Ask HN: Will programmers write more efficient code during the memory shortage?
Maybe there will even be more interest in the invention and use of more advanced algorithms and data structures that use less memory?
In practice I expect most optimizations will come from "stop doing stupid stuff" and not "use fancy advanced algorithms." But that's a cynical perspective so don't be cynical like me.
Its how software is built now in these palces.
From what I’ve seen of GitHub and AWS this year the answer is no. That’s despite me being bullish on LLMs and finding them highly productive.
in aws, some of the core bedrock services have been replaced with the new serving architecture. that thing was written basically with LLMs.
mind you, guy's a distinguished engineer, his team was basically all principals, but you can do it and some of the new teams are copying the style (though with less success, due to lack of technical skill).
An existing project has a well defined code base, test cases etc
If llms aren't able to migrate a good project then they wouldn't be of any use for general purpose programming
We do stupid stuff as a stopgap to meet a deadline and then stupid stuff stays until it starts being a problem.
The root of the problem is much more deeply ingrained in our economic system.
I have known a lot of extremely talented developers, some with more technical skill than me, that simply failed at their job because they couldn’t come to terms with the fact that their job isn’t to produce the most perfect code possible for the problem.
That should be one of those Tech culture “laws.”
I suspect that the dependapocalyse is a significant factor. When every part of an operation has multiple context rebuilds, and resources are not shared across module boundaries, you get inefficient behavior.
But I’m skeptical that there’s a will to rethink that.
We already have something:
"Nothing is so permanent as a temporary government program."
Milton Friedman, Tyranny of the Status Quo
I feel that there is a bit of a difference, though, because that’s really a cynical (and probably accurate) observation on the behavior of bureaucracy, and I feel the temporary fix thing is of a “finer granularity,” and applies to basic human nature.
Shipping is very important, sometimes more important than what you ship :)
Sometimes when some janky p.o.s. solution works well enough, it truly is good enough.
My anecdotal data point: my MacBook neo with 8 gb of ram running mac os is much snappier than my thinkpad x13g1 (ryzen 7 pro, 8c/16t+ 32gb memory) running linux.
Same user (me), same apps, same websites, same data.
I don’t really blame the apps.
> Optimizing memory server-side directly translates to cost savings for companies.
- everyone assumes their program / website is the only thing running at the machine at a given time, and dev machines are always more powerful than user machines
- it's not really lack of advanced data structures and algorithms that result in the bloat most of the time but the fact that programs and websites are delivered by large teams, there are dozens of submodules that are often loaded even when not needed, and doing it properly is hard to architect to without getting into big complexity and gnarly bugs waiting to happen when someone from other team modifies something and does not know full picture. So it's cheaper to just keep things the way they are to reduce complexity of architecture and fragility.
The main place RAM usage is going to get optimized is on the server side, because the client's RAM is a tragedy of the commons. If you're reducing RAM usage while your competitor adds features then the extra RAM saved by your app will just be silently allocated to theirs, the device won't feel any different and the user will prefer your competitor. There is little incentive to optimize RAM usage on the client side because it only helps other companies, so nobody does it - except (ironically) browser devs, who tend to assume they have first dibs on all the RAM of the device.
If you really wanted people to care about it OS devs would need to surface memory usage of apps visibly in a way ordinary users can understand and translate to customer feedback forms, which is difficult.
Agree, but it's going to be a long, hard, potentially ultimately unsuccessful uphill slog. I think the main obstacle is basically 2 decades of inertia. Approximately everyone believes that server side software needs to scale horizontally. So we optimize systems for this property--I can make it work better by just throwing more computers at it. This was a lot more important ca. 2004 (MapReduce) than it is ca. 2026. For resiliency and redundancy, how many nodes does your service actually need? If you're running more than that number, do you have some kind of justification as to why? This is the mindset we need to try to move towards, but it's very different from how we currently do it.
EDIT: Another good question to ask, which we don't do enough: "does it really need to be a separate service?" Using data in RAM that's already been allocated is probably better than allocating more and shipping data over the network...
1. Some event triggers the function (HTTP request hits load balancer or whatever)
2. A VM is loaded from NVMe (possibly on another computer, and transferred over the network) into RAM
3. That VM boots up
4. The function code runs
5. The VM is terminated
Naively, this probably takes a lot more RAM (not to mention all the other computing resources) than to just have another HTTP handler sitting idle on some computer somewhere.
