative ways, while computers demand precision and predictability.
His goal was to humanize the machine through thoughtful interface design—minimizing modes, reducing cognitive load, and anticipating user intent.
What’s fascinating now is how AI, changes the equation entirely. Instead of rigid systems requiring exact input, we now have tools that themselves are fuzzy, and probabilistic.
I keep thinking that the gap Raskin was trying to bridge is closing—not just through interface, but through the architecture of the machine itself.
So AI makes Raskin’s vision more feasible than ever but also challenges his assumptions:
Perhaps, but I don't think we're going to see evidence of this for quite a while. It would be really cool if the computer adapted to how you naturally want to use it, though, without forcing you through an interface where you talk/type to it.
I think it does; LLMs in particular. AI also enables a ton of other things, many of them inhumane, which can make it very hard to discuss these things as people fixate on the inhumane. (Which is fair... but if you are BUILDING something, I think it's best to fixate on the humane so that you conjure THAT into being.)
I think Jef Raskin's goal with a lot of what he proposed was to connect the computer interface more directly with the user's intent. An application-oriented model really focuses so much of the organization around the software company's intent and position, something that follows us fully into (most of) today's interfaces.
A magical aspect of LLMs is that they can actually fully vertically integrate with intent. It doesn't mean every LLM interface exposes this or takes advantage of this (quite the contrary!), but it's _possible_, and it simple wasn't possible in the past.
For instance: you can create an LLM-powered piece of software that collects (and allows revision) to some overriding intent. Just literally take the user's stated intent and puts it in a slot in all following prompts. This alone will have a substantial effect on the LLMs behavior! And importantly you can ask for their intent, not just their specific goal. Maybe I want to build a shed, and I'm looking up some materials... the underlying goal can inform all kinds of things, like whether I'm looking for used or new materials, aesthetic or functional, etc.
To accomplish something with a computer we often thread together many different tools. Each of them is generally defined by their function (photo album, email client, browser-that-contains-other-things, and so on). It's up to the human to figure out how to assemble these, and at each step it's easy to lose track, to become distracted or confused, to lose track of context. And again an LLM can engage with the larger task in a way that wasn't possible before.
Tell me, how does doing any of the things you've suggested help with the huge range of computer-driven tasks that have nothing to do with language? Video editing, audio editing, music composition, architectural and mechanical design, the list is vast and nearly endless.
LLMs have no role to play in any of that, because their job is text generation. At best, they could generate excerpts from a half-imagined user manual ...
Because some LLMs are now multimodal—they can process and generate not just text, but also sound and visuals. In other words, they’re beginning to handle a broader range of human inputs and outputs, much like we do.
Those are not LLMs. They use the same foundational technology (pick what you like, but I'd say transformers) to accomplish tasks that require entirely different training data and architectures.
I was specifically asking about LLMs because the comment I replied to only talked about LLMs - Large Language Models.
At this point in time calling a multimodal LLM an LLM is pretty uncontroversial. Most of the differences lie in the encoders and embedding projections. If anything I'd think MoE models are actually more different from a basic LLM than a multimodal LLM is from a regular LLM.
Bottom line is that when folks are talking about LLM applications, multimodal LLMs, MoE LLMs, and even agents are all in the general umbrella.
Everything has to do with language! Language is a way of stating intention, of expression something before it exists, of talking about goals and criteria. Everything example you give can be described in language. You are caught up in the mechanisms of these tools, not the underlying intention.
You can describe your intention in any of these tools. And it can be whatever you want... maybe your intention in an audio editor is "I need to finish this before the deadline in the morning but I have no idea what the client wants" and that's valid, that's something an LLM can actually work with.
HOW the LLM is involved is an open question, something that hasn't been done very well, and may not work well when applied to existing applications. But an LLM can make sense of events and images in addition to natural language text. You can give an LLM a timestamped list of UI events and it can actually infer quite a bit about what the user is actually doing. What does it do with that understanding? We're going to have to figure that out! These are exciting times!
What if you could pilot your video editing tool through voice? Have a multimodal LLM convert your instructions into some structured data instruction that gets used by the editor to perform actions.
Zooming is a bad example (because pinch zoom is just so much better than that scene hah.) Instead "go back 5 frames, and change the color grading. Make the mood more pensive and bring out blues and magentas and fewer yellows and oranges." That's a lot faster than fiddling with 2-3 different sliders IMO.
