Yes 100% this. A lot of people keep talking about how OpenAI and Anthropic will need to raise their prices. What is less discussed is how they CAN'T raise their prices because competition exists, and sure it's not SOTA, but it's literally an order of magnitude cheaper in many cases and the drive to figure out how to make it work well enough is going on right now (and will only intensify when the SOTA models raise their price).
It's a given that the SOTA models need to raise their prices. It's also a given that they can't. The more they raise the more customers will move to their competition.
So what happens next? Well I think it will suck horribly if you can't move off of SOTA sooner or later, because the Big Two are going to lose customers, and therefore have to raise prices on the locked in customers even more than these projections suggest.
Beyond that if you're looking to start a business, figure out how to use cheap models in new scenarios. Build software which does that and license it. This is kind of contrary to the idea that you shouldn't over optimize for deficiencies in the models that will likely go away in the next generation - for instance a lot of problems were solved when context windows got way bigger. So it's a thin line to walk but I think it's there because a lot of orgs are using Claude today for pretty basic tasks.
The dev who's addicted to SOTA models honestly is going to have to settle for less or get totally screwed. Most applications within business from what I see aside from complex research do not require SOTA. They summarize, they classify, they transform, and doing that accurately has been cheap for a while.
I'd qualify your point that Anthropic and OpenAI can't raise prices, that is as of right now. Once the industry will go through a phase of consolidation and the bigger players will have some moat around their product, they'll have more pricing power.
Your last point is common sense in my opinion, I agree with it. At the end of the day most employees are (by definition) of average intelligence and most businesses are average in complexity. Thus, it is logical that average tools (AI models) should do the job for most people and most businesses.
I essentially owe my career to two great strokes of luck.
The first was that my father purchased a PC in the early 1990s to help out with his self-employed publishing business, and like most PCs of the time, it came preloaded with QBasic and the source code for a couple of games like GORILLA.BAS that an introverted kid with a lot of free time could mess around with.
The second was attending a high school with a reasonably well funded computer lab and an unusually open minded computer teacher. If you demonstrated that you were dependable, he'd basically let you do whatever you want. While my school was mostly a Mac shop, I was a bit of a Microsoft shill in high school so by graduation I'd figured out how to stand up and run a Windows NT file & web server for our school newspaper. Another guy was a Linux nut and had been allowed to do something similar with RedHat for the school's drafting lab.
Inclination met opportunity, one thing led to another, and I went on to work with technology for the next 25 years of my life.
What worries me now is that so much technology is so locked down. It must be a very rare school today that allows the kind of freedom we had. There is no IDE preinstalled on a phone, and even merely installing an "unapproved" app is under fire.
If for no other reason, for the sake of the kids the industry, the tools, the operating systems need to be more open. They need to be tinkerable. That's how the most motivated kids tend to learn. Our best and brightest are not being made because we've closed things down to maximize some hedge fund's ROI somewhere. The financialization of America was a grave error.
I've heard this story several times before and I, myself, had an unusually open minded computer teacher as a youth. I'm beginning to suspect this attribute is is not so unusual amongst computer teachers as has been assumed.
A silver lining is that Javascript code will run on any modern browser+OS and can be created with nothing but a text editor. Even though this is many degrees of abstraction from facing the bare metal that was there in the 80s and 90s, it's better than nothing.
>I essentially owe my career to two great strokes of luck.
>The first was that my father purchased a PC in the early 1990s
>The second was attending a high school with a reasonably well funded computer lab
So, you essentially own your career to your parents being well-off and tech inclined.
It's not like you ended up in that high school by accident.
Sure, being born in such a family is a stroke of luck that many people don't get to have.
I did; my mom was a software engineer in the USSR, and I grew up in the 1990s Ukraine with a PC at home, and went to a great high school in Odesa, and later, in Brooklyn when we immigrated.
Like @susam, I played Digger on IBM PC 286 as my first game when I was 4.
I have a PhD in math and Google/Meta/MS on my resume today.
I owe this to many strokes of luck, but how tinkerable the PCs were was not the most significant one by far.
The most important part was access to production tech, seeing it used, and having a role model that made it a natural consideration as a career choice.
And the "luck" of what was available in my K12 was 100% the work of my parents who got me into those schools.
My parents were broke. I grew up on food stamps. Mom worked retail. Dad had a few good years, his business went bust and we spent a decade below the poverty line.