So this is only a reasonable choice if doing this intensive computation for a short amount of time, and using a large amount of resources for that short amount of time, is somehow better than doing a much smaller amount (maybe none at all) of computing over a longer amount of time. It's kind of hard to tell what the actual conditions are that merit one solution over the other. The other layer is the cloud pricing model, but that's a whole other can of worms..
I have yet to encounter a system that didn't have at least some computers (or cloud instances) that always need to be running. Those computers are almost always a bit overprovisioned (not using 100% of their CPU or RAM or IO all the time). So it has always seemed to me that instead of running a cloud function, it would be cheaper in terms of compute resources to execute the code from step 4 above on one of those machines.
Obviously we have cloud functions for a reason, people like them and find them useful. But is there ever a regime where they're optimal?
EDIT: I guess I've implicitly assumed the code in step 4 is doing a small amount of work. Maybe there's a goldilocks zone of work-per-invocation where this model is more optimal? But then of course there's always the standard executor pool model, and that way you don't have to spin up a VM per task invocation...
CF Workers uses V8 isolates rather than (OS/HW-level) VMs.
Also, the RAM usage isn't really a differentiator between FaaS and any other cloud instance. If a VM is used for FaaS, it's almost certainly also used across the platform for multi-tenancy, invisible live migration, etc.
This is true so long as client-side (desktop, mobile) memory management does not penalise high memory use.
I'm already taking the approach of killing off my largest browser processes regularly, and need to look at more targeted ways of managing memory. I'd really like to see the parent browser process as a lightweight manager over subprocesses such that it can persist (rather than leaking multiple GB of RAM over the course of days), to the point my entire user session falls over with stunning regularity.
I have a shell script (currently triggered manually, hopefully subject to further refinement) which kills off the ten top browser processes. I'll often run that in a shell loop of 10--20 iterations. It barely keeps things manageable, and system hangs/reboots are still far more common than I'd like.
The way to change behaviours is to change costs. This is where OS devs have a choice before them, and application and remote service / SaaS devs and project managers might eventually start feeling the pain.
One reason I favour HN over numerous other options is that the site doesn't absolutely pig out my browser session(s).
And that said, if anyone has tips on both revealing and managing memory usage in Firefox, I'm quite receptive.
Precisely this. We had an incident once where a CSV had a postal code field be interpreted as an integer by pandas, which of course results in stripping any leading 0s. After looking at what the code needed to do, I asked why they were using pandas in the first place, as it was literally just “read the CSV as-is into a list.” Guess what Python’s stdlib csv module doesn't do? Type inference.
Instead of replacing the unnecessary pandas import (which brings along a fairly heavy transitive chain) with stdlib, they added additional code and tests to ensure this wouldn’t happen going forward.
Some devs learn by trying. Others learn by reading docs. Most seem to learn by reading blog posts that use unnecessary 3rd party packages.
If there are no guardrails, loose quality control and nobody cares about performance - why should they replace current pandas with stdlib?
I honestly don’t understand how it’s even helpful to put shit like that on your résumé. If I read that in an interview, my immediate question would be “why did you build this instead of using HAProxy / AWS NLB / etc.?”
Advertisements and tracking.
About 90% of the bloat found on most big company websites comes from these scripts being added all over the place. Ideally removing these would make these sites far more efficient, but the marketing and sales folks probably wouldn't allow it.
"Programmers" don't make this decision, the product owner does.
With LLMs, it's faster to ship even in a more verbose language like Go or Rust.
If you're working on SAP or Salesforce, the language decisions are already made for you. If you're integrating with an existing Electron runtime, then you'll be using something in the JS family, like it or not.
Not solely. The business will have reasons to stay on a mainstream language, for example because
- it offers better guarantees for hiring maintainers in the future
- it has a higher likelihood that security issues will be fixed rapidly for free
- LLMs are better at maintaining code written in it
Even bus-factor comes into it.
I don't even want to recruit or be recruited with such a title.
If the company is using Common Lisp, do you have the 6-12 months to wait for them to ramp up, or do you hire someone who has done Lisp before? That is downstream from the technical decision to use Common Lisp, but it is a huge business impact.
Technical decisions like this have to take into account a lot of factors outside of just the language itself.
Is the language you want to use easy to hire for? Will we have to pay a premium for engineers with a specialised background in the language? Do all our 3rd party dependencies maintain SDKs for the language? Do libraries that meet various certifications we might need (i.e. FIPS) exist for the language?
Something like Typescript or Java is going to win out over Rust/Erlang/FP-of-your-choice on a number of these criteria.
For example, if you're rendering a user account page that has 100 data fields (name, address, etc), that's a few kilobytes at most. If your code is using Node of PHP you're probably using tens or hundreds of megabytes, possibly gigabytes, to turn that into a stream of HTML to send to the user.
