If you're using llms to shit out large swathes of unreviewed code you're doing it wrong and your project is indeed doomed to become unmaintainable the minute it goes down a wrong path architecturally, or you get a bug with complex causes or whatever.
Where llms excel is in situations like:
* I have <special snowflake pile of existing data structures> that I want to apply <well known algorithm> to - bam, half a days work done in 2 minutes.
* I want to set up test data and the bones of unit tests for <complicated thing with lots of dependencies> - bam, half a days work done in 2 minutes (note I said to use the llms for a starting point - don't generate your actual test cases with it, at least not without very careful review - I've seen a lot of really dumb ai generated unit tests).
* I want a visual web editor for <special snowflake pile of existing data structures> that saves to an sqlite db and has a separate backend api, bam 3 days work done in 2 minutes.
* I want to apply some repetitive change across a large codebase that's just too complicated for a clever regex, bam work you literally would have never bothered to do before done in 2 minutes.
You don't need to solve hard problems to massively increase your productivity with llms, you just need to shave yaks. Even when it's not a time save, it still lets you focus mental effort on interesting problems rather than burning out on endless chores.
merlincorey · 1h ago
> * I want to apply some repetitive change across a large codebase that's just too complicated for a clever regex, bam work you literally would have never bothered to do before done in 2 minutes.
You would naively think that, as did I, but I've tested it against several big name models and they are all eventually "lazy", sometimes make unrelated changes, and worse as the context fills up.
On a small toy example they will do it flawlessly, but as you scale up to more and more code that requires repetitive changes the errors compound.
Agentic loops help the situation, but now you aren't getting it done in 2 minutes because you have to review to find out it wasn't done and then tell it to do it again N times until it's done.
Having the LLM write a program to make the changes is much more reliable.
sothatsit · 22m ago
> Having the LLM write a program to make the changes is much more reliable.
I ended up doing this when switching our 50k-LOC codebase to pnpm workspaces, and it was such a good experience. It still took me a day or two of moulding that script to get it to handle the dozens of edge cases, but it would have taken me far longer to split things up by hand.
I still feel like I am under-using the ability of LLMs to spit out custom scripts to handle one-off use-cases.
ch4s3 · 16m ago
That’s not even a very large code base. My experience is definitely that anything with more than 100K-loc really makes the LLMs struggle.
NewJazz · 26m ago
Yeah was thinking about this recently. A semantic patch is more reliable, but prompting an ai might be easier. So why not prompt the ai to wrote the semantic patch.
smaudet · 5m ago
"bam work you literally would have never bothered to do before done in 2 minutes."
And I would never want to use a piece of software written by you ever.
If you think that writing the code was the hard part, your code was probably always shite.
brandall10 · 1h ago
There's also a middle ground, where you have the AI generate PR reviews and then review them manually. So that 2 minutes of code you spat out (really more like 5-10 using CC) takes another hour or three to review, and maybe 5 to 10 more commits before it's merged in.
I've done this successfully on multiple projects in the 10-20k LOC, ~100 file area - fully LLM generated w/ tons of my personal feedback - and it works just fine. 4/5 features I implement it gets pretty close to nailing from the spec I provide and the work is largely refactoring. But the times it doesn't get it right, it is a slog that could eat the better part of a day. On the whole though it probably is a 3-5x speedup.
I'm a little less confident about doing this on projects that are much bigger... then breaking things up into modules begins to look more attractive.
dumbfoundded · 48m ago
It's definitely a middle ground, but PR reviews, are not perfect. So it's easy to miss a lot of things and to have a lot of extra baggage. From reviewing code it's not always easy to tell exactly what's necessary or duplicate. So I agree, this is a middle ground of using LLMs to be more productive. Removing one bad line of code is worth adding a hundred good lines of code.
layer8 · 1h ago
> If you're using llms to shit out large swathes of unreviewed code you're doing it wrong
> bam, x days work done in 2 minutes
This is a bit of a misrepresentation, since those two minutes don’t account for the reviewing time needed (nor prorperly, which vastly exceeds that time. Otherwise you end up in the situation of “doing it wrong” described in your first paragraph.
appease7727 · 57m ago
Most of these cases don't require "review". It either works or it doesn't.
