The 'Hidden' Costs of Great Abstractions
jdgr.net
jdgr.net
But ultimately business reality has changed, largely because achieving business goals is dramatically easier with AI tools. This undercuts a lot of the focus on building solid fundamentals, and in a lot of cases that’ll come back to bite the business. But in many scenarios it won’t, and the industry will rumble on.
Those of us working in marketing or journalism or education have already been forced to accept this new reality decades ago, largely because of inventions by software developers. Now devs are just late to their own party.
Programmers used to work with punch cards, then assembly, then low-level languages with odd quirks. Today few developers even think about first-party code size, micro-optimizations, register allocation, etc. LLMs are just another abstraction.
A developer with the ideal AI code writer (which we’re not at yet) must still think about idea, design, scope, etc. like a product owner or manager. And these concepts have theory, sometimes even math (e.g. time complexity).
EDIT to comment on the article: all abstractions are leaky, but sometimes it rarely matters. Today we do still need to understand code quality and architecture when working with LLMs, or the software will get bad enough that it will affect the company. But maybe not next year. An analogy: stack vs heap, memory allocations, etc. still matter in high-performance software, which isn’t uncommon, but programmers almost never think about register allocation.
You cannot be serious.
LLMs are also quite deterministic if you want them to be - generally, their final token selection is deliberately randomized (the model “temperature”). But the word you’re looking for here is probably not actually determinism, it’s probably something closer to predictability.
In any case, it’s perfectly possible to ensure that the output of LLMs is fully deterministic, debuggable, understandable, and testable.
> You cannot be serious.
I don’t think you’re thinking about this clearly.
There's no comparison to be made between this and, for example, a compiler. It's an incompetent comparison.
> I don’t think you’re thinking about this clearly.
My literal job is dealing with layers of abstraction. I'm thinking pretty clearly when I tell you that, not only are LLMs a super leaky, terrible abstraction, they are also not comparable to any other layers of abstraction. All other layers of abstraction we use are well understood, predictable (as you put it), and DEBUGGABLE.
When claude deletes a fix it did two weeks ago, while trying to fix some unrelated error, do you never stop and think "this is not quite the same as what GCC does"?
With a sufficiently complex specification of a failure mode, you can find problems with anything.
Humans, given sufficiently complex requirements and sufficiently complex codebases, also regularly fail. You're tacitly admitting that LLMs are approaching (if not exceeding) human levels of performance now. We somehow get non-deterministic humans to achieve useful work. In fact, staff provide managers with an abstraction over the work they're responsible for - managers don't know every detail of the systems they're responsible for.
There are effective ways to use LLMs. I recommend using those, not using overly complex prompts, and not letting LLMs freely make changes to large code bases. Just as compilers only compile one source file at a time, LLMs work best if you scope their attention. Same goes for humans, in fact.
> There's no comparison to be made between this and, for example, a compiler.
A simple comparison is that both can generate useful code. You need to be more precise about the issues you're trying to identify.
Anyway, the comparison to compilers isn't really the point. It's undeniable that LLMs are an abstraction themselves, and that they can generate new abstractions. Saying that they're "not another abstraction" is just definitionally wrong.
Sure, they're not the same kind of abstraction as a traditional compiler. They require new ways of working, but actually not that new, as the manager example I gave suggests.
> When claude deletes a fix it did two weeks ago, while trying to fix some unrelated error, do you never stop and think "this is not quite the same as what GCC does"?
I never made the mistake of thinking LLMs were the same as GCC in the first place.
And once again, I've seen human developers do exactly what you just described. That's why we review code. All the arguments you're making are essentially also arguments that humans shouldn't be involved in software development either.
In the shallow sense that any PRNG is deterministic if you set the seed and if you control triggering order.
However that's not usually the situation/scope people are talking about.
My broader point is that it's not really the non-determinism that's an issue. What the other commenter seems to be looking for is something along the lines of repeatable correctness, where correctness is generally a requirement that the model doesn't have full access to. The non-determinism is an implementation detail here.
A compiler simply always preserves semantics of its input. Even when randomness is used (ex. possibly during register allocation).
However, there is absolutely nothing stopping an LLM from "deciding" tomorrow that a fix it built a week ago is no longer real, because not only has that fix left its context, but also the bug was not obvious.