It was a public high school in the USA. It was in a rich neighborhood on the other side of town; our neighborhood was below median income and had one of the worst high schools in the state. My parents encouraged me to apply for attendance at the school across town via a magnet program. I got in and took a public bus 80 minutes each way for 4 years.
I'm not relating this in order to validate precisely what my Privilege Index was. Rather to relate how public investment in resources which were available to anyone who was willing to make a bit of extra effort transformed at least one kid's life. It seems these days that public resources go mostly to those with the most money, or maybe those who were born into the politically correct group du jour, but almost never to the random kid who just wants to take a shot at doing something bigger.
I'm glad I was born when I was. Public policy has changed, that magnet program is now gone, that rich kid school is now for the rich kids only, and it has gone to the dogs in terms of academic performance.
I grew up in a communal flat in post-collapse Ukraine with 5 families to 1 toilet, and then on food stamps when we immigrated to the US. I went to a public high school, and state colleges.
Your point about public investment in resources which were available to anyone who was willing to make a bit of extra effort notwithstanding, many kids from the very same school(s) didn't do so well, and it was far less about "willing to make a bit of extra effort".
>Public policy has changed, that magnet program is now gone, that rich kid school is now for the rich kids only, and it has gone to the dogs in terms of academic performance.
That, sadly, applies to my high school in Brooklyn too (E.R. Murrow High School). It's not at all what it used to be.
> It seems these days that public resources go mostly to those with the most money, or maybe those who were born into the politically correct group du jour, but almost never to the random kid who just wants to take a shot at doing something bigger.
BIGOTRY ALARM BELL
Yeah right. The kids get the resources today because they belonged to a "politically correct group du jour", unlike you, who was merely "willing to make a bit of extra effort".
And also, you know, had a personal computer at home in the 1990s, and lived in a neighborhood with a high school mostly for rich kids.
Uh huh.
>I'm not relating this in order to validate precisely what my Privilege Index was
That much is clear, which is why I'm pointing your attention to it.
Describing yourself as "the random kid who just wants to take a shot at doing something bigger", as opposed to "those who were born into the politically correct group du jour" was absolutely uncalled for.
Instead of "lived in a neighborhood with a high school mostly for rich kids", the line should say "had parents that not only let you tinker with their very expensive machine, but also encouraged you to apply to well-funded magnet schools".
My point - that you owe your success to growing up in that household to a larger extent than other "quirks of chance" - still stands.
Ok, just waiting on some Indian lad whose parents didn't work for ussr state software engineering to chime in, and put this guy in his place, and i can write my post-post-python class decontruction.
>Ok, just waiting on some Indian lad whose parents didn't work for ussr state software engineering to chime in, and put this guy in his place, and i can write my post-post-python class decontruction.
Not sure what you're talking about, and who needs to be put in their place, and why you'd want an Indian to chime in, but OK.
My point was that when it comes to careet, the parent commentor owes a lot more to their parents than strokes of luck and how tinkerable computers were back in the day.
They did attribute their success to luck though, instead of going for the usual self-made-man myth, so I don't know what place they need to be put in either.
I ended up at the same conclusion as you guys but implemented in yet another way: my home office has a desk for digital work, and a 'study nook' for analog (mostly reading, but it has its own little desk as well).
If I had a larger home, a dedicated den or study room would be a pretty high priority for me. I want my office to be a minimalist, businesslike space with no distractions that focuses me on the screen where my work happens. But the study is a place for clearing my head, thinking and reading, more about comfort, dimmer lighting, familiar objects, and no screens unless I bring in a small e-reader.
Why would they die in cold climate? I would expect them to die in hot climate (no AC - heat stroke, no refrigerator - food poisoning), not the cold where they would have wood/gas heating.
In lots of places that get very cold, even if the heat source is combustion, electricity is still used for ignition, control and distribution through the house.
While I agree, I think there's a better way to frame this with the public. We don't need to bring in pedo references. That looks very unhinged to most people.
There's already a lot of support out there, in both public opinion and the law, for the idea that if I pay for something physical like a device, I own it. Any substantial alteration in its functionality, especially a reduction in what it can do, requires my consent. Reduction in what it can do should require my consent. Just because tech made it possible for the manufacturer to brick my phone or my car, start charging me extra for certain features I already paid for, or block the apps the OS vendor doesn't approve of doesn't mean they should or that it's even legal to do so. Additionally once I buy the device the vendor has zero business telling me how I can modify it, or whether I can repair it.