I suspect using Claude to turn all the Node and PHP apps in the world to Rust or Go would massively reduce the necessity for huge datacentres using terabytes of memory.
Enh. I think this only exists for some programmers, who can't write good code fast.
The manager wants the Submit button to submit the form, they rarely care how the programmer does it. It's the programmer that chose to install the 11,000 node_modules, to use React with 3 layers of state management on top with hybrid SSR/CSR and to do Kubernetes because that's what was on Hacker News that day or whatever. And guess what, somehow the Submit button does still not submit the form 10% of the time.
Get rid of the junk and the program will be more efficient AND you'll ship quicker.
I suspect you're imagining something like, just write some vanilla JS and maybe some tasteful but not bloated CSS.
You could do that. Then I'll come along with my old man Win32 skills and write a form with a submit button that sends the data by doing a memcpy into a UDP packet. It will take 300 kilobytes of RAM and start in 0.1 seconds. That'll make it approx 100x more RAM efficient than the baseline you can manage with Chrome, where an empty renderer process takes 23 MB of RAM to achieve nothing and starts so slowly the browser has to cache them in the background.
On the other hand, delivering features your users expect might be a bit harder. Shipping and updating that app will be painful for me, unless I use [plug] https://hydraulic.dev/ to make it easy [/plug], and the styling options are limited to setting solid color backgrounds on things. Things the product manager views as basic, like adding a dark mode, will take a surprising amount of effort. The app is likely to be more crash prone than the web app due to all the manual memory management required. Text zoom won't work. I'll have to write in C or maybe C++ if I'm feeling extravagent.
So there's got to be a balance somewhere. Features do matter.
If you ask your boss, it will most likely be whatever spaghetti code which ensures the contract gets signed on time.
The boss doesn't care if the developer needs 10000 libraries for the submit button.
Then there are industries where the customer complains if code is slow. They will actually hire expensive consultants to analyze and benchmark the code. And while the consultants likely are not more talented than inhouse staff, now you have both sides very interested at looking at the problem from engineering perspective.
In this case "good" includes performance.
But what even is good code?
As an infrastructure guy i see my fellow software engineers endlessly debating and bikeshedding on anything but speed and memory consumption.
I see them arguing about functional vs object oriented, immutable data structures, test driven development, agentic bs, this or that interpreted language… never about reducing memory consumption .
I wrote an encrypted mesh networking library that runs on normal operating systems. A customer asked me if I could make it run on an ESP32 with 520 kiB of RAM. At first this seemed impossible, but it turned out that it was, and not even that hard. While the original library was not memory hungry at all for a desktop CPU, it still wasted space on unnecessarily large buffers. Cutting those out made the library run on an ESP32 while leaving plenty of room for an application.
Also, my first PC was a 200 MHz single-core 32-bit AMD k6 with 32 MiB of RAM. This ran a graphical OS with browsers, word processors, 3D games and so on. Nowadays you can get a CPU with more than that amount of RAM as just built-in cache.
So a good place to start optimizing code would be to actually get a "severely resource constrained" computer and start making your code work on it.
If buyers can't afford the hardware anymore, the studios need to adjust. It's definitively possible to scale games down a lot. There are a few AAA games that were "dumbed down" for the Switch 1 (Hogwarts, cyberpunk, ...). And that's a really low-spec device.
There are two factors: existing gamers not able to afford upgrading. But also new gamers, that might only be able to afford much lower spec PCs than people who bought 2 years earlier.
Why games? Because there is a clear point where people stop buying games. Minimum hw specs are known before buying.
Spot on.
Now with LLMs and desktop app libraries such as Tauri, there is little excuse in choosing Electron to build memory hungry apps other than laziness.
Recently I booted up Insurgency: Sandstorm. With a 5800X and an Intel Arc B580 at 1080p and high graphics, the game runs at around 200 FPS. Meanwhile, pretty much any modern UE5 title (with the exception of Ready or Not and Split Fiction, from what I've seen) runs horribly - the interesting thing is that no matter how much you tweak the .ini files or change the graphics settings you can't get something like STALKER 2 or The Forever Winter or Borderlands 4 to run as well as UE4 with the graphics similar to those old games. Instead you get something that runs at like 10% of the render resolution and still doesn't get 60 FPS (I'm not exaggerating, literally the performance I got in The Forever Winter).