If you have an LLM transform a big pile of structs, you plug them into your program and it will either compile or it won't.
All programmers write countless one-off throwaway scripts. I can't tell you how many times I've written scripts to generate boring boilerplate code.
How many hours do you spend reviewing such tools and their output? I'll bet anything it's just about zero.
the_af · 34m ago
What do you mean "reviewing" throwaway tools and scripts? If you wrote them yourself, presumably you understand what they do?
I've also spent countless hours debugging throwaway scripts I wrote myself and which don't work exactly like I intended when I try them on test data.
p1necone · 1h ago
The implication with that example was it's some editor thing for use during the dev process separate from the actual product, so it doesn't matter if it's disposable and unmaintainable as long as it does the thing you needed it for. If the tool becomes an integral part of your workflow later on you stop and do it properly the second time around.
dymk · 1h ago
It’s not a misrepresentation, they’re saying the time it would take to write the code has been reduced to two minutes, not the reviewing and everything else (which still takes just as long)
seba_dos1 · 1h ago
Reviewing the code you didn't write takes much longer than the one you did.
scarface_74 · 25m ago
Reviewing code another person wrote also takes longer than code I wrote. Hell reviewing code I wrote six months ago might as well be someone else’s code.
My job right now depending on the week is to either lead large projects dealing with code I don’t write or smaller “full stack” POCs- design, cloud infrastructure (IAC), database, backend code and ETL jobs and rarely front end code. Even before LLMs if I had to look at a project I did it took me time to ramp up.
puppymaster · 17m ago
personal favorite of mine - I want to switch data api but I dont have time to port 2 different services so here's their documentation. BAM. Done.
3vidence · 6m ago
YES, this is the way to make AI tools pretty much a strictly positive productivity tool on large codebases.
AlienRobot · 1h ago
Scaffolding is another area where LLM's work great.
I want to create a new project using framework XYZ. I already know how to do it, but I don't remember how to set up it since I only do that once, or I don't know how to set up a class that inherits from the framework because I usually just copy the other from another class in the same project. I can simply tell the bot to write the starting code and take it from there.
The sad thing is for a LOT of use cases an LLM is completely unnecessary. Like why do I even need an LLM for something like this? Why can't I just download a database of code examples, plug it into a search engine that apppears in the sidebar, and then just type "new project XYZ" or "new class XYZ.foo" to find the necessary snippet? A lot of NPM frameworks have a set up script to get you started with a new project, but after that you are practically forced to use Google.
It's crazy that a problem that could be solved so easily with a local file search has been ignored for so long and the only solution has been something impossibly inefficient for the problem it's supposed to solve.
foxfired · 17s ago
I used a similar metaphor in the past referencing "They Machine Stops" [0] by E.M. Forster. Yes, in the near future, we will still be able to read code and figure out what it does. I work on legacy code all the time.
But in the long term, when experienced developers actually feel comfortable letting LLMs write large swats of code, or when the machine no longer needs to generate human readable code, then we will start forgetting how it works.
The article throws out a lot of potential issues with AI generated code, but doesn't stop for a moment to consider if solutions currently exist or might exist to these problems.
- Before LLMs, provided that your team did some of their due diligence, you could always expect to have some help when tackling new code-bases.
Has the author never worked on legacy code before?
- (oh, and it has forgotten everything about the initial writing process by then).
Does the author not think this can ever be solved?
- Because of the situation of a Bus Factor of zero that it creates, vibe coding is fundamentally flawed. That is, only until there is an AI that can generate 100% accurate code 100% of the time, and it is fed 100% accurate prompts.
