The systems analyst analyzed the problem that needed to be solved and designed a solution involving a computer, including system modules, inputs, outputs, etc.
The programmer, supposedly a lower paid profession, transformed that into Computer code. I remember someone telling me that wasn't the best choice and that I should become a "systems analyst". They wore the better suits.
My college even offered a shorter, cheaper degree for people who just wanted to be "programmers".
I think this distinction still survives in government labor statistics.
Perhaps we came full circle with LLMs partially killing the "programmer" job.
Words are just words. In tech especially, it seems anyone can call themselves whatever they want.
Someone who chooses to call themselves "programmer" can be more engineering-minded and see the big picture better than someone that calls themselves an engineer.
I also started out in systems programming and we still call ourselves programmers there, when there is a lot of systems "engineering" (aka trying not to blow up the system with your code running in supervisor state) going on. And there, a lot more attention paid to stuff like IEEE code of ethics, professional development, rigorous designs, etc.
So I'll keep to calling myself programmer amongst my non software engineering friends. It just seems more humble.
That being said, writing the code has like others have pointed out, never been the largest part of the job in my experience. So I can see how it may be confusing to some, especially software engineers. But to non software engineers I think they mostly get it...you are a software engineer trying not to imply some some sort of unearned intellectual superiority.
Who's laughing now? Who still has a job now?
As of September, this is objectively false. When "programming in the small", frontier LLMs write better code than 80%-90% of human devs and are much more thorough about checking their work than most human devs. They are also starting to exhibit signs of good judgement in design and architecture, not just implementation, which is why you need a software architect in order to steer things, but hands-on-keyboard programmers? Yeah, you need far fewer of those now.
When "programming in the large", yeah, they're not there yet. But that's coming.
I think both definitions are still too narrow for what I do daily. I fully agree that we still need humans to decide what to build, how it should all fit together, etc.
But both definitions do not really take into account the huge social dimension of being a good (computer) engineer in a company. You have to translate requests from a wild variety of users into what this means for a system. You have to recognize and explain that what they need is not always what they think they need. You have to be able to say "no" to things. All these things are very hard for an LLM (reading humans, not literally following requests).
Then there are social aspects like building consensus within a project, building a user base of many people that are initially skeptical, using existing relations with people and other companies to slowly adopt what you are building, etc.
Programming used to be a large chunk of the work. I thoroughly enjoyed this aspect of the work when I was younger, but at some point I started enjoying system design and everything around it more. So in that sense, programming with LLMs is much nicer (though I still review every single line of code and often have to go through a couple of iterations because even Opus 5.5 often doesn't get it 'right' in the first take). But the remaining part of the work is extremely hard for an LLM to replace.
How do you justify saying "no" when there is a magic machine invented 3 years ago that'll do it? Your gut says it's a maintenance nightmare, your LLM says it's already done implementing it.
I regularly meet (recently examined) people who are trained "software architects". Common among them is that they are often delay projects championing their pet idea, sometimes bringing negative value to the project. The best programmers are just so much better software architects also. So I expect history to repeat at this point.
You don't forget CS concepts so easily, but all the syntactic tricks of new language variants and working set vocabulary of APIs, data formats, and frameworks keeps churning away from whatever you last had contacted in a real, hands-on way.
It is fascinating, wondering where this is going as programmers turn into their own middle managers on top of agents. Particularly, the school who claim that you don't really need to understand all the code and instead treat it like a black-box experiment. If this fad takes over the main stage, then everyone becomes one of those less technical managers.
But, in this new scenario, there is no human in the loop to inject judgement calls and "manage up" anymore. There is no programmer culture except the models and agents, which don't really have the same social connections to drive a programming culture. What will happen? A culture per agent product? Per enterprise client of the agent vendor? Per agent operator? Per codebase?
About half a year ago I noticed there were no people called 'Agile something' at our org, so I asked about it. They looked at me like I was a space alien. I remember, it was like 3 years ago that management bet the house on 'Agile transformation' talking about how Agile practices are the bread and butter of a good organization.
Now they're talking about AI the same way.
The memo has NOT gotten around, though AI now makes it easier to make all the various slide decks and schedule all the 50+ person meetings that power this process.
Negative value to the project, perhaps; but if these "software architects" are kept around despite several rounds of this, it may be because they're providing positive value to the company, in terms of producing for the company a thing whose existence (no matter how flawed) serves as collateral for:
- the executive team, to show to investors, to convince them that they're acting strategically in the market; or
- the marketing team, to get breathless puff-piece articles written about the company, in places where dumb CTOs/CIOs who make IT buying decisions will read them. (Not necessarily to then buy the specific product that came out of that project — hopefully not, in fact! — but instead to remember the company itself the next time they're comparing vendors for something else);
- or the executive team again, this time to show to a potential acquirer, to convince the acquirer that their employees have a higher asset value for having a certain skill / experience with certain paradigms, that they gained working on this project.