> Zooming is a bad example (because pinch zoom is just so much better than that scene hah.) Instead "go back 5 frames, and change the color grading. Make the mood more pensive and bring out blues and magentas and fewer yellows and oranges." That's a lot faster than fiddling with 2-3 different sliders IMO.
Eh. That's not as good as being skilled enough to know exactly what you want and have the tools to make that happen.
There's something to be said for tools that give you the power of manipulating something efficiently, than systems that do the manipulation for you.
> Eh. That's not as good as being skilled enough to know exactly what you want and have the tools to make that happen.
I mean, do you know that? A tool that offers this audible fluent experience needs to exist before you can make that assessment right? Or are vibes alone a strong enough way to make this judgement? (There's also some strong "Less space than a Nomad. Lame" energy in this post lol.)
Moreover why can't you just have both? When I fire up Lightroom, sure I have easy mode sliders to affect "warmth" but then I have detailed panels that let me control the hue and saturation of midtones. And if those panels aren't enough I can fire up Photoshop and edit to my heart's content.
Nothing is stopping you from taking your mouse in hand at any point and saying "let me do it" and pausing the LLM to let you handle the hard bits. The same way programmers rely on compliers to generate most machine or VM code and only write machine code when the compiler isn't doing what the programmer wants.
Training LLMs to generate some internal command structure for a tool is conceptually similar to what we've done with them already, but the training data for it is essentially non-existent, and would be hard to generate.
My experience has been that generating structured output with zero, one, and few-shot prompts works quite well. We've used it at $WORK for zero-shot stuff and it's been good enough. I've done few-shot prompting for some personal projects and it's been solid. JSON Schema based enforcement of responses with temperature 0 settings works quite well. Sometimes LLMs hallucinate their responses but if you keep output formats fairly constrained (e.g. structured dicts of booleans) it decreases hallucinations and even when they do hallucinate, at temperature 0 it seems to stay within < 0.1% of responses even with zero-shot prompting. (At least with datasets and prompts I've considered.)
(Though yes, keep in mind that 0.1% hallucination = 99.9% correctness which is really not that high when we're talking about high reliability things. With zero-shot that far exceeded my expectations though.)
This is something I keep tossing over in my head. Multimodal capabilities of frontier models right now are fantastic. Rather than locking into a desktop with peripherals or hunching over a tiny screen and tapping with thumbs we finally have an easy way to create apps that interact "natively" through audio. We can finally try to decipher a user's intent rather than forcing the user to interact through an interface designed to provide precise inputs to an algorithm. I'm excited to see what we build with these things.
"no" .. intelligent appliance was the product that came out of Raskin's thinking..
I object to the framing of this question directly -- there is no definition of "AI" . Secondly, the humane interface is a genre that Jef Raskin shaped and re-thought over years.. A one-liner here definitely does not embody the works of Jef Raskin.
Off the top of my head, it appears that "AI" enables one-to-many broadcast, service interactions and knowledge retrieval in a way that was not possible before. The thinking of Jef Raskin was very much along the lines of an ordinary person using computers for their own purposes. "AI" in the supply-side format coming down the road, appears to be headed towards societal interactions that depersonalize and segregate individual people. It is possible to engage "AI" whatever that means, to enable individuals as an appliance. This is by no means certain at this time IMHO.
Oh, I haven't heard the name "Dr. Dobb's" for quite a few years now.
I remember when they finally shut it down, they promised to freeze the website and preserve it in a read-only format, for all the computing history it contains. https://drdobbs.com/ certainly looks the part, but I'm afraid something seems to be fundamentally broken, because every article links just leads to an error message.
When I was in my late teens, right about the time this article came out, I was an ignorant, naive, first-time computer user. My college had some Canon Cats in the computer lab. I didn't know the first thing about computers, and I didn't understand the difference between the PCs and Macs in the rest of the lab and the Canon Cats. There was never a line to use a Canon Cat, so I tried it. By Raskin's standards, I should have been the perfect Cat user. Being completely ignorant, I had no preconceived ideas about how a computer should operate. I found the Cat utterly incomprehensible. Then someone demonstrated a Macintosh to me, and I immediately understood it. Take that anecdote as you will.