I own the thing I bought, fucker. It's my property and I have property rights. The corp has no right to steal away part of the thing I bought or change the terms after the fact. It's potentially criminal if they try.
This framing resonates with a lot of people.
The guy who really exemplifies this positioning at the moment is Louis Rossman and by focusing on these widely understood and popular concepts, he's gained the ability to direct an enormous amount of attention to an issue. He can absolutely swamp a legislature with letters from angry constituents for example when he gives an issue visibility.
Frame it as theft because it is. If they push an update without my consent that removes functionality or sabotages my ownership of the device, it's theft. At the very least product liability laws should apply. Some part of what I bought stops working, that goes to product liability. But I'd take it a step farther and say we're dealing with straight up theft.
Speaking of ways to - rightly or wrongly - veer off on a tangent, and convince large numbers of people that the anti-Big Brother side is unhinged...
A better counter argument to "catch the pedo" is to bring up cases of creeps who were insiders - law officers, or just techies with access - and used the "well-intended" tech to get at their victims.
> A better counter argument to "catch the pedo" is to bring up cases of creeps who were insiders - law officers
Certainly. You mean like that time an Israeli Cyber Directorate division chief fled Nevada for Israel after being investigated for soliciting a minor for sexual purposes?
Populations are large. You can draw a pattern in any population.
I think if I were saying this to convince people that hardware freedom is a good idea, they might think I care less about hardware freedom and more about memorising evil Israeli people to make sure I always have a negative example of an Israeli to mention in conversation.
The problem with the reasonable framing you suggest is that it gets thrown out of the window the moment someone utters Protect the Children®. I'm willing to bet that most people, including those with kids like myself, don't truly believe that surrendering our basic rights to better protect the children is a rational thing to do, but they would never dare to push their opinion publicly. The few that do get all but labeled as, you guessed it, fraudster terrorist money launderer drug dealer pedophiles.
It's the the Emperor's New Clothes in real life but for morals. No amount of Rossmanning is going to help society walk back its collective hypocrisy.
I don't actually believe this. People don't actually believe every car should have a GPS tracker so that if a pedophile drives a car, the police can track it. That is a ridiculous argument, and if they make it, there should be something you can say to make it blow up in their face. Unfortunately, as we've all now discovered, winning arguments isn't about being right, so I don't know which words you can say to make the obviously stupid argument sound obviously stupid.
> People don't actually believe every car should have a GPS tracker so that if a pedophile drives a car, the police can track it.
It's not about what people believe, but what they are willing to publicly push back against. If such a law was proposed today, I bet it would pass because the only discussions around it would be whether the data can be kept safe and what punishments to dole out if the car owner access this data. Arguments about privacy will be waved away or dismissed without debate.
In fact, let's make a pointless bet: I bet my imaginary internet reputation that the US or EU will pass a law within the next 10 years that requires the continuous recording and collection of data that not only includes GPS, but also face and audio data whenever a car is in motion. This law will impose severe punishments on any owner that accesses this data or deletes it.
I desperately fear for my family and want things to improve, but we are going to lose this battle.
Since July 2022, Driver Drowsiness and Attention Warning (DDAW) systems have been required in all new vehicle types within the European Union (EU). They will be mandatory for all newly registered vehicles from July 2024.
I think most people would think, and say, that giving every car a GPS tracker so that if a pedophile drives a car, the police can track it, is a terrible idea.
People already showed that they will swallow anything as long as it's attached to "protect from the terrorists" label. Protect the children is an even more powerful extension. Few people ever really have to worry about terrorists but kids, that's a different story.
My logical assumption is that all terrorists and pedophiles will concentrate in the areas where they have legal exceptions from being monitored by multiple different parties at any given time. Legislators and the like. To play one of their cards, why would people who love to say "innocent people have nothing to hide" have something to hide?
I have decided that if they'll play dirty then I will. If someone says "protect the children" then I smear those saying it, e.g.
Kier Starmer wants to protect children? He put Mandelson into government even though he was mates with Epstein. Doesn't sound like someone who cares about protecting children to me.
Rinse and repeat for any politician or political side, they are all only a step or two away from someone who's done something horrible to children. It doesn't matter to me whether I really think it's true or not (though in the example I've used, that is my opinion, who employs someone like that and really cares about children?) but *it does not matter*. This is an us versus them situation, and they are making proponents of freedom out to be criminals at best, paedos at worst. They can take some of their own medicine, and anyone who parrots their line. If ad hominem is the name of the game then let's play, I'm on firmer ground than they are.