Companies have spent collective billions (trillions?) of dollars on projects they know will go nowhere, just to be able to say they did a project that is made a certain way or does a certain thing or has certain certifications, and to have a shitty codebase they can pull out of their back pocket if needed to back that up. In all such cases, even if they need to present the product, the product will never be evaluated for its quality or for its fit-for-purpose-ness; so it doesn't matter that the project failed to produce something of quality or that is fit-for-purpose. It just matters that the project happened.
This is a holdout from the time when the whole software development field was still stuck with the idea that waterfall/big design up front was an adequate way of doing things.
The belief was that, like any engineering field, systems would first be designed in detail before anyone could even open an IDE.
This approach was quickly proven to be wrong, expensive, and fruitless. Agile started to emerge as an effective approach to the problem of developing and maintaining software systems. The uselessness of the system analyst role, when contrasted with the fact that mere programmers could easily implement features without guidance from an enlightened individual far removed from the problem domain, was readily noted.
Milt realized as early as the 1970s, if not long before, that programmers had way too much power in an organization because most organizations adopted a "tool-oriented approach" to computer technology: here's a tool, how do we make use of it? They let the programmers run wild, and the result was software that was a buggy mess. What's needed is a management-oriented approach in which the management of the company calls the shots on what's needed for the company to function.
In PRIDE, the systems analyst is centered. This is explicitly a business role, not a technical role, and their job is to understand the business in such a way that the information systems (not programs, nor apps) that drive the business can be described in diagrams and text. Only once the requirements, definitions of terms, and data definitions have been established and standardized can the implementation phase begin. Bryce estimated that if you're not spending 65% of the effort in systems analysis and only 15% of the effort in programming, you're doing it wrong.
This is how grown-ups did information systems for decades. And then the internet happened, "hacker culture" happened, and the programmers were back in the driver's seat, producing things like "Agile".
Now we have AI, and to quote another commenter elsewhere in the thread, the programmers have had their tails cut off. That 15% number asymptotically approaches zero. But the systems analysis still needs to be done, and for the foreseeable future anyway, that needs a human touch.
Become a business person, not a "software engineer". They are the ones who will drive things forward in the agentic era.
A counter to this is "requirements change quickly, therefore the system needs to change quickly too." But the actual case is: most organizations don't really know how to do systems analysis. So this results in one of two scenarios: either the organization wastes a lot of time and money in the analysis phase struggling to do what they cannot, or they punt on requirements analysis altogether and wait for requirements to surface at implementation time. This is called "Agile" today, but was described by Milt in 1983: "do a superficial feasibility study, do some quick and dirty systems design, spend a lot of time in programming, install prematurely so you can irritate the users sooner, and then keep working on it till you get something accomplished."
Today, with AI, there is no reason not to do requirements analysis and design first. An implementation phase that used to take months can now be performed in weeks or even days. So an executive who observes shifting requirements can document them, work with an AI to update the design specification, and then have a swarm of agents perform the implementation work, and roll out the new system before next quarter if not next month. So Agile really isn't necessary anymore.
PRIDE is like a textbook on how to build a system, from requirements analysis through design, implementation, and maintenance. It was the first methodology for doing so, and it transformed information systems from arcane art into a proven, repeatable science. It should be the industry standard, but is now at risk of being forgotten, largely due to programmers holding the industry hostage.
AI is taking programmers out of the equation. PRIDE is relevant again.
Maybe that will change with LLMs, but you’ve got the causality arrow backwards. Systems analysts didn’t disappear because of Agile; Agile exists because that way of working didn’t work.
Source: I was on the front lines of the first Agile wave starting in late 1999.
(And then what people actually adopted as “Agile” generally wasn’t what was created by the first wave of the Agile community, but that’s a whole ‘nother topic.)
But there's a more general kind of "systems analyst" that still survives (and is mostly what the BLS stats refer to): a person whose job is to map out the current business domain of a system (consisting of not just computers, but also people, infrastructure, institutions, etc.) and identify where it will need to be modified (but not necessarily how it will need to be modified) to make it capable of being put to use in a certain way for a given plan/mission.
Systems analysts in this sense tend to be one of the actual workhorse jobs inside "management consulting" firms and "think tanks." Governments, militaries, and large bureaucracies with many independent business units (think IBM) are their customers.