I'm curious if the manual was made available to you. The reason I ask is that in the linked article he said the manual was part of the user interface. Also, from what I recall, his opinion on intuitive was that it was more important that it worked well after learning how to use it, vs somehow knowing it without being taught. (I'll double check my copy of The Humane Interface when I get back to it.)
edit: back, here's a quote:
> [...] I asked people unfamiliar with the mouse to use a Macintosh. My protocol was to run a [game that only used clicking, with the keyboard removed]. I would point to the mouse and say, "This is the mouse that you use to operate the game. Go ahead, give it a try." If asked any questions, I'd say something nonspecific, such as "Try it." The reaction of an intelligent Finnish educator who had never seen a Macintosh but was otherwise computer literate was typical: she picked up the mouse.
> Nowadays, the might seem absurd, but [mentions the scene in Star Trek where Scotty does the same thing]. In the case of my Finnish subject, her next move was to turn the mouse over and to try rolling the ball. Nothing happened. She shook the mouse, and then she held the mouse in one hand and clicked the button with the other. No effect. Eventually, she succeeded in operating the game by holding the mouse in her right hand, rolling the ball on the bottom with her fingers, and clicking the button with her left hand.
> These experiments make the point that an interface's ease of use and speed of learning are not connected with the imagined properties of intuitiveness and naturalness. The mouse is very easy to learn: All I had to do, with any of the test subjects, was to put the mouse on the desk, move it, and click on something. In five to ten seconds, they learned how to use the mouse. That's fast and easy, but it is neither intuitive nor natural. No artifact is.
> The belief that interfaces can be intuitive and natural is often detrimental to improved interface design. As a consultant, I am frequently asked to design a "better" interface to a product. Usually, an interface can be designed such that, in terms of learning time, eventual speed of operation (productivity), decreased error rates, and easy of implementation, it is superior to both the client's existing products and completing products. Nonetheless, even when my proposals are seen as significant improvements, they are often rejected on the grounds that they are not intuitive. [He goes on to talk about how if it going to be significantly better than it will end up being different than what people currently know, but the clients still want it to be similar to Windows...]
The Humane Interface section 6-1
Having refreshed myself on what he said, and re-reading what you wrote, I don't think he would say that you should be able to walk up to his computer without having someone show you how to use it, or looking at a manual. And as you said: "Then someone demonstrated a Macintosh to me" just like when he said he'd show people how to use the mouse.
> Let me make a typical error. I want to move the cursor to the word good, so I should press the left Leap key and type “good.” I'll press the right Leap key and type. It found it anyway. The system does one thing that all systems should have done from day 1: If you tell it to search one way for something and it doesn’t find it, it searches the other way in case you made a mistake. Most systems didn’t do this because if you did find it then you’ve lost your place. In this system if you want to go back, you just bang on the keyboard. (Raskin slams both hands on the keyboard, and the cursor returns to the point in the document at which his search began.)
We're supposed to idolize this as some sort of hyper-enlightened version of interface design? Hell no.
I get that this design worked for Raskin. It worked for him the same way that my hacked version of GNU Emacs' next-line function does for me when the cursor is at the end of the buffer, or how I needed a version of its delete-horizontal-space but that would work only after the cursor.
I get that Raskin's "oh, you probably made a mistake, let me do something else that I suspect is what you really meant" might even have worked for a bunch of other people too. But the idea that somehow Raskin had some sort of deep insight with stuff like this, rather than just a set of personal preferences that are like those of anyone else, is just completely wrong, I think.
You're making the error of judging Raskin's approach with the knowledge of user interfaces that a person in 2025 has. It's been 40 years since that interview.
Many people today weren't even born yet.
In that 40 years, many UI conventions have sprung up, and we've internalized them to the point that they're so familiar we actually say they're intuitive.
But if you go back to the state of computing in 1986, or even earlier, when Raskin was developing his UX principles for the Canon Cat and the SwyftCard, he was considering computer interfaces that were almost exclusively command-line interfaces.
You're not supposed to "idolize" any designer or engineer. But I would highly encourage you to read The Humane Interface, learn about the underlying principles of usability and interface design, and consider how you'd apply them to a UI today, 40 years later. The execution you'd come up with would be different. But the principles he started from are foundational and very useful.
Emacs and a couple of key plugins can get you pretty darned close to the Cat interface. But here's the thing, Emacs is still behind the Open Genera interface which I believe, predates the Cat. And the extensability of the Lisp Machine/Emacs is superior to that of the Cat.