Why does it require 100% accuracy 100% of the time? Humans are not 100% accurate 100% of the time and we seem to trust them with our code.
AntwaneB · 2h ago
Hello, author here. Thanks for your comment.
I agree with your first point, maybe AI will close some of those gaps with future advances, but I think a large part of the damage will have been done by then.
Regarding the memory of reasoning from LLMs, I think the issue is that even if you can solve it in the future, you already have code for which you've lost the artifacts associated with the original generation. Overall I find there's a lot of talks (especially in the mainstream media) about AI "always learning" when they don't actually learn new anything until a new model is released.
> Why does it require 100% accuracy 100% of the time? Humans are not 100% accurate 100% of the time and we seem to trust them with our code.
Correct, but humans writing code don't lead to a Bus Factor of 0, so it's easier to go back, understand what is wrong and address it.
If the other gaps mentioned above are addressed, then I agree that this also partially goes away.
scarface_74 · 22m ago
How is that any different than any other legacy code where the reason for the decisions being made have long been forgotten?
ndneighbor · 2h ago
Yea, I found the article to be overly reductive. I work on a shared ownership codebase, which at times, I am not going to be able to pull the original author at all times when I work on my branch.
At least having a partner explain some aspects is a huge unlock for me. Maybe the OP shadowboxing a world with no humans, but no humans is already a status quo issue that my team and I face sometimes.
lovich · 2h ago
He’ll, have they never had to maintain applications where the source code was lost? I haven’t had to do it too often, and I wouldn’t claim I’m very good at it, but on more than a handful of occasions I have had to decompile binaries to figure out what the fuck it was doing before I could write a wrapper around it to try and mitigate its unfixeable issues
sfink · 2h ago
The author was just saying that this will become the norm, not the exception. Ok, not this bad -- we can at least expect to have access to the AI-generated code for at least a little while longer. (I can imagine a future where AI programming gets just robust enough that some early adopter types will take their whiteboard sketches and prompts and prose test case descriptions, generate the code, compile it to a binary, then throw out the code. It's more agile! You can tweak the spec slightly and just turn the crank again!)
raincole · 2h ago
> Why does it require 100% accuracy 100% of the time?
I've said this before, but I'd say it again: anti-AI people, instead of AI users, are usually those who expect AI to be magical panacea.
The vibe reminds me of some people who are against static typing because "it can't catch logical error anyway."
bgwalter · 2h ago
No, we treat it as a bureaucratic, distracting and infantile way of software engineering that does not work.
TuringNYC · 57m ago
The Bus Factor was an issue long before LLM-generated code. Very few companies structure work to allow a pool of >1 individuals to understand/contribute to it. What I found is -- when companies are well structured with multiple smart individuals per area, the output expectation just ends up creeping up until again there is too much to really know. You can only get away from this with really good engineering management that specifically tries to move people around the codebase and trade-off speed in the process. I have tried to do this, but sometimes pressure from the stakeholders for speed is just too great to do it perfectly.
Shameful plug, i've been writing a book on this with my retrospective as a CTO building like this. I just updated it so you can choose your price (even $0) to make this a less shameful plug on HN: https://ctoretrospective.gumroad.com/l/own-your-system
I dont think anyone has the perfect answer, yet, but LLM-built systems arent that different from having the system built by 10 diff people on eLance/Upwork/Fiverr...so the principles are the same.
the_af · 28m ago
The Bus Factor was indeed an issue before LLMs, and in fact it's a jargon term that has been in use since forever.
What TFA is arguing is that never before we had a trend towards Bus Factor zero. Before, the worst was often 1 (occasionally zero, of course, but now TFA argues we're aiming for zero whether we're aware or not).
TuringNYC · 16m ago
True, but when the bus factor is 1, it might as well be zero -- soon you end up with employees (or contractors) who legitimately want more compensation realizing their critical nature. I totally sympathize from the employee's perspective, esp if the 1-factor means they cannot take holiday. Really, it is the company's job to control the bus factor (LLM or human) -- it is good for both the employee and company in the long run.
scarface_74 · 19m ago
Is there really a large difference between 0 and 1 when the average tenure of a software developer is 3 years or less at any given company?