There's no good technical reason for things to be that way (Unity still exists, and the games made in it struggle less) other than the devs or the higher ups choosing higher fidelity but more expensive rendering technologies and using upscaling and framegen not as something that helps laptops or when you need the spare GPU capacity (e.g. encoding a video recording of the game), but rather as something that's supposed to be used to even get to 60 FPS in the first place.
I don't know what needs to change for things to get better.
I also don't see anyone particularly caring about regular software, Electron et al are just too convenient to develop in (having to create per-platform UIs sucks in already-overworked teams).
Studios need to start creating custom engines again, for one. We'll get better games with less unsatisfying jank, some of the projects will actually cost less (which is paradoxical to some) and performance is likely to jump significantly. Off-the-shelf engines have as many costs as they have benefits, but like a lot of technology people refuse to look at the choice as a trade-off, and to the extent that they acknowledge it's a trade-off the implicit admission is always that it's a trade-off that the user/player is paying the most for, so it's OK.
If companies start creating custom engines en masse again it will also help solve part of the competency crisis in the industry, because they'll be forced to actually learn and educate people on how things work.
I think certain games like Robocop are awesome on UE5.
Unfortunately, for the consumer this can also mean skipping some games altogether: for example, I might not be able to play the latest Indiana Jones or Doom game because they refuse to let RT be disabled (unless they get enough pushback, but we can all see where things are heading).
At least there are some (usually indie) games that let you do that, like Incursion: Red River but even with Lumen and other features turned down, the performance is still worse than UE4 games of comparable scene fidelity (not necessarily complexity). I think the industry might have jumped into Nanite and all the adjacent tech way too eagerly.
A lot of RT features makes game development faster (or more efficient) so you won't be able to play the game without them.
Of course there is. What people gets presented is look how new graphics is shining. What devs get presented is look how much less manual work you need to get this graphics out of the door. Look at any UE5 presentation aiming at devs. You will be able to see a lot of 'just do this and technology will handle the rest of it'. There is no going back to manually making 3-5 replacements for each and every thing in the game. And the same goes for lights every few meters of a game world.
As a gamedev I don't really care about the regular software. You can see that the main problem for devs is to get paid for it. All kind of schemes with subscriptions and online services and such were tried. People just don't want to pay for the software. The mentality is 'we will get it for pennies on a sale' is the same like with the steam sales. Or even worse people will choose 'free' version with 'promotions' and data collection and whatever else that saves them pennies. Look at 'free to play games' steam category for the example of the horror show.
Oh and I don't think devs are the saints there. You can find a lots of examples that prey on gullible customers or trick people to buy 'digital goods' they don't need or outright bad things with gambling addictions and more.
The only thing consumer can do is to only vote with their wallet and push their representatives to regulate. The stop killing games is an example of the latter. The former is often deemed inefficient but Imho it is the only thing that will separate surviving studios and the shattered ones. As you may see in the press the names and past successes don't save studios from closing their doors now.
hardware costs must come down or every consumer segment is going to be renting, not owning, everything.
Is this not exactly what companies want
In specific sectors I do think we will see more optimization. If you're working on cloud compute or AI training / large scale data processing, there will be a big focus on optimization as prices are very large at that scale and shortages have a bigger effect.
Also in gaming I think the next cycle will be different. Big game studios used to push for the best possible graphics that might require the newest consoles or high end gaming computers, but the next releases might not be as much of an upgrade. The next gen of consoles or graphics cards themselves might be delayed, or be less powerful, or be too expensive and flop, as chip manufacturing companies continue focus on more lucrative markets and leave average consumers behind.
I work with dynamic language developers. Perl, React, Python and so on. Most currently don't care about memory usage. Some devs are actively wasteful. Usually in the name of code clarity or speed of delivery.
Some big-tech orgs (that have their own hardware) will take costs into account, but they already do that. The "optimization" is more likely to be business-optimizations; "this can be slower if it uses less memory", rather than inventing new stuff.
Note that I am excluding any of the big AI labs. They are definitely going to be working to figure out how to use less memory, but that's primarily not related to the direct cost.
Even today it is possible to use languages and programming techniques to be ultra efficient when it comes to memory. It's not that difficult to use them but the world don't care.
Actual example: I have built same utility using Rust, Zig and Haskell (and Go, humorous writeup[0]). Rust binary is 450kb, Haskell binary 30mb. Zig was unfinished but I constrained memory to 512kb just for fun.
Thus it's not about memory per se but more about convenience. Many applications could be scaled by 1-3 degrees of magnitude in memory if they reused techniques but... no one cares.