I have a geographical memory, that worked well with paper and books, but I haven't printed more than two pages in the last 7 years and none of my work uses books anymore (I did buy a book for academic study last year).
Zooming, from a building, to a room, to a bookshelf, to a book/folder/boxfile, to the content and the location within the content worked with my brain. With digital files it just seems like a swamp I have to wade through. Microsoft are so antagonistic to my 'location' based thinking because Windows conceals where files really are.
Completely agree. An infinite canvas sounds great until you need to find anything, then you end up with having to create a structure to it, and you rapidly end up grouping stuff into "pages" - Miro's "frames".
Further, I have read several sections of The Humane Interface, and I think it does contain some real insight, some of which we have unfortunately lost.
But I do not think that Raskin was channelling some remarkable stream of insight into these matters. And yes, "idolize" was more poking fun at people who use superlatives to describe him, in my opinion without much justification.
I read his book during my early years of software development, as a big fan of the (early) Mac and with a real passion to build better user experiences in industrial software.
I often wonder what Jef would think of the iPhone (or even the current Mac), if he were still with us. I suspect he'd be deeply disappointed.
Jef Raskin got cut out of the Macintosh project by Steve Jobs and he held a grudge about that. Unclear if he'd be deeply disappointed for personal or technical reasons.
I'm currently working on making a custom keyboard with the leap keys below the space bar like he made for the Canon Cat, and planning on using it with either a microcontroller or 65c02 (edit: or a single board computer) and building up the software in Forth, so this was a really fun article to run across.
What an interesting piece of history. Imagine times when users were reading user manuals for the software/hardware they used. RTFM is already a thing of the past.
I love, love, love flipping through old magazines. Look, an ad for a commercial Emacs! A C compiler for just $495! A port of vi to MS-DOS for $149! A sort command for just $135! A PC card with a 68000 coprocessor for heaven knows why!
The good old days were fun for their sense of everything-is-new adventure, but there's an awful lot I don't miss.
Remember that these are 1986s dollars, worth nearly $3 today.
You don't miss being able to make a living from writing your own software?
That part sounds pretty wonderful, but it also means you were paying through the teeth for everyone else’s software, too.
There are still plenty of small shops making a nice software living today. I don’t have the nerve to do it myself, though.
No one ever talks about how OSS literally devalued software, which resulted in the best hackers being forced to take corporate jobs.
If you listen to RMS’ talks, arguably this is a feature, not a bug.
Most people pick up on the idea of freedom as in liberty with OSS: that you should be able to amend software at will. Strangely there are limits in his mind (at one talk I attended he insisted he didn’t care about being able to amend the software in his microwave, so didn’t care if it was Free or not).
However when asked about how programmers should make a living, his economic argument is that we should be paid for our hours, not for our software. In his telling, the right economic model is not the author who writes something once and is paid forever by everyone who wants to consume that “art”, but more like the trading crafter who makes and sells things on an on-going basis.
In support of that model, the “devaluation” of software, has meant we have a planet running on it that would not have been possible if every library and application on ever machine had cost $100-$500 each. The advances in scientific and medical research powered by that software driven World would not have been achieved yet, but neither would the damage caused by social media and adtech.
I’m not sure which side of the fence I sit on. I’ve had a good career being paid to write software because of its growing influence in the World, which likely would have stalled if OSS didn’t exist. But sure, I like the idea of spending a year writing something and living the rest of my life off the proceeds, like most people would.
The return to 90's style licenses kind of makes the point of everyone realizing someone has to put into the money, capitalism doesn't work with pull requests to upstream.
"Borland introduces Turbo Prolog, the natural language of Artificial Intelligence" just $99.95!
I subscribe to Microcenter's emails and the one with the miscellaneous lesser priced accessories always completely loads in Gmail and it gives me a little bit of that feeling of flipping through an old computer shopper. I think those are all being scanned too and uploaded to the archive I think https://archive.org/details/computer-shopper-may-1996-images...
For the young'uns among us, imaging that you didn't have access to the Internet, but you still needed to research and build a PC from parts. Now imagine someone mails you a 896 page (!!!) magazine describing basically every part you could ever actually need or want, along with a gazillion vendors for each.
And you got a new version every month.
It was pure magic, I tell you.
It was wonderful. It's like having the entire industry in your hands. If it's for sale, chances are Computer Shopper has it.
The magazine was nearly 100% ads and I could spend a long time doing nothing but consuming ads. Nonetheless, I never felt annoyed by them like I do with animated and pop-up ads.