> That said, you might be surprised to learn that some of the models from 3b-9b could probably replace 80% of the things nonvibe coders use chatgpt for.
Really? I'm a total amateur when it comes to doing anything with local models but I tried a few in this range using ollama at this point, and they didn't seem to know much about anything, and I couldn't figure out how to get them to search the web or run other tools, so that was where the experiment ended.
A small local model that can use bash would be a bit of a game-changer for me.
Local models are improving quickly so if you keep an eye open you’ll find something soon enough. But from experience, I’ll warn you that local models can lose the plot very quickly. Their little self arguments when they get stuck usually come down to:
- It failed? This must be a mistake, I’ll try it again. It failed? This must be a mistake, I’ll try it again because then I will complete the task (repeat about every six seconds until you rescue it).
- You know, the best way to deal with a permissions problem is to erase the entire system. That’ll definitely solve those pesky permissions and I’ll complete the task.
The latest small models are now reliable enough at simple tools like web search I think. It's just afaik none of the user friendly harnesses like ollama or LMStudio have a real one-click setup flow for this. You'll need to download models and do a fair bit of tool configuration.
I'm no fan of car culture but I think to say it's the primary cause of living a sedentary life at home is an overstatement. I deliberately moved somewhere where I could walk to everything I needed including a fantastic central train station, I no longer even own a car, and yet... over the years my habits changed and I now spend a ton of time at home. My motivation to go out has simply declined.
There is a relevant concept in psychology called activation energy, James Clear provides a good introduction to it. Certainly in recent years screens seem to be incentivizing more stay at home behavior. People used to not own a TV, many quite intentionally, before our other screens were invented. But it is a very complicated topic.
Of course our activity levels change and in some cases go down as we age, but I'd like to submit that is a given, and that car-only infrastructure is an additional barrier on top of those natural tendencies.
It's simply much easier to walk to a coffee shop, or park, or wherever for those who have maintained their mobility (probably in part by living in a walking-centric environment) than it is to hop in a car, sit in traffic, for small things. It's less of a barrier.
There's a pub in my city where the staff greet you when you walk in and actively encourage you to sit at a table that's already occupied. If you come in alone you'll probably end up sitting at a two person table where inevitably there will be a stranger right across from you, with a couple similar tables right next to it. Of course if you want to be alone you can go sit in a corner somewhere.
Needless to say everyone starts talking to each other after a drink or two. This bar is enormously popular. I've never seen it not be packed. It's an incredibly successful strategy for them. With all the complaints about the death of third spaces, I'm baffled that more places don't do this. I see no reason a cafe couldn't do it as well.
All this to say I think it's a great loss that younger people aren't going to bars as much. I wouldn't say they're the best way to form deep connections, but I have zero fear of ever lacking random social interactions, because I know I can just go to a reasonably busy pub in the evening, sit at the bar, and sooner or later either I'll start a conversation or someone else will. It's also a great way to get good at handling opinions that are different from yours - if you have a thin skin or live in a bubble, being subjected to drunk people from every walk of life
will rectify those issues quickly lol.
I'm of half Mediterranian heritage and there was an Italian restaurant that I started eating at out of spite at the French one next door where I had a horrible experience. A little commercial strip across the street from a cemetary. Great food, big hearts.
Nobody ever suggested sharing a table, but if you offered and there was a seat available they'd seat you / someone there. They got busier, and the line started going out the door; people were doing this in line, because if you said you had a "full table" they would seat you at a family table and you could often get ahead of at least part of the line.
The French place closed and the Italian place moved downtown. The end.
I didn't end up with any enduring friends, but I met some great people who I shared food with, learned some interesting things. Riding Amtrak in first class (on the Starlight) was similar.
The peak of TUIs is now. Take a look at Omarchy, an entire operating system built around terminals and config files, it's nirvana. I can only imagine how much farther down this road things may go as we enter a world where the primary interface is conversation with the machine in text. I'm sure I'll get downvoted for that last part because Reddit -- (cough) I mean Hacker News - hates AI, but I'm genuinely excited for the future.