Systems analysis is sometimes done structurally over an entire organization: someone at a very high level has some new vision for an initiative, and hires these companies/think-tanks to help them determine how to "align" their organizational architecture with what a military might call a new "doctrine" or a company might call a new "paradigm." In these cases, the systems analyst performs the first step of that alignment: understanding "the system as it is today", and identifying what parts of it are already aligned vs. which parts will need to change. Their output is a report to the person driving the process.
But structural analysis is more often applied to specific strategic/business units, as the result of that strategic/business unit being chosen to execute a given strategic mission, as an enablement process to get that unit into the specific shape it needs to be in to execute that mission. In this case, there is likely an assigned "mission {planner, architect}" (who may also become the mission's director; or who may hire for such a director role and then depart to go plan the next mission.) The systems analyst works for/with the mission planner as their "eyes and ears", interacting with the business unit to gather information for the planner to use.
In both the structural and strategic cases, there are usually also one or more systems engineers who determine the "how" for implementation of these changes. (A mission planner is usually a very senior systems engineer themselves.)
However, unlike the systems analyst, who is considered more trustworthy to central leadership when they exist outside of the system they're analyzing (and so mostly work for external firms like think-tanks rather than for the orgs themselves), there's much less of a motivation to separate out systems engineering in that manner. So systems engineers (including mission planners!) tend to be regular [if high-paid] career employees of the central organization. These folks are frequently "general staff" in a military (with no real analogous term, AFAIK, for the equivalent people within a conglomerate's parent company.)
Due to their value coming in part from existing in separate firms that can negotiate on their behalf, systems analysts in this sense are well-paid indeed! But also worked hard, bounced around, and signing NDAs constantly. And possibly a bit bitter, in that they often see how things are broken with no chance to be part of the fix.
This is not legal advice but allegedly lawyers were instructing CSA category TN applicants to never ever mention that they are “coding” or “programming”
AI has already delivered huge productivity gains in my opinion; the whole StackOverflow/Google search part of my life is gone. Instead I get answers from Claude tailored to my specific environment, use-case, problem, etc. Writing well-defined, simple code with lots of prior art in the same repo is easily handed off. But the AI industry and a lot of leadership in big tech seems to be expecting further miracles from AI. Maybe one day a CEO or CPO will just be able to speak their thoughts and a fleet of agents will implement it flawlessly, but we are nowhere near that reality right now.
I think this is a different point, though. OP is saying, we haven't really made progress on the problem of figuring out what to build (and just as importantly - what not to build) and how to convey that effectively and set up the right feedback loops to get AI to actually deliver on that. Whereas you're saying, "Yes, I agree, the code that the agent writes has flaws in readability and design". If we can get the first part actually solved (and I'm not sure what an answer looks like - other than what we have today, which is that The Market ultimately dictates whether you've built the right thing, built enough, not built too much), then nobody other than the AI agent actually cares about solving your problem (readable, maintainable code) anymore. Or to put it another way, it's only in the agent's interest to solve that problem, for the same reason that today it's in the SWE's interest to solve that problem, rather than the product manager or the business owner.
I believe that LLMs will, eventually, reach a level where they can provide decent architecture, but I suspect that will suffer the same issues affecting code generation. It will be "decent," but not "awesome." I think exceptional people will still be able to beat LLMs at architectural design.
The issue is that the definition of "exceptional," means that it doesn't happen very often, and most people manage to convince themselves that they are the exception.
AI has absolutely nothing to do with any of this. It merely accelerates the "how". You still need a customer to determine everything else.
The path to success in software is varied, there is no one hallowed way. Saying everything must start with talking to customers risks eliminating a whole swath of useful creative approaches. A dev slaving away for months or years on their own pet project can still come out with something that achieves major success.
Second, I'll concede that revolutionary innovation happens primarily top-down. Common folk don't know anything about the state of combustion engines or transformer architectures, or how those can be composed into new categories of product.
Third, odds are if you're building software, you're doing nothing of that sort, but rather dealing with infinitely more modest innovation. You would be, in fact, improving a horse, if that were at all possible.
Fourth, when improving the proverbial horse, you'd better have an outstanding sensitivity to the finer problems that customers face. Whether you develop that sensitivity by asking, or sensing in some other way is another question. But asking may not be a bad idea at all.
Fifth, I'll close with an anecdote. The worst seed-stage software startup tire-fire I've seen in the last couple of years was led by a CEO that published the Henry Ford quote, verbatim. A couple years in and they still don't have a paying customer. I'll generalize by saying that I've seen pretty much everyone around me with "the don't ask, envision" attitude fail, while the people who ask or are otherwise exceptionally attuned to the customer problem do much better on average.
The quote is a great example of imagining user needs instead of just asking them.