Indeed. The ads were the point of Computer Shopper. The reason it worked so well is the advertisers knew their most aggressive price competitors would also have ads, so it would be pointless to advertise unless you knew you could compete. While companies always prefer to avoid such side-by-side comparisons, the number of readers of Computer Shopper was simply too large to resist for many.
I recently have also been thinking about Jef Raskin’s book The Humane Interface. It feels increasingly relevant to now.
Raskin was deeply concerned with how humans think in vague, associative, creative ways, while computers demand precision and predictability.
His goal was to humanize the machine through thoughtful interface design—minimizing modes, reducing cognitive load, and anticipating user intent.
What’s fascinating now is how AI, changes the equation entirely. Instead of rigid systems requiring exact input, we now have tools that themselves are fuzzy, and probabilistic.
I keep thinking that the gap Raskin was trying to bridge is closing—not just through interface, but through the architecture of the machine itself.
So AI makes Raskin’s vision more feasible than ever but also challenges his assumptions:
Does AI finally enable truly humane interfaces?
> Does AI finally enable truly humane interfaces?
Perhaps, but I don't think we're going to see evidence of this for quite a while. It would be really cool if the computer adapted to how you naturally want to use it, though, without forcing you through an interface where you talk/type to it.
"Does AI finally enable truly humane interfaces?"
I think it does; LLMs in particular. AI also enables a ton of other things, many of them inhumane, which can make it very hard to discuss these things as people fixate on the inhumane. (Which is fair... but if you are BUILDING something, I think it's best to fixate on the humane so that you conjure THAT into being.)
I think Jef Raskin's goal with a lot of what he proposed was to connect the computer interface more directly with the user's intent. An application-oriented model really focuses so much of the organization around the software company's intent and position, something that follows us fully into (most of) today's interfaces.
A magical aspect of LLMs is that they can actually fully vertically integrate with intent. It doesn't mean every LLM interface exposes this or takes advantage of this (quite the contrary!), but it's _possible_, and it simple wasn't possible in the past.
For instance: you can create an LLM-powered piece of software that collects (and allows revision) to some overriding intent. Just literally take the user's stated intent and puts it in a slot in all following prompts. This alone will have a substantial effect on the LLMs behavior! And importantly you can ask for their intent, not just their specific goal. Maybe I want to build a shed, and I'm looking up some materials... the underlying goal can inform all kinds of things, like whether I'm looking for used or new materials, aesthetic or functional, etc.
To accomplish something with a computer we often thread together many different tools. Each of them is generally defined by their function (photo album, email client, browser-that-contains-other-things, and so on). It's up to the human to figure out how to assemble these, and at each step it's easy to lose track, to become distracted or confused, to lose track of context. And again an LLM can engage with the larger task in a way that wasn't possible before.
Tell me, how does doing any of the things you've suggested help with the huge range of computer-driven tasks that have nothing to do with language? Video editing, audio editing, music composition, architectural and mechanical design, the list is vast and nearly endless.
LLMs have no role to play in any of that, because their job is text generation. At best, they could generate excerpts from a half-imagined user manual ...
Because some LLMs are now multimodal—they can process and generate not just text, but also sound and visuals. In other words, they’re beginning to handle a broader range of human inputs and outputs, much like we do.
Those are not LLMs. They use the same foundational technology (pick what you like, but I'd say transformers) to accomplish tasks that require entirely different training data and architectures.
I was specifically asking about LLMs because the comment I replied to only talked about LLMs - Large Language Models.
Multimodal LLMs are absolutely LLMs, the language is just not human language.
At this point in time calling a multimodal LLM an LLM is pretty uncontroversial. Most of the differences lie in the encoders and embedding projections. If anything I'd think MoE models are actually more different from a basic LLM than a multimodal LLM is from a regular LLM.
Bottom line is that when folks are talking about LLM applications, multimodal LLMs, MoE LLMs, and even agents are all in the general umbrella.
Everything has to do with language! Language is a way of stating intention, of expression something before it exists, of talking about goals and criteria. Everything example you give can be described in language. You are caught up in the mechanisms of these tools, not the underlying intention.
You can describe your intention in any of these tools. And it can be whatever you want... maybe your intention in an audio editor is "I need to finish this before the deadline in the morning but I have no idea what the client wants" and that's valid, that's something an LLM can actually work with.