4K resolution frame is ~40mb uncompressed. Another 10mb is 80k lines of text. Thus fitting in 50mb for 99.8% of applications should be perfectly doable with today's tech.
If the price increase would be per each next 1mb then maaaaaybeeee. But if it's like 100% over GB then I'm sure it won't have any impact.
[0]: https://xlii.space/eng/the-four-language-waltz-a-tale-of-all...
I wanna hang out with coders who care, I wanna read articles about caring and how you can achieve more with less code, and I want to see video content that visualizes the sheer insanity of it all. Like those "scale of the universe" videos, except the complexity isn't fascinating, but embarrassing.
Make it uncool, make software bloat a goober thing you people do at most because they "have to", for pointy-haired bosses, or because they're in a rush or under other constraints. Not something that just gets normalized because hey, thinking is hard, caring is hard, so let's all conspire and pretend we're not being deplorable slobs, under the "leadership" of even worse slobs. There is little technical difficulty here, it's mostly social. Ignorance and greed are in cahoots and don't want to get called out, so call them out.
Though I think to start with, you have to not give a shit about the majority, and huddle with people who care and make better things. If the majority comes around, great, if not, still better than standing on the beach, wondering if the ocean will ever spontaneously take this or that shape.
This very much reminds me of an expression coined by Loris Cro (VP of Community at Zig Foundation): "Create software you can love".
Looks like it's even got a website these days: https://softwareyoucan.love/
Young devs have not been exposed to this and they can get away with writing inefficient code most of the time, sometimes all of the time.
And even if there is a RAM shortage, we are still in the Gigabyte age. So I don't think we will go back to cycle and byte counting.
I can't even get some devs to care if their query runs in 3s or 8ms.
If we can insert "some" then Yes.
> use of more advanced algorithms and data structures that use less memory?
I don't think so.
At least at work there is a push to decrease memory usage but the way I've seen it playing out is not using some O(N) data structure instead of O(N lg(N)) per-say but instead replacing `int[]` with `byte[]` or in-lining some fields to remove some indirection costs.
If your backlog is full of feature requests you will add features.
If your backlog is full of tasks to reduce memory usage, speed up areas etc., then that is what you will do.
Most programmers will have no choice whatsoever.
I think Rust's rise in popularity will probably lead to some benefits.
Games will probably get more efficient but they're easier to scale down to the memory that's available.
Precisely. And all that extra resource wastage is completely free! (paid by your customers).
Perhaps if there were any big software companies who were so iconoclastic as to write fast software and avoid wasteful patterns like using Electron, pressure to do better could be felt, but every company that ships software[1] behaves the same so if anyone tries out competition due to performance beefs, they'll have no relief. They'll be forced to blame their hardware and upgrade it.
[1] Only exception of course is some indie developers. I don't know of any companies that have more than about 2 devs who haven't adopted the 'modern' approach, where we only fix performance issues that completely block using the app.
Android, too. Not much Electron to be found there. Both Android and iOS were heavily optimized for low memory environments back at the start.
The fact that nothing on my modern day iPhone feels any faster than my iPhone 4 did, when the current phones are 10x more powerful tells me that developers consistently spend >= 100% of the yearly Moore's Law dividend on framework bloat, spyware, and useless animations. If we could run iOS 6 and its matching apps on modern hardware, and set animation duration to 0 like jailbreaks used to allow, imagine how fast our phones would be!
Google has already downgraded memory for their upcoming Pixel 11 while at the same time imposing local running models as a first-class feature. Both decreasing the memory pool and increasing the demands on it.
The key is that they own the full stack (hardware and software) and can demand the OS be more efficient, along with perhaps forcing that goal on app developers as well via tightened background limits etc.
The problem is that things go quickly in the other direction between these sprints.
This isn't where memory goes or real-world speed comes from. For most applications, it's abstractions like React running in electron that hurt performance. There are also richer resources backing parts of apps--higher resolutions, better quality, higher framerates.
I still often notice that servers on Linux use <1GB of RAM even with relatively high use. I don't think that's really changed massively in 20 years.
Even then, a lot is required for most businesses to prioritize this (presumably) temporary issue at the cost of things like: participation in the AI race, other features, bug fixes, new markets etc.
Heck, sometimes software is so inefficient that it costs developer and tester productivity but a fix is not prioritized for years.
For cloud apps where the costs are largely hidden from users, the user has no way of doing that analysis. I agree with the second part of what you said, though: I expect businesses to just raise their prices in those cases rather than systematically focus on actual difficult engineering problems.