You don't think that AI can (or soon will) be able to talk to the customer and discover what they want?
I don't know if this is a perfectly fair of their business model, but it has that feel in terms of dealing with them.
I'm surprised people are agreeing with this take. LLMs can model things at a conceptual level, and they are actually better than some software developers at this task (though, not better than experienced developers). There has been progress on this specific ability.
Now that execution is more of a commodity, everyone seems to be saying execution is cheap and ideas are everything.
I don't disagree with you. I'm just noting an observed 180.
I’ve never been one to trust the VC crowd without knowing where their money lay.
Derivative ideas? Like "let's make a $thing with $variant"? Yes those are cheap, and AI made them pointless to pursue because most people can now whip up a hacked up variant of $thing with GPT and be happy with it, until they realise that it didn't add that much value to begin with.
But good, truly good ideas? Those that turn everything around? Those are cheap in the sense that your brain doesn't consume that much energy having them, but they are rare.
LLMs kind of confirmed we're indeed constrained by ideas: the homepage of HN isn't full of things like "we reimplemented X in Y" because we were full of revolutionary ideas to begin with. IMHO LLMs have further hampered the spread of truly great ideas - because nowadays if you have a truly great idea you have a fork:
- your idea can be quickly implemented by an LLM. You do it, and then everybody simply copies it using their LLM.
- your idea requires time, iterations, ... so the LLM speeds up delivering it, but only a bit. We're still on a path similar than before - even before you could just hire more people and try to deliver stuff faster, as long as you had all already clearly planned beforehand.
So yes LLM are a big boon to productivity, but IMHO cast a long shadow of "why would I do it at all" on things. In the before times, I'd spend my day coding a dumb library for fun, just to brag about it a bit on Github and HN. But given that nowadays LLMs have reduced the value of code to ~0, the one thing that still has value is the idea behind. Which now people have less incentive to put out in the first place.
I'll just observe that execution remains ridiculously difficult despite AI. Even building the software itself. Then there's all the other stuff, mostly relationships (with investors, customers, employees, lawmakers, society at large) to which AI contributes nearly zero or may even cause damage if used recklessly. The amount of posts I'm seeing from VCs complaining about vibeslop decks!
(Well, they still need handholding and regularly demonstrate spectacular stupidity... but so do we, and we've been getting along just fine!)
The former are fun but irreproducible ; the latter are boring (index funds, sobriety, compound interest, don't be stupid, and some luck to avoid the bad timing of the crashes) unless the author adds some snark (again, Munger)
I expect it will be more than 2 people trying that, and almost all of them didn't become multi-billionaires.
So it's the difference between "depending on things being average" and "depending on things being exceptional".
[1] https://root.cern.ch/TaligentDocs/TaligentOnline/DocumentRoo...
It was my first years of being a "software engineer", versus the solo hacker kid who liked making art of over phone lines. So I devoured books which would teach me how to make software with other people and for other people.
Books like the "BeOS Filesystem book" (Practice File System Design) was another classic from the era [1]. It was like today's LLM white papers, learning how a company did something unique, except companies seldom published them because it was a time of IP hoarding.
Unfortunately the books that eventually became the most practical were those by Sutters and Meyers; compendiums on the warts and jewels of C++ that you needed to know to be good at programming in it. Such nitty gritty, much of which was eventually fixed by language design. I really don't miss having to know all that, freeing my mind to consider architecture, design, and delivery.
[1] https://www.nobius.org/dbg/practical-file-system-design.pdf
I think marginally is an understatement though because it feels like it is compounding. I don’t know other people’s experience but I’m building software that would not have existed before LLMs and coding agents arrived.
And this software is without a doubt better from the end users perspective. Maybe the code is crap, who knows? I don’t look at it. I judge based on 25 years of writing software by hand, deploying it to real users, and being a user myself.
I hope there is a resurgence of indie software like existed in the early 2000s.
There’s no hope for enterprise software. It will always be terrible.
I implore you to look at it! I recently vibed too close to the sun: I knew at a high-level what I was doing, but never looked at the code.
Two weeks after and 50k LOC, I’m having a look and I can’t believe my eyes. There are whole failure modes that don’t need to exist. Moreover, how can I sale this software in good faith, when something happens and I’m largely a meat proxy because I don’t know how the code works?
I will go through the entirety of the code with my naked eyes, and make improvements (with AI).
But this time, I won’t fall for the illusion of “productivity”.
If you get into the regular habit of having the LLM look for issues with the code base it will find them: structural, bugs, design pattern issues, duplicate code, or same functionality not sharing a single path.
Honestly it feels like working with a real codebase. Tech debt always builds up, and yes the AI does it, too. It does notice if you ask to look.