TuringNYC · 15m ago
> Is there really a large difference between 0 and 1 when the average tenure of a software developer is 3 years or less at any given company?
Spot on. 1 might as well be zero. Totally unfair to the worker also, who now cannot take time off.
scarface_74 · 10m ago
When I was an architect for startups between 2016-2020 doing mostly green field development using new to the company AWS technologies, I made damn sure that any knowledge was disseminated so I could both take a vacation without being interrupted and I could “put myself out of a job”.
I considered it a success when I realized a company doesn’t need me anymore and I can move on and talk about what I did at my next interview in STAR format.
hobs · 54m ago
Yes, its very much a goldfish problem, where work needed grows to fill what is possible, not what is advisable or good.
The only way I have seen people "solve" this is by putting a bunch of speed bumps in a process, and generally it just makes everyone lazy and deliver stuff at the last second anyway, not use the additional time to make something polished.
TuringNYC · 14m ago
>> The only way I have seen people "solve" this is by putting a bunch of speed bumps in a process
I solve this by sufficient compartmentalization with good inter-component interfaces. Worst case, you excise part of your system and rebuild. Possibly you can take the schema and docs and rebuild with an LLM :-)
I talk about this in my upcoming book on the topic (link above.) Most good systems are rebuilt 3 or 4 times anyway.
pratikshelar871 · 39m ago
I recently joined a team with a very messy codebase. The devs were long gone, and even the ones maintaining it didn’t really understand large parts of the code. The bus factor was effectively zero.
What surprised me was how useful AI was. It helped me not only understand the code but also infer the probable intent behind it, which made debugging much faster. I started generating documentation directly from the code itself.
For me, this was a big win. Code is the source of truth. Developer documentation and even shared knowledge are often full of bias, selective memory, or the “Chinese whispers” problem where the story shifts every time it’s retold and never documented. Code doesn’t lie, it just needs interpretation. Using AI to cut through the noise and let the code explain itself felt like a net positive.
spankalee · 1h ago
There is a really important point here, and it's critical to be aware of it, but we're really just at the beginning of these tools and workflows and these issues can be solved, IMO, possibly better than with humans.
I've been trying to use LLMs to help code more to mixed success, honestly, but it's clear that they're very good at some things, and pretty bad at others. One of the things they good at obviously is producing lots of text, two important others are that they can be very persistent and thorough.
Producing a lot of code can be a liability, but an LLM won't get annoyed at you if you ask it for thorough comments and updates to docs, READMEs, and ADRs. It'll "happily" document what it just did and "why" - to the degree of accuracy that they're able, of course.
So it's conceivable to me at least, that with the right guidance and structure an LLM-generated codebase might be easier to come into cold years later, for both humans and future LLMs, because it could have excellent documentation.
jsjsksksksksk · 1h ago
There is a really important point here, and it's critical to be aware of it, but we're really at the end of these tools and workflows and these issues can't be solved.
siliconc0w · 1h ago
The problem is that our brains really don't like expending calories on information we don't repeatedly use so the further you get from something, the less you understand it or remember it.
So even if you aren't even vibe coding and are trying to review every change, your skills are atrophying. We see this all the time as engineers enter management, they become super competent at the new skills the role requires but quickly become pretty useless at solving technical problems.
It's similar to why it's so hard to go from level2 to level5 in driving automation. We're not really designed to be partially involved in a process - we quickly loose attention, become lazy, and blindly trust the machine. Which is maybe fine if the machine is 100% reliable but we know that isn't the case.
No comments yet
s1mplicissimus · 1h ago
I agree with the premise and the conclusion, but over almost 20 years of writing, adapting and delivering software I've more than once been in exactly the same situation. Noone to ask, the only person even vaguely familiar with software development left half a year ago. Half of the processes have changed since the software was written, and the people who owned them have left, too.