We had "opinionated" TUIs with emacs, and Omarchy will never surpass emacs' ease, shallow learning curve, and configurability. Emacs is the operating system of the future, and you can already integrate AI with it. It provides everything you need or want or don't know you want except a decent text editor.
I used emacs for years, switched to neovim last month. It's just too old and crufty for modern use, and modernizing frameworks like Doom / Spacemacs only add to the complexity. Nice modal workflows just go against emacs' nature.
Plain text coupled with non-deterministic interfaces (AI) is not great. It’s like a hybrid: some of the best of old school tech coupled with the most sketchy high tech.
I will now get to have Kafkaesque conversations with computers in MarkDown.
I'm imagining a TTY-like interface that you can simultaneously type into, speak and gesture at, and whatever else I'm not thinking of (maybe with the "shell" creating a list of suggestions/ anticipating future tasks in the background based on voice input?) Doubt it would be at all practical, if only because the keyboard as a primary input device might not be as much of a thing when you can generate most code/text, but kinda fun to think about.
To be fair, ShellGPT does a lot of heavy lifting for me. Weird thing I need to do, ask it a question in my terminal, it suggests a bash one-liner.
I haven't looked into voice/dictation yet. I'm dumber when I talk and smarter when I type so while it's super cool I imagine the net productivity gain may be a little overstated. Probably pretty straightforward with OAI's API.
You easily have 4k pixels, why use a tiny subset of those in a very inefficient way? We have proper hardware to make a bunch of these computations actually fast, and yet we should stuck with drawing relatively expensive text everywhere?
If you only care about the UX of TUIs, that I can stand behind (though mostly as a guideline, it doesn't fit every workflow), but you can do that with a proper GUI just as well.
> If you only care about the UX of TUIs, that I can stand behind
This is a confusing concession. Of course we love TUIs because of the UX, what other reason is there?
Constraint breeds consistency and consistency breeds coherence.
Take 1,000 random TUI designers and 1,000 random GUI designers and plot the variations between them (use any method you like)—the TUI designers will be more tightly clustered together because the TUI interface constrains what's reasonable.
Yes of course you CAN recreate TUI-like UX in a GUI, that's not the issue. People don't. In a TUI they must. I like that UX and like that if I seek out a TUI for whatever thing I want to do, I'm highly likely to find a UX that I enjoy. Whereas with GUIs it's a crapshoot. That's it.
> the TUI designers will be more tightly clustered together because the TUI interface constrains what's reasonable.
It constrains what’s possible, not what’s reasonable. For example, one could typically fit more text on a screen by compressing it, but most of the time, that’s not the reasonable thing to do.
I’m saying most of the time because of the existence of English Braille (https://en.wikipedia.org/wiki/English_Braille#System) which uses a compression scheme to compress frequently used words and character sequences such as ‘and’ and ‘ing’ shows that, if there is enough pressure to keep texts short, humans are willing to learn fairly idiosyncratic text compression schemes.
One could also argue Unix, which uses a widely inconsistent ad-hoc compression scheme, writing “move” as “mv”, “copy” as “cp” or “cpy” (as in “strcpy”), etc. also shows that, but I think that would be a weaker argument.
Try a 300 baud modem for a few months and good money says something terribly modern like Get-MrParameterCount would get compressed, a lot. Here's Bill Joy on the topic:
> No. It took a long time. It was really hard to do because you've got to remember that I was trying to make it usable over a 300 baud modem. That's also the reason you have all these funny commands. It just barely worked to use a screen editor over a modem. It was just barely fast enough. A 1200 baud modem was an upgrade. 1200 baud now is pretty slow. — "Bill Joy's greatest gift to man – the vi editor". The Register. 2003.
> It constrains what’s possible, not what’s reasonable.
Why do you say "constrains what’s possible, not what’s reasonable", as though it's one and not the other? Does possibility conflict with reasonability? I would think it's not an either/or, it's a both/and.
The set of reasonable things is bounded by the set of possible things. So if the constraints of TUI design make certain things impossible, surely they make those same things unreasonable at the same time.
I'm sorry, excellent GUI with Blender? With the 2.5 interface things were ass backwards but you had a bunch of stuff you could do with only the mouse. With the 2.8 interface suddenly a bunch of stuff was hidden behind arcane key combinations, options disabled by default, and the loss of important visual data like the bounding box view and having both the UV and cursor coordinates in the same tab in the UV/image editor. No matter what the controls are different with every sub-window type, and interface panels flip from top to bottom and left to right for best readability without thought spared for consistency. There's a reason why someone can learn FL Studio in a few weeks, but take months or even over a year to become competent in Blender. I love it's jank and have been using it for eleven years, but I would never call the UI more than serviceable.