When I wrote features by hand, I used to have like 2-5 bugs. Most of which were just human clumsiness.
When I started using AI and remained in the driving seat, most of the human clumsiness bugs went away.
Now, when I vibe, all I have are my verification loops and asking Opus5.5 to find and fix bugs.
Opus tells me it’s done. Only for me to start a new loop and it finds 10 more things or maybe I go in myself and it says “oh you are right”.
For me, this is not progress. I would rather spend an extra 1 or 2 hours a day in the driving seat building software that is correct and that is so simple that cost of changes is low.
It wrote them to begin with. Now you're handing of this discovery to literally the same LLMs that wrote this.
Oh, they will find some issues. Are you sure they will fix them and/or rewrite them without introducing more issues? How do you know if "looking is not necessary"?
I have automated the entirety of testing and Claude can validate it's own inputs. I then focus on marking requirements fulfilled, or bugs found. The bugs it does find are usually on the side of incomplete requirements. The few technical ones were due to a lack of testing. And I am not perfect, so I would have made all kinds of bugs myself.
If you keep to a mindset of, and build processes for, the management of software development rather than coding, results are better. No this isn't a magic prompt like "make no mistakes".
It is approaching the project with the mindset of knowing the limitations of the tools like one would a junior/mid engineer, and then mitigating. This is dependent on the codebase, product, and desired requirements for what processes to use and to what degree.
If you find yourself still needing to look, you have insuffiently automated the creation and validation of written software.
Its solved. Adapt and learn to use them effectively or get left behind; seems to be the unfortunate reality.
When I release features, players report minor bugs, which I can fix in minutes by simply dumping their complaint into claude.
I worked on this codebase prior to LLMs and made slow progress, with many bugs. I did not know how to approach the performance problems, nor was I able to refactor the project to effectively have test and production environments reproducible. Claude did just fine.
It is 350k LoC, with over 138k as an Elixir app, 60k in tests/scripts, 65k Vue components, 45k JS, and more.
Coding is solved. Many people can't manage software engineering effectively. I don't know how much larger or how much more comparison is needed before this is accepted. Maybe it is "circular" reasoning like the dead comment says.
But it literally just works for me.
Every time. Every single time. People who say they never look at code anymore always say the trite things about "management of software development", how they have no issues, and that if you have issues, it's the problem with your skills with a side of snide remarks about your intelligence.
And then you press them for details.
After a few rounds of more and more trite answers they either go away or it turns out that their "automated everything" and "processes" and "coding is solved" are:
- a bunch of "make no mistakes" aka untold numbers of random markdown files aka skills, "guidelines", "memories", "ADRs", "technical docs" etc. Which are routinely ignored by any and all models.
- a bunch of tools that can't or don't catch all the mistakes the models routinely make. E.g. there are very few tools that reliably detect code duplication (especially code duplication with subtle changes in behavior), or reliably detect bad data access patterns, or reliably detect suboptimal algorithms etc.
- an overreliance on tests. The very same tests that are written by the very same very models
Also, if the models could actually "validate inputs" and other "coding is solved" things, you wouldn't actually need dozens of tools around to catch the mistakes they do. Ironically, you need "processes" and "engineering management".
---
As a related aside.
I have multiple tools from linters to semantic analysis running as routine checks, and still. Here are the responses by Claude for some of the things I caught it doing in just the past few days:
- "The [test data] generator can sometimes fail, and create duplicated data, so I increased the number of times it runs to 100" (instead of fixing it)
- "The tests check this in an indirect way. They send megabytes of data and watch the memory of the whole VM. Unrelated processes change that number. The direct check is the number of body bytes that the server read. Do you want me to rewrite these tests" (Claude wrote all these tests to begin with)
- "The previous research covered only a few CSS files. It did not cover the design system documents, the components, or the pages that already use them." (so much for "No this isn't a magic prompt like 'make no mistakes'" as it had an explicit requirement to use the design system)
- "The too_large_message/1 helper has three copies, two of them private, with no good reason for them..."
And that's just some I documented for myself.
I don't use memories, markdown files, guidelines, etc.
I design the software like I would as an principal engineer and dictate the high level requirements. Opus 5.5 really does just work. 10 months ago this wasn't quite true, it required more steering of the technical requirements.
I don't know why your results are different. I run a live multi-week long campaign game that streams updates over WSS from an Elixir app. I can literally see if Claude is doing as it claimed. It is trivial to verify when it gets it wrong.
I dunno what some others are doing to get such poor results. How much larger must my codebase become, how much more complex before those that see poor results see that it really is solved?