Yeah, and we've never had deterministic tools like GCC suddenly fuck up commonly-relied-on undefined behavior between releases. Sure.
I get what you're saying, but again, to the vast majority of devs, none of that shit matters. Whether that's a good thing or a bad thing is a different discussion.
While I agree with your thesis, I don't think your example actually happened. From my experience teaching CS, the fundamental skill when it comes to programming a computer is algorithmic reasoning, that is, the ability to split a task into subtasks until you reach base tools that you exist in your toolbox. Whether those base tools are MOV or document.getElementById() is largely immaterial, IMO. Obviously it's quicker, and easier, for us to not have to drill all the way down to assembly all the time, but if you have a firm grasp of algorithmic reasoning, you are capable of it.
Algorithmic reasoning is a valuable skill across the economy, one of the most valuable, in fact, right after "risk evaluation" and "having a lot of money", but it's difficult to teach. For some people it seems to fit right into their mental model, and it's simply a "duh". For most, though, it either takes a ton of grinding their head against it, or more often they simply bounce off and decide to do something else. I think that is responsible for the relatively limited supply of developers, which combined with software being wildly profitable because of the whole "copies are free to make" scaling created a shortage, which drove up wages.
They drove up wages so much that approximately everyone who has any aptitude for algorithmic reasoning is now funneled into software development. I think that, more than anything, contributed to the explosive growth of the number of developers. Adults may be hesitant to jump industries, but subsequent generations flooded in.
Not so with LLMs. I can give you an english prompt, and several different LLMs the same prompt and you would all understand it differently. There isn't a way to move between abstraction levels.
I agree that current LLMs are a bad abstraction, and any non-superintelligent AI is worse than a mechanical abstraction in important ways (it's much more complicated, which makes it harder to debug, impossible to prove correct, creates unnecessary coupling, etc.)
But it's still technically an abstraction. A project's source (or part of it, e.g. a single function) could be a sequence of prompts, and even a non-deterministic LLM, if it's good enough, will always output correct code.
A prompt isn't an abstraction. Say you evaluate an llm completely deterministically. Fixed seed for the sampling, or maybe zero temperature. You give the prompt to an engineer and ask them what it does. He can't really say for sure. 'Try "compiling it". It doesn't work? Well try adjusting your prompt a bit. How? I have no idea; try renaming some variable to shirley or something see what pops out.'
They didn’t even earn anything close to what they were worth. According to Marx’ Labor Theory of Value anyway.
However the dice fall now, one of the possible outcomes is that the tech billionaires take that 100K USD for themselves. The very deserving individuals whose job is to sit their arses on automation assets.
Meanwhile workers from other sectors can gloat about how they are now in the same boat as them. The boat of accepting your ever-meagre reality.
In Germany for instance I've seen many a company that treated their programmers as a cost center and they actually were (probably a mutually reinforcing self-fulfilling prophecy).
Too many instances of programmers being deployed in such a way that I couldn't possibly see a way that they would get back even that meagre investment that was being made. Fully irrational dev teams doing useless busy work.
Most German "startups" used to be replaceable with Zapier and Pipedrive. That has probably only gotten worse with the advent of LLMs.
I'm OK with this now, it is what it is, but these years weren't smooth as there were ups and downs and a down after an up can be stressful if you're not ready for it.
If I would return today to Portugal, I would probably earn less after taxes than in the dotcom days when working for Altitude Software, by moving into my home town.
Sure it is above minimum wage, yet plenty of office workers get similar salary levels, provided they have an university background.
I am also aware that IT salaries in Greece and Italy aren't that great, versus other office workers with high education background, and we all three enjoy our unpaid overtime.
That doesn’t mean we should accept mediocre. Businesses might not care. Few businesses have bought a product based on how many lines of code it has or how easy the code is to maintain.
Even building software for them for nearly 3 decades it became apparent early on that businesses don’t care. It has always been a point of contention: the struggle to ship now, faster and making sure we ship the right thing and do it well. We had to learn when to give ground and when to pull hard… because in the end there are times when it absolutely matters.
Just because business can’t recognize when it’s about to shoot itself in the foot doesn’t mean we should let them.