You could double or quadruple the number of pixels, and it wouldn't make any difference in how much information humans comprehend easily. You would be using more computing power and more memory to deliver the same amount of useful information less efficiently.
A "proper GUI" is rarely better than a well-designed TUI for communicating textual information, IMO. And the TUI constraints keep the failure-states for badly-designed UI tightly bound, unlike GUI constraints.
What about a map, or an image? We can surely agree that humans can take in a lot more information than a readable letter-grid allows, depending on the type of information.
Sure, of course sometimes an image conveys things better than a thousand words. But a very large percentage of what most people do with computers is primarily text, with more images in ads than useful content. By and large GUIs don't use images to convey information better, they just make text worse.
Modern terminal software supports displaying images, for what it's worth.
> Modern terminal software supports displaying images, for what it's worth.
In a worse, and dramatically overcomplicated way. Like it's kind of funny that largely the same people that is all for this supposed ultra minimalism would be celebrating a Rube Goldberg way of doing graphical interfaces? (Because in the end it is a graphical interface).
Sorry for the late response; it's been a busy week.
As a user, I don't care how complicated is the means of displaying images in a terminal session. I would only want to do so when I'm deep in a text-oriented context and there is a suddenly a need for an image. Not a chart or a graph, but an actual image. As a user, whatever contortions are necessary at that point are fine, because it's an unusual circumstance.
One of TUI advantages over GUIs (including modern web sites) - all text can be selected/copied (you may need to use modifies in some TUI). It's a bit frustrating when GUI shows text but I cannot select and copy it.
It's not only about buttons. A web-app of trading platform I use doesn't allow to copy-paste a fund name (both in web and in the mobile app). I don't think they disallow this intentionally, likely an artefact of GUI framework they use.
When you are "drawing text everywhere", you end up not having to draw all that much text. 3d models have more and more polygons as graphics cards improve, but the 80x24 standard persists for terminals (and UX is better for it). And I'm not even that convinced of "relatively expensive". Grokking UTF-8 and finding grapheme cluster boundaries has a lot of business logic, but it isn't really that hard. And unless you're dealing with Indic or Arabic scripts that defy a reasonable monospace presentation, you can just cache the composed glyphs.
I'm curious: Do you have a nice set of GUI applications that come with the UX you'd expect of TUIs?
(I'm not actually sure what the UX of TUIs is I love so much. Relative simplicity / focus on core features? Uff, notepad wins this one on vim. Fast startup times? I use gomuks, that takes a minute for the initial sync. No mouse? Moving around in TUI text editors with hjkl is slow. I either jump where I want to go with search or use the mouse. Lightness over SSH/network is the only thing I can't come up with a counterexample for.)
Blender? There you have to use a mouse because you have a much much bigger state space to control.
Also, Intellij is perhaps a better example. You can fully control it via only the keyboard, yet no amount of plugins would turn (neo)vim into something as capable as it is. And it makes good use of the extra pixels - human can take in much more information than a text grid.
Not just TUIs the whole stack is converging back to text. I run ~15 personal tools and every one that survived past the first month stores data as JSON/markdown in git repos.
Text in git gives you versioning, sync, grep, and you can hand the whole thing to an LLM with zero serialization. It's perfect for me.
I hope it’s inevitable. Most users’ computing workloads by far are text oriented. The terminal is capable of flexbox now. Current GUIs create massive complexity and power draw relative to their value. Over a long enough arc, economic inefficiency is doomed.
That’s the tip of a conceivable iceberg but exactly. Also look at kitty graphics protocol.
Look at the amount of engineering resources we pour into OS GUI toolkits and then browsers. Those layers of complexity aren’t there because we stood back and said, “given what we know in 2026 how should we design a GUI compositor?”. The majority of the stack is written how it is by archeological happenstance. One generation adds on top of the prior since the 60s.
I’d say start from the terminal, fix the rendering limitations that drove the split from terminal and then to the browser. If we pin down efficient GUI, we could have machines that cover non graphics workloads which is the vast majority with solar and the equivalent of a 6502.
The amount of energy wasted on modern stacks relative to the tasks being delivered is incalculable.