And if 350k LoC isn't enough for your use-case. Well, that wasn't quite achievable 12 months ago, but it was 8 months ago. In another year, I imagine if my codebase grew to 1M+ LoC, the latest models would still be keeping up.
Its solved. And the number of codebases it can't apply to shrinks every 12 months.
It doesn't have a summary of the codebase. Or at least a useful summary.
> I design the software like I would as an principal engineer and dictate the high level requirements
Let me quote myself:
--- start quote ---
And then you press them for details.
After a few rounds of more and more trite answers...
--- end quote ---
This is now your third reply with trite answers (two to me and one to a sibling comment).
> I don't know why your results are different.
Because I actually look at code it produces.
> It is trivial to verify when it gets it wrong.
Yes. Yes it is.
E.g. in my codebase it is trivial to verify that it does 15 reads from the database on a page when just 2 or 3 would be enough. How do I know it? I check the logs, or see it in the REPL (I also use Elixir).
It's also trivial to see when it does shit like "oh I didn't read the design system". Because I actually see what it's doing.
It's trivial to see a bunch of other issues it routinely produces because tools like these exist: https://github.com/elixir-vibe/ex_dna and https://github.com/Cinderella-Man/credence (or even my own vibe-coded https://github.com/dmitriid/llamex and https://github.com/dmitriid/ex_css_analyzer)
> And if 350k LoC isn't enough for your use-case. Well, that wasn't quite achievable 12 months ago, but it was 8 months ago. In another year, I imagine if my codebase grew to 1M+ LoC, the latest models would still be keeping up.
Yeah yeah. "Keeping up". Where by "keeping up" you mean "I duplicated the code 15 times, and "I didn't read even half of what is needed to do this task".
It works for you because you give it only the most superficial tests. Wouldn't surprise me if you can cut all of that code in half and raise the fps 5-fold. But for that you actually have to look at the code.
----
BTW lol at the tests. Literally 10 minutes ago it produced this amazing test: https://x.com/dmitriid/status/2121723533567115357
I mean. Nothing wrong with it is there? The code works. I can see it work in my browser. What other checks do I need lol.
Or "high level tasks".
Me: "Do not show invites panel if there are no invites, otherwise it shows as a horizontal line"
Claude: "The hidden card keeps its border because the box-lx { display: block; } rule overrides the browser's [hidden] style. I'm adding a primitive rule that sets display: none for hidden next"
You know, instead of not rendering it in the first place.
Have a fourth, then.
I hope you continue to ship for use-cases your users care about in your work, and continue to find satisfaction in software engineering.
I am.
All I can say is: there's a reason your software had multiple performance issues, couldn't support larger maps, and couldn't get past 20fps. It's the same reason why you're so impressed with the results and don't look at code anymore.
Adieu.
Every model, including $latest_model, makes mistakes and hallucinates things. If you aren't looking at the code, you are producing bad software, end of story.
> Its solved. Adapt and learn to use them effectively or get left behind; seems to be the unfortunate reality.
Ah, there's the FOMO push. It just isn't an AI bro post without false claims of "you're going to be left behind!".
What I’m most anxious about is that, because I don’t read code, if I don’t have access to an agent for any reason I’ll have a hard time fixing a critical bug because I don’t have a mental model anymore of how things work. And that’s not professional and I can’t sell that software.
I think this is a growing problem. I pulled back. I'm much faster than I used to be. BUT, now I also forcibly insert myself into the loop so not as purely fast as I could be. I've thoroughly iterated on a design with the LLM before starting. I've personally looked at and often iterated on every line of code before merging. I regularly walk through the generated code base, both with and without LLM assistance, in order to ensure I have a solid mental picture of what's there.
This behavior has been informed by seeing things go to hell, both for myself and with other people, clients, open source projects, etc
When you have a deep understanding of something, you just can’t help thinking about all the ways you could improve it. Maybe you realize it could be generalized, or that a different problem you’ve had could be solved with a similar technique. Sometimes you get oh “OH DUH” moment where you realize you were making the slightly wrong thing this whole time, and NOW you know EXACTLY what to make.
With LLMs I just get basically what I asked for, and then I’m done. Obviously this is sick and I love it.
But for a lot of my projects recently, I fall into this trap
- I have a problem, I think a program could help. But I want to know if there’s a viable approach before I invest a bunch of time by myself, so I boot up claude
- LLM writes the entire program, I don’t know what’s going on in there
- the program kinda solves the problem or it doesn’t. At this point I already have something that has vaguely confirmed or denied the viability of solving the problem, so I think “welp I’m not gonna spend a whole week of my life reading an LLMs output”, and that’s that
I rarely have cool insights about the problem after that. In the olden days, even if my approach didn’t work, I would come away with a new understanding of that problem, and all similar types of problems. “Oh that’s why people say cache invalidation is unsolved”. “Oh that’s why generics can’t infer xyz”
Don’t really get that when you vibecode. But the siren song of just having something that works is so sweet, and I can never ever resist it
This! Especially this! is the real reason I started building my own commercial software again.