HOW the LLM is involved is an open question, something that hasn't been done very well, and may not work well when applied to existing applications. But an LLM can make sense of events and images in addition to natural language text. You can give an LLM a timestamped list of UI events and it can actually infer quite a bit about what the user is actually doing. What does it do with that understanding? We're going to have to figure that out! These are exciting times!
What if you could pilot your video editing tool through voice? Have a multimodal LLM convert your instructions into some structured data instruction that gets used by the editor to perform actions.
Compare pinch zoom to the tedious scene in Bladerunner where Deckard is asking the computer to zoom in to a picture.
Zooming is a bad example (because pinch zoom is just so much better than that scene hah.) Instead "go back 5 frames, and change the color grading. Make the mood more pensive and bring out blues and magentas and fewer yellows and oranges." That's a lot faster than fiddling with 2-3 different sliders IMO.
> Zooming is a bad example (because pinch zoom is just so much better than that scene hah.) Instead "go back 5 frames, and change the color grading. Make the mood more pensive and bring out blues and magentas and fewer yellows and oranges." That's a lot faster than fiddling with 2-3 different sliders IMO.
Eh. That's not as good as being skilled enough to know exactly what you want and have the tools to make that happen.
There's something to be said for tools that give you the power of manipulating something efficiently, than systems that do the manipulation for you.
> Eh. That's not as good as being skilled enough to know exactly what you want and have the tools to make that happen.
I mean, do you know that? A tool that offers this audible fluent experience needs to exist before you can make that assessment right? Or are vibes alone a strong enough way to make this judgement? (There's also some strong "Less space than a Nomad. Lame" energy in this post lol.)
Moreover why can't you just have both? When I fire up Lightroom, sure I have easy mode sliders to affect "warmth" but then I have detailed panels that let me control the hue and saturation of midtones. And if those panels aren't enough I can fire up Photoshop and edit to my heart's content.
Nothing is stopping you from taking your mouse in hand at any point and saying "let me do it" and pausing the LLM to let you handle the hard bits. The same way programmers rely on compliers to generate most machine or VM code and only write machine code when the compiler isn't doing what the programmer wants.
So again, why not?
Training LLMs to generate some internal command structure for a tool is conceptually similar to what we've done with them already, but the training data for it is essentially non-existent, and would be hard to generate.
My experience has been that generating structured output with zero, one, and few-shot prompts works quite well. We've used it at $WORK for zero-shot stuff and it's been good enough. I've done few-shot prompting for some personal projects and it's been solid. JSON Schema based enforcement of responses with temperature 0 settings works quite well. Sometimes LLMs hallucinate their responses but if you keep output formats fairly constrained (e.g. structured dicts of booleans) it decreases hallucinations and even when they do hallucinate, at temperature 0 it seems to stay within < 0.1% of responses even with zero-shot prompting. (At least with datasets and prompts I've considered.)
(Though yes, keep in mind that 0.1% hallucination = 99.9% correctness which is really not that high when we're talking about high reliability things. With zero-shot that far exceeded my expectations though.)
Deckard. Blade Runner.
> Does AI finally enable truly humane interfaces?
This is something I keep tossing over in my head. Multimodal capabilities of frontier models right now are fantastic. Rather than locking into a desktop with peripherals or hunching over a tiny screen and tapping with thumbs we finally have an easy way to create apps that interact "natively" through audio. We can finally try to decipher a user's intent rather than forcing the user to interact through an interface designed to provide precise inputs to an algorithm. I'm excited to see what we build with these things.
Highly recommended, timeless read!
"no" .. intelligent appliance was the product that came out of Raskin's thinking..
I object to the framing of this question directly -- there is no definition of "AI" . Secondly, the humane interface is a genre that Jef Raskin shaped and re-thought over years.. A one-liner here definitely does not embody the works of Jef Raskin.
Off the top of my head, it appears that "AI" enables one-to-many broadcast, service interactions and knowledge retrieval in a way that was not possible before. The thinking of Jef Raskin was very much along the lines of an ordinary person using computers for their own purposes. "AI" in the supply-side format coming down the road, appears to be headed towards societal interactions that depersonalize and segregate individual people. It is possible to engage "AI" whatever that means, to enable individuals as an appliance. This is by no means certain at this time IMHO.
Dude is responsible for one-button mouse ...
Oh, I haven't heard the name "Dr. Dobb's" for quite a few years now.