This has been the excuse of mediocre developers for decades too. It’s how we ended up with sloppy code in production. Terminals that can’t scroll without flickering or handle much data. Apps that have loading screens on super computers. Software that sometimes works. Ship fast and break stuff.
Would love to see the business and manager types manage software and infrastructure. What's the worst that could happen? Go on, do it. Every time a foot gun goes off it'll be followed by a condescending chuckle.
I used to see 'passion' as the defining factor of how to stay in the field and do well, and that was advice given to people who wanted to join the industry -- who showed the minimum of interest. Now we're going to have these non-technical people who definitely aren't interested and definitely don't have passion for it try to make and manage quality software?
This sort of thing happened to, for example, Maplin.
The big poster child is sadly Twitter. A lot of people said it would collapse without 90% of the staff, and that hasn't materialized. I suspect they can't deploy huge changes to the backend, but they never did that much anyway.
(also, those of us not in the US and not in FAANG always wondered how such a steep salary differential could have been maintained forever; more than doctors and lawyers? Comparable to finance bros or the fabled quants? All of those are much more onerous jobs with much harder entrance criteria!)
If everyone asked for returns, or sued, software companies the same way they deal with other goods, the atittude would most likely have changed by now.
Not to mention the whole EULAs ("we don't have any idea what we are doing here, please sign") disease.
How can they do that when they probably had to sign an arbitration agreement to use the product? Can't pull that shit with a box of oatmeal or most other physical items without software components.
It's not clear if it meant anything anyway. Half the point of an EULA is to scare you away from trying to enforce your rights even if a lawyer would rip it up.
If I'm a business owner, and my point-of-sale software was wildly unreliable, I'd switch if I wasn't constrained by a contract. I can't exactly move off of Facebook as my marketing strategy, though.
And look at Twitter--much less reliable.
This is the state of the homebuilding industry right now. Most of these homes are sold to people who don't make payment contingent upon passing an inspection. The business knows quality doesn't matter to buyers and you can save lots of money by cutting corners.
This is not just homebuilding, this is the future of every industry. We're putting distilled water in the hydroponics. In tech, it was already happening before AI, just look at the usability decline in every Windows and OSX version.
not to say AI-tools do not contribute, they lower the bar of entry to the profession after all, but any C-suite/hiring manager is much more arbitrating labor expense than AI-subscriptions.
What happens in the above scenario is that software gets evolved to a point where LLM cannot make any progress anymore without deep additional direction that a human is unable to provide at that point.
Most effective SW development with LLMs today happens under system architecture/engineering controls by humans — remove that and you converge into a closed-loop situation.
My analysis of this is that "best practices" have evolved to represent a sweet-spot between _running_ complex systems and _evolving_ complex systems, and removing humans out of that equation does not change what works or does not work (I had a sufficiently hard time coaching humans to adopt the best practices as well).
Don't even really have anything else to say other than that, but maybe commenting it somewhere helps someone else realize they're not alone. I don't know how that helps you or me, but that's what I got. Maybe there's still something for us somewhere, but it is very difficult to stay motivated, and I don't have an answer.
I enjoy software development and hopefully one day I will return to it, but I am but one tiny kernel of corn in such a mighty ocean of shit so I might as well right the waves instead of fighting them. Maybe your calling is scamming Indians or scamming Americans or scamming Indian scammers. You aren't alone but the attitude you have will never stop mattering. See if you want to go back to school, start a tutoring program for kids. Motivation is for morons, do something.
I have spent months adjusting my resume, applying for all
jobs where my skill set may be of use, building
proof-of-concepts using Claude, and doing cold outreach to
anyone who may be interested in my potential products or my
services. The well has gone dry.
It is analogous to email spam in the 90's, before anti-spam technology was mature.
Staffing companies have recruiters which vet candidates to varying degrees of success. At minimum, they establish the candidate:
- is a human
- lives where they claim to live
- has worked where they claim to have worked
- has eligibility to work for one or more of their clients
If nothing else, the above eliminates much of the "99% resume fraud" problem companies are dealing with now.
This is what recruiters in quality staffing firms do. Granted, there are many staffing firms which are worthless body-shops. But those are not reputable. :-)
> ... and something the equivalent of a casting director at companies instead of the current "cram leetcode" mode of evaluation.