Apple recently released an 8GB MacBook Neo, which bucks the previous trend of increasing RAM in low-end SKUs, and signals that software needs to be prepared to run in that configuration for a long time to come. I expect we'll see similar moves in smartphones if RAM prices stay elevated.
For most people; computers just suck now.
I work in AAA video games, and I'm often told that consoles hold back games. This is true in the most essential sense: for consoles we are forced by the first-parties to optimise our games so that they function on consoles at a reasonable framerate with not hitches.
Those optimisations help PC players too; and they have more hardware headroom so they can crank things up even higher: without those constraints games would be much heavier and would require you to be on a faster moving hardware cycle.
The same is true when I worked in ecommerce; there was a common verbiage which I don't remember verbatim but the spirit of it was "we lose 10% conversion every 100ms" - so we built our systems to do as much as possible within 100ms (including network RTT).
If people actually have target hardware, and a metric that they're pushed towards, it can happen, people can learn how to navigate the abstractions or do better with the abstractions they have... but I think this wont happen, because in the examples above there was an actual cost to things. Conversion rate is money, not being permitted to launch on console is money.
But you know what modern MBAs see as being money? Time to market & developer salaries. If every company has the same mentality: then it doesn't affect conversation in the same way. Those exogenous constraints simply don't exist in the majority of technology today. It's like a collective action problem.
If there was going to be a collective action against this for financial reasons we would have seen it, shipping laptops with 16G of RAM instead of 4G (or 8G) has a really large headline cost in a company of a few thousand people; but those companies never pushed hard on Teams or Slack.
But I think they should. Not only for the practical reason, but because it leads to better software in my opinion
Limitations foster creativity. Abundance kills innovation.
In my company: - All computers are refurbished and at least 10 years old. - Cloud deployments use Hetzner or self hosted machines. - All hardware in general is a bit dodgy. Our main CI/CD setup is on a very unreliable network with varying availability and bandwidth
It can be annoying once in a while but it reflects our customers reality. And it catches a lot of usability and performance bugs that we wouldn't otherwise catch
I can only recommend it
Niklaus Wirth wrote his A Plea for Lean Software in 1995. Pdf at https://people.inf.ethz.ch/wirth/Articles/LeanSoftware.pdf
People did not bother and then the Web made everything worse. We are now so used to layers and layers of abstraction, humongous libraries and frameworks that most do not know anything about how to write "lean" software.
Nothing is going to change in the application/services layer.
I will say though, optimizing memory usage is easier than ever. Doing a scan of your codebase with an LLM for _large_ memory gains can probably shave a decent chunk off of any application with ease. You don't have to go down the rabbit hole, but taking the top 3 large things it catches would probably result in notable gains for minimal time usage.
Hell, I've hunted down code that was causing an OOM from a coworker that ran a regex that would have to search the entire string of a base64 payload that was often around 24MB, when it very easily could have been a basic string operation on the first 50 characters or so. I caught it myself, but went back after with an LLM about 8 months ago and it also spotted it with a vague point in the direction of where it could be.
There are probably a lot of small slip ups across a codebase that are simple fixes that add up over time, but we never catch because we're not actively profiling.
So, this carefree attitude directly shapes all code that runs client-side (JS + native apps) since the only impact on the company is whatever happens on the dev's laptop. The rest of the impact is "free" since it's paid (in either money or in misery) by customers.
For server side, I also highly doubt it, as being more memory efficient on the server side has always had a benefit to the company who pays the bill. The only things that may change may be the relative cost, but if management comes and says "AWS bill going up, help?" devs will say "OK, we can find ways to save RAM, but the team won't be doing any new features or fixing any customer-facing bugs during that time" and management will say "Okie dokie, we'll just pay the bigger bill then."
And not all exorbitant RAM use is about sloppiness. Sometimes you can trade more RAM use for lower complexity. Bugs and development time were expensive. RAM was not. So sometimes there is calculation rather than sloppiness behind certain types of heavy RAM use.
The whole Electron “ship a browser for each app” ideology will die first.
EDIT: It's also possible that the era of hypergrowth might end. Most of the world's addressable eyeballs are already being monetized. There's no new coal seam to mine. If tech becomes more about doing the same with less, and less about explosively growing at all costs, we'll probably see more efficient systems. And fewer people building them. I think this kind of larger systemic change would be more likely to drive efficiency improvements than the (probably more temporary) RAM crunch.
Most games are written for the 3% who can afford a luxury gaming PC. That's how the market is working for a while now.
However that's more or less unity fault
A web browser and the basic mobile app will be fine.