I remember when they finally shut it down, they promised to freeze the website and preserve it in a read-only format, for all the computing history it contains. https://drdobbs.com/ certainly looks the part, but I'm afraid something seems to be fundamentally broken, because every article links just leads to an error message.
When I was in my late teens, right about the time this article came out, I was an ignorant, naive, first-time computer user. My college had some Canon Cats in the computer lab. I didn't know the first thing about computers, and I didn't understand the difference between the PCs and Macs in the rest of the lab and the Canon Cats. There was never a line to use a Canon Cat, so I tried it. By Raskin's standards, I should have been the perfect Cat user. Being completely ignorant, I had no preconceived ideas about how a computer should operate. I found the Cat utterly incomprehensible. Then someone demonstrated a Macintosh to me, and I immediately understood it. Take that anecdote as you will.
I'm curious if the manual was made available to you. The reason I ask is that in the linked article he said the manual was part of the user interface. Also, from what I recall, his opinion on intuitive was that it was more important that it worked well after learning how to use it, vs somehow knowing it without being taught. (I'll double check my copy of The Humane Interface when I get back to it.)
edit: back, here's a quote:
> [...] I asked people unfamiliar with the mouse to use a Macintosh. My protocol was to run a [game that only used clicking, with the keyboard removed]. I would point to the mouse and say, "This is the mouse that you use to operate the game. Go ahead, give it a try." If asked any questions, I'd say something nonspecific, such as "Try it." The reaction of an intelligent Finnish educator who had never seen a Macintosh but was otherwise computer literate was typical: she picked up the mouse.
> Nowadays, the might seem absurd, but [mentions the scene in Star Trek where Scotty does the same thing]. In the case of my Finnish subject, her next move was to turn the mouse over and to try rolling the ball. Nothing happened. She shook the mouse, and then she held the mouse in one hand and clicked the button with the other. No effect. Eventually, she succeeded in operating the game by holding the mouse in her right hand, rolling the ball on the bottom with her fingers, and clicking the button with her left hand.
> These experiments make the point that an interface's ease of use and speed of learning are not connected with the imagined properties of intuitiveness and naturalness. The mouse is very easy to learn: All I had to do, with any of the test subjects, was to put the mouse on the desk, move it, and click on something. In five to ten seconds, they learned how to use the mouse. That's fast and easy, but it is neither intuitive nor natural. No artifact is.
> The belief that interfaces can be intuitive and natural is often detrimental to improved interface design. As a consultant, I am frequently asked to design a "better" interface to a product. Usually, an interface can be designed such that, in terms of learning time, eventual speed of operation (productivity), decreased error rates, and easy of implementation, it is superior to both the client's existing products and completing products. Nonetheless, even when my proposals are seen as significant improvements, they are often rejected on the grounds that they are not intuitive. [He goes on to talk about how if it going to be significantly better than it will end up being different than what people currently know, but the clients still want it to be similar to Windows...]
The Humane Interface section 6-1
Having refreshed myself on what he said, and re-reading what you wrote, I don't think he would say that you should be able to walk up to his computer without having someone show you how to use it, or looking at a manual. And as you said: "Then someone demonstrated a Macintosh to me" just like when he said he'd show people how to use the mouse.
> Let me make a typical error. I want to move the cursor to the word good, so I should press the left Leap key and type “good.” I'll press the right Leap key and type. It found it anyway. The system does one thing that all systems should have done from day 1: If you tell it to search one way for something and it doesn’t find it, it searches the other way in case you made a mistake. Most systems didn’t do this because if you did find it then you’ve lost your place. In this system if you want to go back, you just bang on the keyboard. (Raskin slams both hands on the keyboard, and the cursor returns to the point in the document at which his search began.)
We're supposed to idolize this as some sort of hyper-enlightened version of interface design? Hell no.
I get that this design worked for Raskin. It worked for him the same way that my hacked version of GNU Emacs' next-line function does for me when the cursor is at the end of the buffer, or how I needed a version of its delete-horizontal-space but that would work only after the cursor.
I get that Raskin's "oh, you probably made a mistake, let me do something else that I suspect is what you really meant" might even have worked for a bunch of other people too. But the idea that somehow Raskin had some sort of deep insight with stuff like this, rather than just a set of personal preferences that are like those of anyone else, is just completely wrong, I think.