The equivalent has traditionally been hiring managers who work with approved staffing companies, both to ensure those companies provide value as well as to foster an understanding of the people/skills needed by the organizations.
Wise organizations use multiple staffing firms and perform internal audits in order to minimize complacency/corruption.
At least this is what I recall.
Meta: this is probably the first time this year where I use the word agent to refer to a human. Feels odd even.
If someone who contributed to the Linux kernel has their resume on par with a spammer who lied about it how do we know who is correct unless there is a verification system?
Put simply, organizations needing skilled personnel can delegate post-application screening to their set of approved staffing firms. Said organizations can employ initial screening techniques to filter out obvious fraud, such as requiring online applicants prove they possess the email/cell phone provided via industry standard mechanisms (while employing deny-lists as applicable). Given the remuneration commitment each hire represents year-over-year and the ability to hold staffing firms accountable, their fees are typically quite reasonable.
Why you introduced the question "what impact have LLMs had on productivity?" in this context escapes me.
When the problem definition is "companies want applicants who are known to be humans having a minimally vetted work history", Occam's Razor[0] implies people can do so efficiently. If for no other reason than it being trivial for one person to converse with another.
How would the above result in:
... of course a reduction in industrial productivity and a
negative contribution to economic growth.
Everyone that praises how they get more productive always forgets that means big corp now needs less of us.
I work on enterprise consulting, and have watched how the change into managed cloud infrastructure, followed by low-code/no-code tooling, has had an impact on team sizes, meaning less devs for the same outcome.
AI driven development is reducing those team sizes even further.
In many European countries, gettting jobs at a later age is really an almost impossible task, the easiest solutions end up trying to get early retirement status, or go self employed, which also isn't without its own set of complications.
By now... I see in my country high prices for laptops with only 4Gb of Ram and Celerons.
It could do wonderful things if in 2000s people didn't buy the argument that hardware is so cheap so lets write unefficient code. Same hardware that could play an Youtube video in 2000s today cannot even open the website. Electron send hugs...
Now people are mad about AI until when? Oceans be drought like in Oblivion movie?
And professionals? The generation of specialists will pass... and people will blindly depend on Ai soon if the course of things doesn't stop or at least be corrected.
I think the author could have brighter days in future (and still thing in present in some hidden niches) as knowledge will always precious.
The main lesson I have is buy less TI and every buzz promises and find the place where knoledge and craft walk side by side.
Marcus Aurelius the historical figure was a monster who killed a measurable portion of humans alive at the time
We are not so one-dimensional that good mind habits are the one and only thing we do and act on.
More and more places just want Jira tickets done fast instead of someone that's going to push back or question if this is the best way to build some thing. They want the thing, they don't care if it works well. They don't care if it's efficient. They want it now.
We've been moving to React, replacing an internal framework that's worked wonders for us we've been using for over a decade. The biggest part of the move is "hiring".
My general sense is that nobody understands how React works under the hood. The answer I get when I ask questions is generally just "don't worry about it".
Everything is giant overbuilt and terrible because most people never bothered to learn even a single level up from where they do most of their work. The people that do become unhirable. Everything takes hundreds or thousands more cycles and electricity it should because people can't be bothered to understand what they're doing.
That's one thing I never care to do unless I'm the one making the technical decisions. What I do is to build the thing, but with defensive programming in place. I take care of making that my code is good, then harden any interface so that I can demonstrate that I'm not the cause for new bugs. People will be careless, so make sure that you have blast doors between your work and theirs.
And I do take time to learn about the abstractions of the new shiny tools, even when it's overengineered. Going blind and making mistakes is not my cup of tea.
This is one of the most alienating things about the modern software engineering industry. Someone who grew up just fucking around with computers since they were 5 is supposedly now on even footing with someone who took a 16 week bootcamp and a Claude subscription and has never seen a terminal before.
I was at a drum and bass show recently and talked to one of the other people there. It was obvious I didn't really listen to that much drum and bass as I couldn't name anybody except the most popular artists. You see peoples' reactions change slightly when they discover you are not really part of their music scene - you're an outsider, or a tourist, or even a poser. That's not even a problem, that's just the way subcultures are - you've either lived and breathed that way of life, or not.