The iPhone 17 Pro has the most RAM and it's only 12GB. Hell the iPhone 16 Pro only had 8 GB. The vast majority of consumer cases don't need it. I doubt Apple and other manufacturers will go beyond that to keep prices down.
As a consequence of this fixed RAM/core ratio, substantial software development effort goes to either making jobs fit in 2GB or if that is not possible then to utilize multithreading. Generally, particle physics processing does not particularly benefit from MT except in this fixed RAM/core situation. Sometimes large memory jobs are needed (inherently or because of bloat that is too costly to improve). When run on the "grid", these jobs must allocate multiple cores just to get "their" memory. If those jobs can use the extra cores, overall throughput does not have to suffer.
That's for conventional software, which still makes up the bulk of the computing. The situation for the growing amount of GPU-accelerated software is different and more varied. One trend can be seen relating to VRAM. Research groups with easy access to big GPUs like A100 write code to fit or exceed the relatively copious VRAM limits of the data-center GPUs, while groups that lack easy access to DC GPUs but have access to more modest "gamer" GPUs write more advanced software that can fit the smaller VRAM. In some cases, they write the software so it can scale the computation, keeping GPU utilization high while staying just under the VRAM limit.
General budget crisis and limited resources in the particle physics field are in part responsible for all of this tailoring of the software to fit the hardware. If better funded, particle physicists could spend more time doing physics and less time squeezing last drops of processing power.
In particle physics, as more of the code is being GPU-accelerated, there is now another integer ratio to worry about optimizing: CPU core per GPU device. Across the landscape, some jobs have zero GPU acceleration, others may need 100 cores to keep a GPU busy, or only 1 core. Yet others can tune their CPU/GPU ratio to optimize throughput given what hardware ratio a given facility provides. Only a fraction of the software in the ecosystem takes up this challenge.
Most physicist pretending to be software developers or vice versa who are involved in the field do not consider any of these computing realities. At some level, that's natural and excusable. It's hard enough to develop the simulation and reconstruction and analysis algorithms. Simultaneously optimizing their implementation for throughput on a given hardware assumption is even harder. Harder still is to do that optimization over the variety of hardware assumptions. There are only a few cases where this holistic thinking has driven the design of the software.
I feel like we're about to learn a similar lesson with generative AI. Things don't always get easier/better/faster.
It's not a memory shortage like the world had run out of gas for cars.
It's not like we suddenly only have 1MB of RAM to fit stuff into.
The memory shortage means stuff costs more that's all.
I know we are having issues with china and they can't produce the latest tech but I suspect they will produce ram so cheap relative to the competition that companies will find away to buy from them.
When I started out, the mini computers were 16 bit and there was a limit to program size. You could do a lot within those limits, but once in a great while I did exceed the limits and the compile (if lucky) would fail or the program would dump. That was well before GUIs.
I believe the GUI portion of a program are the largest use of memory, so I do not even know if people today can survive without GUIs :)
stepping outside of garbage collection, and managed runtime systems could teach a lot; there was a book written 20 or so years ago;
Until Electron based projects have viable competitors that aren’t using Electron, this nightmare will continue. There is absolutely no justification for consuming 300Mb to 4Gb of RAM when a native app doing the exact same thing typically uses 5% to 0.05% as much memory.
They're suggesting that instead, consumers will just be forced to pay more for inefficient SaaS products being run on more expensive infrastructure.
At my current job we have Windows virtual machines with 4 old Xeon cores, running at 2.4Ghz and 16Gb of RAM, no GPU. Such config will force them to write efficient UI.
“The new MacBook only has 8 gigs of sheep - we need to go resource-lean!!1!”
Feedback loops are the important bit. If you want to reduce your service's memory footprint, don't at the code look at the memory profile and monitoring. You will find something like "oh shit 30% of our RAM is used by these buffers that we could basically eliminate if we tweak the flush frequency".
If you automate/regularise those investigations you will get an efficient service.
Same is true of every other performance metric, and reliability. It comes from your engineering processes (alerts, qualification, prod experimentation) not "write better code".
But most 'programmers' are just Python programmers and not real programmers.
2. Cut out the macho gatekeeping nonsense around what makes a "real" programmer. It's tiresome.
But I’m hopefully optimistic that we’ll see a renewed emphasis on speed and responsiveness, since users do notice that despite most products ignoring it.
Unless that changes we won't see any spike in optimization
Computer: Rewrite this python code in Go. Make it memory efficient.
Or is user choice illusory, nonexistant
Economics will invariably alleviate the memory crunch. It just takes long and requires a huge upfront CapEx.
They have been burned in the past and are hesitant to over invest, worried that the bubble might burst.