You're making the error of judging Raskin's approach with the knowledge of user interfaces that a person in 2025 has. It's been 40 years since that interview. Many people today weren't even born yet.
In that 40 years, many UI conventions have sprung up, and we've internalized them to the point that they're so familiar we actually say they're intuitive.
But if you go back to the state of computing in 1986, or even earlier, when Raskin was developing his UX principles for the Canon Cat and the SwyftCard, he was considering computer interfaces that were almost exclusively command-line interfaces.
You're not supposed to "idolize" any designer or engineer. But I would highly encourage you to read The Humane Interface, learn about the underlying principles of usability and interface design, and consider how you'd apply them to a UI today, 40 years later. The execution you'd come up with would be different. But the principles he started from are foundational and very useful.
Emacs and a couple of key plugins can get you pretty darned close to the Cat interface. But here's the thing, Emacs is still behind the Open Genera interface which I believe, predates the Cat. And the extensability of the Lisp Machine/Emacs is superior to that of the Cat.
You're making the mistake of thinking I wasn't using computers in 1986 :)
I used GNU Emacs as an example for precisely this reason.
Raskin was a fan of "zoomable interfaces" as I recall. Remember reading about a huge canvas which you navigate and can zoom in and out of.
Today we have Miro and it works like that. I hate it.
I have a geographical memory, that worked well with paper and books, but I haven't printed more than two pages in the last 7 years and none of my work uses books anymore (I did buy a book for academic study last year).
Zooming, from a building, to a room, to a bookshelf, to a book/folder/boxfile, to the content and the location within the content worked with my brain. With digital files it just seems like a swamp I have to wade through. Microsoft are so antagonistic to my 'location' based thinking because Windows conceals where files really are.
Completely agree. An infinite canvas sounds great until you need to find anything, then you end up with having to create a structure to it, and you rapidly end up grouping stuff into "pages" - Miro's "frames".
Further, I have read several sections of The Humane Interface, and I think it does contain some real insight, some of which we have unfortunately lost.
But I do not think that Raskin was channelling some remarkable stream of insight into these matters. And yes, "idolize" was more poking fun at people who use superlatives to describe him, in my opinion without much justification.
I read his book during my early years of software development, as a big fan of the (early) Mac and with a real passion to build better user experiences in industrial software. I often wonder what Jef would think of the iPhone (or even the current Mac), if he were still with us. I suspect he'd be deeply disappointed.
Jef Raskin got cut out of the Macintosh project by Steve Jobs and he held a grudge about that. Unclear if he'd be deeply disappointed for personal or technical reasons.
He was selling a product called swyftcard for Apple II, see https://hackaday.com/2014/04/06/vcf-east-the-swyft-card/
I'm currently working on making a custom keyboard with the leap keys below the space bar like he made for the Canon Cat, and planning on using it with either a microcontroller or 65c02 (edit: or a single board computer) and building up the software in Forth, so this was a really fun article to run across.
The article in a scan of the magazine at archive.org: https://archive.org/details/1986-05-dr-dobbs-journal/page/32...
It's amazing how much his son, Aza Raskin, looks like him!
By definition, an operating system is the program you have to fight with before you can fight with an application.
One for the quote file.
What an interesting piece of history. Imagine times when users were reading user manuals for the software/hardware they used. RTFM is already a thing of the past.
Cookie popup keeps popping up if I allow only necessary cookies on an article about frustrating interactions with technology. (FF on Android)
"Our name for the wordprocessing program you get with the Macintosh is Macwait. If a little clock ever appears on a computer of mine, I'll shoot it."
1995 article by Raskin (page 9) in the Journal of the Computer History Association of California: <http://www.bitsavers.org/pdf/chac/CHAC_Analytical_Engine/2.4...>
I suspect that Jef Raskin would not be down with "prompt engineering' at all.
I think that Mr Raskin's opinion would be that it should be obvious how to use a piece of software.
Prompt engineering, it seems to me, is about the most obvious way to use a computer: Tell it in plain English what you want.
you are assuming clarity and good faith on the other side of the model providers
Wow. EGA graphics on the cover. That's not something I've heard about in a while...
Someone built a legit commercial game with EGA graphics just last year: https://store.steampowered.com/app/1098770/The_Crimson_Diamo...
I haven't played it, but looking at the graphics alone brings up some deep feelings of nostalgia
Yeah, she really likes her EGA palette. That's cool.