What LLMs are doing is they are automating the manufacture of posers and cultural appropriators at scale - you don't really understand the nooks and crannies of this territory, you never actually lived on IRC or in the bash terminal - but you can sure wave around these oversimplified maps of the territory with all the back alleys and laneways missing, and use your pocket book of translated phrases to pose as a native.
> My general sense is that nobody understands how React works under the hood. The answer I get when I ask questions is generally just "don't worry about it".
The problem in software is it seems that we are losing the ability to distinguish between appropriators of computer geek culture and those who do "speak" programming languages natively. The bar has fallen so low that I can't even expect people to understand the difference between runtime and compile time. Anybody who brings up such advanced and esoteric (read: high school level computing) topics is viewed with scorn, as if their ability to expose ignorance on foundational topics presents an existential (or career) threat.
There's been a rise of anti-intellectualism in software from people with non-STEM backgrounds who actually disdain seeking out and possessing such knowledge. It's utterly useless to study - just like math. I find it harder and harder to locate hobbyists, especially here in Toronto, who bother to go below the abstractions not just because they want to, but because they are compelled to understand.
I guess because I’m in game dev maybe, but in all my jobs knowing about the underlying stack has either been necessary knowledge or highly regarded.
I can’t think of any time in my career where knowing about the internals of the stack was ever frowned upon or where it’s been anything other than an advantage (especially when hunting bugs). I must have been lucky.
Instead, what we have today is a computing ecosystem dominated by powerful players who care about money and control. Speaking from the standpoint of a Bay Area resident, since roughly 2012, the field has been increasingly taken over by people who are in it for the money. Combine that with Alan Kay’s observation that computer science is a “pop culture” that often lives in the moment and has little regard for the past, and also combine that with the “move fast and break things” attitude that permeates modern software development, and this has created an environment that seems hostile to the types of nerdy pursuits that the industry once encouraged. The working environments of many major software companies and the products they release are a reflection of the values of the companies’ executives, managers, and shareholders.
While I’m not anti-AI, I see agentic coding as another step in the direction that the software industry was already heading towards, where it can move even faster and break even more things.
There is still wonder, joy, and freedom in computing, but I feel this is increasingly confined to the hobbyist world and certain niches in research environments.
Design time, code time, compile time, run time. Why all that potentially wasteful upfront work?
The next step are shipped applications whose help menu is a chat interface that responds to all user questions of the form "How do I ...", with a short pause to add a new hack to the existing pile, and then some upbeat instructions.
In theory this should be nirvana. No more vibe coding! Everyone is a power user. Zero dependencies. But there will be much weeping.
If I had to sum up the zeitgeist of the '90s techno-optimism it would be this persistent, confident prediction that once people just learned _how_ to use computers, and everyone is a power user everything will be fine! Despite the mounting evidence that actually, no, like everything else in reality the distribution of skill is a bell-curve with the median sitting uncomfortably low for those who, to quote OP, "lived on IRC or in the bash terminal".
Free universal education didn't fix this problem, LLMs won't fix this problem. Man's natural paucity is no longer in the availability or accessibility of knowledge. The liberal ideal that all we must do is empower the individual turns out to not have been the solution to everything forever.
But hey, being self-aware enough to make productive use of this new technology is probably _some_ kind of edge.
May as many as possible survive.
Yep: It's a cheap tiny factory you control that's big enough for your unique priorities! Every person would own the means of production!
I don't think the assumption was that everyone would write software, but that they'd at least they'd choose in a way that favored future independence of choice. It's depressing to see the arc from "tool that grants autonomy" to "tool for delivery of product from megacorp."
It's a method that worked out, for those companies who needed it. There's still people who're like you they're just 'hidden'. This isn't just done in the tech industry, devaluing of workers is done everywhere. Look at the luddites or those working in older industries in America/Europe. At some point government/industry believed they had too much bargain power/there's labour shortage/it made economical sense and they could be replaced in Korea, China .... It's the same thing just in the tech industry it kept growing within the US.
Not saying it’s wrong, but plenty of folks successfully program in high level languages, never knowing a single ISA at all.