I’m betting that the details will still continue to matter.
Absolutely true! Some month ago I discovered huge memleaks by coincidence because Opus nailed tons of LINQ calls together and duplicated data as hell.
It is. And the trouble is, once you do that, the speed gains evaporate. It takes just as long to carefully review the LLM's code as it would've to write it yourself, because the typing wasn't what took time, but the understanding of the problem. The only way to get productivity gains from an LLM is to vibe code, at which point the quality of your software craters.
"Technical debt is a qualitative description of the cost to maintain a system that is attributable to choosing an expedient solution for its development.
While an expedited solution can accelerate development in the short term, the resulting low quality may increase future costs if left unresolved."
AI is right now buying down huge amounts of technical debt (see the flurry of CVEs). But we're already onto the self-improvement treadmill: future more capable systems will be better at solving any issues introduced today - it's literally not worth the time worrying about now (this is the same reason you don't get a cancer prognosis over a longer then 10 year time frame - the expectation is future treatments will be better).
The technical debt my code written in Opus 5.5 might have is already going to be handled by Fable 5.2, for example. It's going to be better and faster then I am. But also: if the rate of technical debt production from AI is even slightly lower, then it gets more and more manageable in a future where nothing changes but the increasing volume of AI compute is concentrated on less and less of it.
I mean certainly it would be nice if I could implement a feature without Claude finding a security bug and tripping cyber protections every couple of hours.
No, we can't. We can believe this will be the case
> The technical debt my code written in Opus 5.5 might have is already going to be handled by Fable 5.2, for example.
"Let's generate more and more garbage in hopes that there will be some magical tech that solves the issues" is not that good of a strategy.
E.g. even for the amazing magical benchmark-shattering Fable 5.5 Anthropic's guidance is still the same: your CLAUDE.md should not exceed 200 lines and claude only reads 200 lines of memory index because otherwise it quickly loses adherence.
Like all problems, we'll solve them when we get there but its not like anyone has an abundance of useful alternative advice: I'm getting useful software out now, whether I have to fix it later has the same problem it has always had - when we need to, we'll do it.
It's already here. People are already producing unmaintainable slop at never-before-seen levels, and software reliability is at all-time low.
> its not like anyone has an abundance of useful alternative advice
Yes, the advice is the same: look at the code produced and understand it.
> whether I have to fix it later has the same problem it has always had - when we need to, we'll do it.
Yet you don't. You're quite literally saying "who cares some future model will solve this for me".
But the model I buy today is cheaper in terms of capability then the one I could've bought a year ago.
But software can be "done". There isn't an endlessly acruing technical debt, the product can be finished.
Yes my frontend is nicer. I do not hang myself up debugging a small issue for hours. I also don't have to search questions online, and get even complicated answers fairly quick.
But I'm not sure if I ended up being so much productive, or I'm just building a lot of open source stuff nobody needs. Last year i would have solved a narrower problem and relied on others to contribute the rest.
Why hope for a resurgence of indie software without also a revolution in enterprise?
If there ever was a time for either entrenched players to improve their UI/UX (using LLMs) or for rival upstarts to build lots of feature parity, now seems like it would be the time.
Perhaps if Agents are given some autonomy to buy enterprise software trials it would be easier to break the stranglehold of the biggest vendors.
Where I'm finding AI exceptionally useful is now being able to address tech debt and developer experience for my team WITHOUT having to argue for spots in the backlog to do it.
So, if nothing else, it sucks a lot less to work on enterprise software if you can let a team lead or principal engineer loose from the PM to just fix things.
But, how is AI fixing this? Unless the perverse answer is: no one’s reviewing code anymore so I can just merge my technical fix.
So "Where does this function get called from", "What does this function do?", "Rewrite this function as-is with better variable names", "Make a document outlining the structure of this project", or "Review this PR and watch out for any places that may be lacking the fix" are all reasonable things to ask of an AI.
When you're in a big project and aren't sure where this is supposed to go or how things get A from B, AI can be extremely helpful.
In which sector is your company active where this is allowed? Or do you run on your own infra?
Yes, I imagine this is trickier in more sensitive applications, but I'm sure this isn't a new concern. AWS has Private Cloud, surely you can get a LLM with whatever assurances a more sensitive project needs at this point. But in my current environment that's not really a concern, so I've not personally looked into it.