I expect high prices to stick around for a while, but I would be surprised if this was permanent.
Which means to me, that price pressure probably won't be the driving force for writing more memory efficient software.
For those who want, I expect AI to make it easier to do that, assuming it's done right, i.e. not vibe coding it.
If you have a subscription to The Economist, I recommend listening to this Money Talks podcast. They talk about the shortage and the economics behind it.
Can anything stop South Korea’s bull run?
https://www.economist.com/podcasts/2026/05/21/can-anything-s...
“ The China card
Two Chinese chip makers—CXMT, the country’s top DRAM producer, and Yangtze Memory Technologies, which focuses on NAND storage—are growing fast and want to expand their global clientele. China is the closest thing to a quick fix for the chip shortage, but the solution is at best partial.
Yangtze Memory is building three new factories in China that would more than double its current capacity by the end of 2027, people familiar with the plans said. Meanwhile CXMT is seeking to raise $4 billion in an initial public offering in Shanghai, and it is building new factories. It said its revenue rose by more than 700% year-over-year in the first quarter of 2026, though it acknowledged that its products still trail those of the three industry leaders.”
Gift link: https://www.wsj.com/tech/why-the-memory-crunch-is-almost-imp...
Who’s going to tell them?
(Just refactored an app to do this. 60k lines to 20k)
Besides, have you heard about "virtual memory" and "swapping"? Nowadays, SSDs (and especially NVMe) are quite fast, so thrashing is much less of an issue.
Those that didn't still won't
Yes. No. Yes.
I've worked in gamedev, helping ship code that ran on consoles. Nice fixed hardware targets. You OOM, you crash. We crashed a lot, and cut and saved and optimized and explicitly invoked the garbage collector twice on level transitions, because a single full scripting language GC doesn't work when finalizers must run to deref C++ objects to unroot script objects, and committed other horrific hacks.
The memory shortage may eventually impact fixed targets like this. Or the minspec publishers will swallow for fuzzier targets like "PC". But it takes awhile for new targets to roll out, and for newly bought PCs to make a significant dent on total percentage of PC ownership. Steam Machine's about to release with 16+8GB and while price and market saturation may be affected by the memory shortage, I'd be suprised if the actual spec changed.
> Maybe there will even be more interest in the invention and use of more advanced algorithms
Not for typical gamedev IMO. There the focus will be more on "reduce the amount of content loaded in memory simultaniously". That means less detail, smaller scale, or less variation. Going from 4096x4096 to 2048x2048 textures quarters your texture memory usage. The surface of meshes also has some square density nonsense going on. Basic animation tends to scale with bone count and keyframes, which are more linear, although shape keys are more per-vertex nonsense.
And of course, reusing the same mesh or texture multiple times doesn't use more memory, just more memory bandwidth.
Audio is more a factor of "how many sound effects (and variants) do we need preloaded just in case there's suddenly an event that triggers them".
These are the big ticket items for memory use and advanced algorithms don't help much. Rather, the algorithms help stretch whatever amount of memory you do have to provide the best amount of quality you can, and the small constant shift in total memory availability doesn't change the calculus for how important that is very much (which depending on the game ranges from "unimportant, everything fits in memory easily" to "critical, we're doing open world streaming and we've got terrabytes of raw data already, 16 vs 64GB of memory doesn't change that much")
> and data structures that use less memory?
Some bit packing type stuff comes up for smaller logical data structures in gamedev, and that can be useful for saving memory bandwidth, but I'm skeptical of how critical it is for total memory usage outside of the occasional Factorio.
You could be a senior dev with over a decades experience at this stage doing nothing but chucking react slop over the fence, performance may have never once come up as a consideration in that time.
> Will programmers write more efficient code during the memory shortage?
The hackers doing the drone software in Ukraine will go to enormous lengths to get something onto Jetson Orin, but once it needs Thor? New cost model.
I think what the parent might be asking is whether the cost of DRAM will be passed along to the most powerless actor in the system, and that depends on whether it's a real competition (war, HFT) or a pillow fight between frenemies (consumer Internet, too big to fail AI lab).
Us liberals have this quaint idea that good referees make capitalism an adversarial contest instead of a rolodex contest, but that idea is out of fashion at the moment.
Abundance and limitations are a bit of ying/yang phenomena in terms of driving things, you don't have one without the other.
Igor Stravinsky: "Constraint drives creativity"
(I also don't see Amdahl's Law --which is fundamentally about limitations -- going away any time soon.)
I do agree that there are compounding abundancies present.
int main(){}