When I graduated, I was shocked to learn that no company cared about any of the architectural concepts that I had learned. UML class diagrams, sequence diagrams, ER diagrams, etc... had been on the way out. At one point, as internet companies where scaling up, there was a brief resurgence of interest in sequence diagrams... Especially as a communication method when explaining complex bugs or complex message-passing scenarios. But it didn't really last. Nowadays most software is riddled with race conditions and deep exploitable architectural flaws. Cryptocurrencies have been victims to many such attacks. Billions of dollars have been lost to race conditions... And that's just the ones which were discovered. They are notoriously difficult to find post-implementation.
The programming primitives that we're using today aren't optimized to avoid race conditions or even try to encourage good concurrency patterns; quite the opposite; they encourage convenient but disorganized parallelization and they're optimized to put the focus on type safety which is a far less concerning issue. A lot of people who were rightly alarmed by gaps in schema validation (which is critical at API boundaries) became overly obsessed with type safety (which is a broader concern). I have built some async primitives for Node.js, nobody cared! NOBODY! Other developers have had the same experience with most other languages. I think only a few niche languages like Elixir actually treated it as important. But nobody even acknowledged that the problem could be remedied in existing languages. It's so bad that it seems as though some people wanted it to be that way.
The term 'concurrency safety' doesn't even exist! Some have a vague idea about thread-safety OK, that's very specific to one particular concurrency primitive... but what about the concurrency of asynchronous logic (much more common nowadays)? I have felt thoroughly suppressed in that regard in my career.
The only voice on the subject of architecture that got through to the 'mainstream' was Martin Fowler (one of the inventors of Agile software development). After that, there was Dan Abramov of Redux fame. Some notable opinionated architecture books were published but none really identified the underlying essential philosophy to good architecture.
The best, most succinct quote I ever read on the subject was from Alan Kay (inventor of OOP) who said "I'm sorry that I long ago coined the term 'objects' for this topic because it gets many people to focus on the lesser idea. The big idea is messaging."
I like that quote for many reasons; firstly because it shows wisdom, secondly, it tells you what the issue is, very simply and, thirdly, it hints at the importance of 'focus' in this discipline where we are saturated with thousands of complex overlapping and partially conflicting ideas.
I think the FP trend was somewhat of a red herring. Same with Type Safety. Yes, they were useful to some extent, there are some really good ideas in there, but people got so caught up in them that the most fundamental area of improvement was ignored entirely. To me, the core value proposition of FP can be reduced down to "pass by value is safer than pass by reference." Consider that in the context of Alan Kay's "The big idea is messaging." - Is an object reference a message? NO! A live instance is not a message! Precisely! His point supports pass-by-value, furthermore, it encourages succinct/minimal parameters.
Good architecture is rooted in 2 core concepts. 1. Loose coupling. 2. High cohesion and you achieve those by separating logic + structure from messaging. The biggest mistake people make it passing around structure and logic as parameters to other logic. You should avoid moving around logic and structure at runtime; only messages should move between objects; the simpler the messages, the better. And note that 'avoid' doesn't mean never but it means you have to be extremely careful when you do violate this principle and there should be a really good commercial reason to do so. I.e. You should exhaust other reasonable approaches first.
Can you provide an example for this?
Let's not have dashboard access the temperature by doing `GetSurroundingCar().engine.temperature`
If the dashboard needs to get the temperature from a sensor in the engine, it should be able to "talk" to the sensor, without going through car object.
In ideal OOP, a "method call o.m(...)" is considered a message m being sent to o.
In practice, field access, value and "data objects" etc are useful. OOP purism isn't necessarily helping if taken to the extreme.
The pure OOP idea emphasizes that the structure of a program (how things are composed) should be based on interactions between "units of behavior".
Maybe there are other languages?
Then you should burn the whole thing down and start over!
I work on tractors not, cars, but I'm sure the abstractions are similar. Our engine objects are things like [service]AutomationEngine. Our dashboards objects are for a collection of things on secondary displays - meanwhile the thing you would point with your finger as the actual dashboard doesn't have or need a dashboard object. There is an object for things like RPM gauge, or check engine light - this later is a generic fault indicator with a icon field; it is connected to a different messages; and placed in different positions depending on the fault.
The point of the above isn't how tractors are designed, it is how the objects you need to design a real OO system rarely have anything to do with that looks like objects. Nobody ever writes dog and cat objects derived from animal; nobody ever writes triangle objects derived from shape.