Nit: not all of those questions require statistical language models. IDEs like IntelliJ have been adept at syntactical queries and refactors for many years using purely deterministic data structures and algorithms.
Like if you have some code that's been intentionally obfuscated, or made by somebody who names everything "a", "b" and "c", a LLM is likely to be of at least some help.
And yeah, you can do things like finding all the places where func() is called, but a LLM can also be helpful in saying where something needs to be done, which is a much fuzzier task.
Usually, If you don’t know that, you shouldn’t modify the system either. Just because you know how to unscrew a bold, you shouldn’t go about removing the head cover of an engine. Locating the area of interest takes a lot less time than actually understanding how it integrate, the guarantees provided by other systems and the guarantees that it offers.
Modularity in software architecture. By reducing the links between parts of the code, it’s easier for someone else to navigate the system and answer these questions. But it does require careful design and caring about the code. This and other things is what the engineering is about.
In my own projects I'm also quite capable of forgetting stuff. Double checking with a LLM whether I forgot about one of the more obscure corners doesn't hurt.
That can be a sign that what you're now building is perhaps not worth building? At least my experience from dayjob, we're building more and faster, but it's either things we really shouldn't do (stuck in backlog forever for a reason), or things that just complicate the main product when features get added willy nilly.
Of course this calculus drastically changes when maintainability and other stakeholders come into the picture. There vision an architecture are more important than adding features. Architecture isn't too hard to solve, but sadly good product owners have always been hard to find
I have the opposite view. Why would anyone build and sell software now when anyone can one shot anything? Genuinely interested in people’s thoughts.
The value added by commercial software will have to be much higher than what it is now, or most companies will have little reason not to code their own, fully customized to their own needs.
But it is really hard to imagine what the future of software will be, even in 2/3 years time. We live in wild times.
A company usually wants support / integratoin / maintenance / updates / new features etc. - for sure there will be some companies building 100% on their own, and yes I guess this number will increase but I do not see any mid-level company or above replacing Oracle or Salesforce or Microsoft with their homegrown solution.
Which is to say, the form of your statement I'd agree with is: if anyone can make an MVP in a few days, what use is an MVP?
A church is going to pilot it to 450 people. Some people actually do find it useful.
https://jaisenmathai.com/articles/sojourn-for-ios-was-45-one...
But the fact AI has made things easier for most people doesn't mean the barrier to entry has been lowered to 0 for everyone. My mom could probably use AI to write a compiler but first she'd have to even understand what a compiler is.
And the fact that some apps can be made in one prompt doesn't mean every app can. I've been working on a fully AI generated project for 8 months, it's taken billions of tokens and continuous prompting. The amount of human time spent on this project is just as much as I'd spend before AI, just now the project is much larger in scope. So even if some people have the ability, they still might not want to spend their precious human time on it.
I agree that the ease at which it can be created feels a little…how do you say it…demoralizing.
In the long run, I don’t think this will be a net positive. But in the scope of this conversation about absolute productivity and code quality I seem to disagree with a lot of HN.
...the problem is instead of doing that people take the first prototype and ship it
1) Knowing what to build, keeping the design appropriate, not over-engineering the spec, not prematurely optimizing, et al. Note that this is different from just having an idea about a product.
2) Making it.
3) Getting people to use it. Nobody gives a shit about your software. Getting them to even try it is extraordinarily difficult in most cases. Selling is hard work. This will get a lot more difficult.
#2 has had the difficulty dialed back in many cases. In exchange #3 will increase drastically.
Absolutely nothing has gotten easier net. If anything the total will become harder net, the difficulty of acquiring users will get harder beyond the extent to which making it gets easier - users will trust software even less in the future, and the popular AI systems (personal agents or whatever they end up being called) will reduce the amount of software the typical user ever needs across a lifetime (narrowing the market).
Making architectural mistakes is now exponentially cheaper - as you can refactor the entire code base in a day.
So even if LLMs are not quite there yet in terms of architecture, they definitely lessen the time one needs to spend on making the right choices.
and in areas like FE for more internal tools design really becomes a non issue.
When AI solves hardware design, then we might see some new room for software again.
Totally agree with this.
I see the future of what an IDE is more like an "architecture designing interface" which designs tests, architectural decisions, type systems, and contracts between components. Tasks then should be delegated in their implementations to agents so they are just responsible for the "wiring parts together" tasks.
Currently I imagine it as something similar to how workflows are usually designed around strong types and schema validations in safety-critical areas like ASPICE, Ada Spark, etc.
That's mostly why I don't use existing agentic environments and started earlier this year to experiment with my own one, where I try to enforce the design and architecture related tasks through capabilities mapped on tools for each of the task-scoped roles involved.