I 100% agree, and this isna big reason why I find the current state of education so suboptimal. Everyone just goes on to do webdev, completely ignoring the lower levels and taking it all for granted. The thing is, there's no real innovation to be done that high up the stack. When you're that high you mostly just write glue code to stick parts someone else wrote together. Real innovation comes from quite a few levels down the stack, starting at the native code level downwards.
Like you pointed out, the current stack is heavily unoptimized and has a terrible architecture; it's only the way it is because of happenstance and tides of the market (companies always reaching for faster over better). An actual "nirvana" in computing like the other guy said would require bulldozing a good chunk of our current stack, keeping only kernels and core utilities, if even.
I really wish we had a bigger focus on getting good foundation instead of making yet another JS framework and SaaS, but then again, who's paying developers to actually do something of quality nowadays?
Comments like this is proof that the old-school hacker spirit is alive and kicking. This kind of pride in efficient and artful use of computers is needed more than ever.
I was confused about why your comment was being downvoted; it sounded like an honest opinion... Until I got to the last sentence. You wrote a self-fulfilling prophecy.
Their solution will be to push mandatory and nonconsensual updates to your devices which limit your device and your freedom in the name of security. Like Google is doing to Android in September. You will no longer be able to install "unverified" software on anything. To address prompt injection attacks they're probably working on an approach where your data all has to be in the cloud and subject to security scans. That's already basically the model for Google Workspace, Google Drive and Chromebooks.
The model will get full access to your data, but in the name of security, you will only be permitted to have data that is cloud-hosted; local storage will effectively just be cache.
The era of the general computer will end, and the products you purchased from these companies will be nonconsensually altered and limited.
I'm so glad I switched to Linux more than a decade ago. At least on the PC there will still be an open source ecosystem for a long time to come, it may have less features but I'm willing to accept that.
Knowing that they can change what you bought overnight with a single nonconsensual update, think very, very carefully about who you purchase all of your future technology from. Google's upcoming nonconsensual degradation of Android should be a lesson for everybody.
>Google's upcoming nonconsensual degradation of Android should be a lesson for everybody.
Google is almost certainly doing this because the iOS was not found to be a monopoly, while Andorid was. It came up in Google's appeal of the Epic case verdict, where they directly asked the judge about it. Turns out you can't be anti-competitive if you don't have [allow] any competitors.
Nope. I'm still going to blame Google for their own actions. Nice try, though. I'm old enough to remember when Google pretended to take responsibility for not being evil. Even had it as their motto.
> I'm so glad I switched to Linux more than a decade ago. At least on the PC there will still be an open source ecosystem for a long time to come, it may have less features but I'm willing to accept that.
Wait until age verification is mandatory everywhere. :)
I can already see that happening, e. g. to access financial transactions or government apps, one needs to verify the id, and that will not work without age verification that can not be tampered with. So Linux will either submit to the same or be excluded.
(That free developers will be able to run Linux fine for much longer will also be true, but I guess they only care about catching the 95%, not the 5% linux users ... and 5% is a high guesstimate).
Edit: To clarify the above, one already had to provide personal data for financial transactions, of course, so a bank knows who is who, but the recent age verification go hand in hand with the attempt to get rid of vpn, and applications now make it a new standard to query the age of users, with the claim to "help protect kids". And some people buy into that rationale too. I don't, but I have seen many non-tech savvy people submit to that justification.
There's always the zero knowledge proof tech alternative, but I don't have the feeling we are moving in that direction - it's not the most profitable business is it.
It's a given that the SOTA models need to raise their prices. It's also a given that they can't. The more they raise the more customers will move to their competition.
So what happens next? Well I think it will suck horribly if you can't move off of SOTA sooner or later, because the Big Two are going to lose customers, and therefore have to raise prices on the locked in customers even more than these projections suggest.
Beyond that if you're looking to start a business, figure out how to use cheap models in new scenarios. Build software which does that and license it. This is kind of contrary to the idea that you shouldn't over optimize for deficiencies in the models that will likely go away in the next generation - for instance a lot of problems were solved when context windows got way bigger. So it's a thin line to walk but I think it's there because a lot of orgs are using Claude today for pretty basic tasks.
The dev who's addicted to SOTA models honestly is going to have to settle for less or get totally screwed. Most applications within business from what I see aside from complex research do not require SOTA. They summarize, they classify, they transform, and doing that accurately has been cheap for a while.
reply