2. Avoid overly complex function parameters and return values. Stick to passing simple primitives; strings, numbers, flat objects with as few fields as necessary (by value, if possible). Otherwise, it increases the coupling of your module with dependent logic and is often a sign of low-cohesion. The relationship between cohesion and coupling tends to be inversely proportional. If you spend a lot of time thinking about cohesion of your modules (I.e. give each module a distinct, well-defined, non-overlapping purpose), the loosely-coupled function interfaces will tend to come to you naturally.
The metaphor I sometimes use to explain this is:
If you want to catch a taxi to go from point A to point B, do you bring a steering wheel and a jerry-can of petrol with you to give to the taxi driver? No, you just give them a message; information about the pick up location and destination. This is an easy to understand example. The original scenario involves improper overlapping responsibilities between you and the taxi service which add friction. Usually it's not so simple, the problem is not so familiar, and you really need to think it through.
We understand intuitively why it's a bad idea in this case because we understand very well the goal of the customer, the power dynamics (convenience of the customer has priority over that of the taxi driver), time constraints (customer may be in a hurry), the compatibility constraints (steering wheel and fuel will not suit all cars). When we don't understand a problem so well, an optimal solution can be difficult to come up with and we usually miss the optimal solution by a long shot.
[1] I don't think I invented "Object-oriented" but more or less "noticed" what was really powerful about just making everything from complete computers communicating with non-command messages. This was all chronicled in the HOPL II chapter I wrote "The Early History of Smalltalk". — Alan Kay
Pass by value would indeed make a big difference to how programs are structured and make it easier to reason about programs.
I just want to point out that "concurrency safety" is very much a word, although "thread safety" is more common. These are broadly part of memory safety, which is a topic mainly due to security concerns but also academic study.
The two perspectives are not perfectly congruent. Non-concurrency-safe languages like go can also be considered broadly memory safe. The pragmatic rationale is that data races in GCed languages are much less exploitable. From a academic, principle based view this is unsatisfying and unconvincing as one would prefer safety to be matter of semantics. See also https://www.ralfj.de/blog/2025/07/24/memory-safety.html
Rust uses "fearless concurrency" as a slogan. Rust offers more options than passing by value (Copy) while still guaranteeing safety through static type checking.
There is also research for GCed languages to establish non-interference eg Scala capture checking.
Concurrency is recognized as difficult (at least by people who are knowledgable) and programs language design usually involves pragmatic choices if you need concurrency. If the language does not provide the primitives or spec that enables safety, then you are left with patterns and architecture.
The science is still evolving, it is certainly not the case that nobody cares. Rather, progress is slow and moving ideas from research industry is even slower. How much value we ascribe to correctness, safety and performance in industry depends very much on the context.
- https://lamport.azurewebsites.net/pubs/time-clocks.pdf
- https://en.wikipedia.org/wiki/Chandy%E2%80%93Lamport_algorit...
- https://www2.eecs.berkeley.edu/Pubs/TechRpts/2006/EECS-2006-...
I think the era of "poor software for fantastic hardware" is coming to an end.
UI development is an even more special case here. The customer buys the machine which runs the code, not the company. So sadly "good enough" is the standard.
One example for me here is the "switch product option" button on Amazon listings (e.g. switch green to blue color, smaller to larger model). On my phone this sometimes takes >5 seconds to properly load. Horribly optimised.
Just as an example, the cost of one week of engineering time corresponds to tens of thousands of vCPU-hours, which is many years of CPU time.
As such, it only ever makes business sense to optimize code either when it has bottlenecks that can’t be fixed by throwing hardware at it, or when it’s so inefficient that it can be sped up by several orders of magnitude.
Especially true when you're paying for neither hardware nor labor.
Writing inefficient client-side software, whether it's desktop or webshit, makes the customers / users pay for the hardware, and pay with their time.
Jokes aside, if you don't need two-way data binding, using react frameworks pulls in a lot of crap that you never need.
The majority of web apps have no need for react
‐--------
[1] I always joke that the reason I am atheist is not because I don't know much about your religion, it's because I know too much about your religion.
By that they mean outsourcing.
Can you share what this internal framework is?
That's because they don't know what to build that will be a successful product, so they essentially try to brute force this question of "what to build" by trying different ideas quickly and see which one will stick. And in this quick iteration loop people just throw bunch of stuff together to make something and once that something gains traction they will keep piling on top of that shaky foundation.
there will be a reset at some point, and software developers will be needed. especially when every piece of software stops working. idk if that will happen before or after an economic collapse tho.
i have no idea where things will go in the future, but i doubt it will be much fun
I don't think most of those jobs will be in the West, though.
I do suspect that developing software in California in 10-20 years may be looked at as if one were proposing a sweatshop to sew pants there right now.
Some “Java in the 90s” understanding of abstraction. Proper abstractions break complexity into composable elements. Hence, fidelity of our understanding increases.
That's why you see hundred level call stacks, polymorphism with a single implementation and still errors are hidden or root causes hidden behind "exception caught".
“Duplication is far cheaper than wrong abstraction." I spoke a million words
They didn't mean that much to me
They rang around my head
Like empty tuneless harmonies
Love's great abstraction mineThe fact that modern tech has disintermediated people with problems to solve from the need for a "priest class" to commune with the machine to solve the problem is a great thing. It's the goal. The more we do it the better we are making the world for humans.
... the fact that people need to work to eat or provide anything above a subsistence quality of life is not only tragic, it's increasingly abhorrent in a world where automation and simplification via machines has freed up this much raw resource and free time.
If we're pitting LLMs against people's ability to provide for their families, we have lost the thread on why we're doing any of this.
However there are tasteful ways of doing it. And google and meta in particular certainly are not.
Those resources are being redirected to create entertainment areas for the rich like golf courses, 7 star luxury hotels and villas. This is the modern predicament.
just share the damn thing, someone may have something for you ;)
...I've kind of rarely seen these ppl complaining about work actually sharing their resume or a condensed description of their skills, knowledge and experience
so, a path could be picked from what you know:
1. devops/sre - really hard to get above entry-level without real experience and you _will_ be competing head on with AI ...ouch
2. cyber - same with whitehat as with devops/sre ...basically go full red-team / blackhat / offesinve for a while, the get certs and portofilio, then job in "real cyber" ...BUT ppl that do this tend to have a "very specially broken brain", so if you haven't done this already you're probably not one of them [probably for the best]
...but they're probably all bad, so better DO SOMETHING ELSE ENTIRELY:
...gtfo of software, you're likely not gonna become an "agents hearder" with skillset, mentality and experience - in the US probably going full on on agriculture [recent US protectionism and isolationism will give you decent levels and shield for globalized markets], learning some minimal hardware tinkering to automate drones and later manage android workers, software for planning farming automation etc... hire hands for physical labour and BUILD AND MANAGE A FARM or something like that (maybe farm + restaurant or smth else form tourism / hospitality)
E: Farm automation probably has some juice though, regardless of how close the androids I keep seeing in demos actually are.
In the 1980s, we also relied on abstraction during development, but removed much of it as we moved closer to the hardware.
Abstraction wasn’t something permanent — it was something we used and then deliberately reduced.
It sounds more like a packaging issue. I know he's attempted to edit his resume, but there's missing information here that OP may not even be aware of.
For instance, I recently became the last of two candidates interviewing for a great opportunity that I sadly lost. When I received feedback, it turned out the hiring committee had a completely different sense of one aspect of my work than I had attempted to convey. I'm glad I got the feedback, but it was frustrating to lose after so many interviews.
Then just recently, I interviewed a candidate at my current company who reminded me of OP. Laid off worker, very nice guy, but he had no idea how to portray himself as a dev at the level he was applying for.
I wanted to call him up and coach him, but it didn't seem appropriate, especially since he didn't ask for feedback.
If you are in this position, find a free coaching program that can help you revamp and resell what you have to offer.
It's not fair to have to do that just to get a chance to be paid a fair wage. But companies get thousands of resumes a month and do dozens of interviews.
We try to give candidates a chance to show us who they are, but if what they are showing us doesn’t line up with the role, or their strengths are buried, there’s only so much we can infer. It sucks, because the resume and interview are not the job. But they are the gate you have to get through before anyone sees the work.