So while I agree that LLMs will accelerate this process, in my opinion it's not a new flavor, just more of an existing problem. Glad to see this kind of thinking though.
trod1234 · 1h ago
The author though neglects what a bus factor of 0 means in real terms and how it gets there, aside from the description of the definition upfront where all knowledge lost.
A company acceptable with a bus factor of zero is a company that is not willing to pay the economic advantage to the expertise required to do the work.
The economic demand, of humanity competing with AI is zero because AI does things its good at with an order of magnitude difference in cost, and the deception and lies surrounding the marketing coupled with communications channel jamming lead to predictable outcomes. What happens when that demand and thus economic benefit go to zero? Any investment in the knowledge in the first place has no return. No one goes into it, no one learns, and that's quite dangerous in economies based in money-printing.
So while there may not be a problem right now, there will no doubt be problems in the next proverbial quarter. Career development pipelines are years in the making. They are sequential pipelines. Zero into a sequential pipeline is zero out with the time-lag between the two (which is years).
~2 years without economic incentive is when you lose your best and brightest. From there is a slow march to a 10 year deadline after which catastrophic losses occur.
I had a chance to have a interesting discussion with a local community college Dean. Apparently they have had to lower the number of Computer Science related program sections because of lack of demand. In intro courses, for 18 sections there were 13 people who had declared for the major, most students when asked citing AI concerns and lack of career development pipeline.
What happens when you have no expertise that you can pay at any price to fix the processes involved because you put them all out of business using what amounts to a semi-thinking non-sentient slave.
Without supply, there can be no demand. Where supply becomes infinite because of external parties, there can be no demand. There can be need, but demand is not need.
So this all started in 2022. Best and brightest are re-skilling. There's a glut of lower competency talent too. Bad things happen when you mess with the foundations of economics, and they happen on a time lag, where you can't react fast enough after-the-fact.
What will it take? At some point there will be a crisis where they will have to treat it as triage on the battlefield. The people in charge didn't want to hear about this when it could have made a difference.
juancn · 2h ago
I think the article underestimates how much intent can be grasped from code alone. Even without comments.
Humans (and I strongly suspect LLMs, since they're statistical synthesis of human production) are fairly predictable.
We tend to tackle the same problems the same way. So how something is solved, tells you a lot about why, who and when it was solved.
Still, it's a valid point that much of the knowledge is now obscured, but that could be said too of a high employee churn organization.
sfink · 1h ago
> I think the article underestimates how much intent can be grasped from code alone. Even without comments.
I agree, the human thought process always ends up getting embedded in a which of several possible ways any one thing might be done. But it's still a process, and a vastly inferior one to having a knowledgeable person on hand. Reverse-engineering has up until now been reserved to times when it is necessary. (I mean, we all still do it, especially on legacy codebases, but it's not good for productivity at all.)
> Humans (and I strongly suspect LLMs, since they're statistical synthesis of human production) are fairly predictable.
I disagree with the parenthetical. That's what stands out to me the most about LLM code: there are definitely intentions embedded in the code, but they're a hodgepodge of different intentions all tangled up together. You can't infer much of anything from a small snippet, because the intention behind that snippet was likely relevant to a completely different origin codebase that was piled up into a compost heap to grow an LLM out of. It's actually harder to make sense of an LLM codebase because the intentions distract and confuse the reader -- just as with text, we implicitly assume that a generated artifact means the same thing as if a human wrote it, and it doesn't.
bigiain · 2h ago
> I think the article underestimates how much intent can be grasped from code alone.
That's very scale related.
I rarely have any trouble reading and understanding Arduino code. But that's got a hard upper limit (at least on the common/original Arduinos) of 32kB of (compiled) code.
It's many weeks or months worth of effort, or possibly impossible, for me to read and understand a platform with a hundred or so interdependent microservices written in several languages. _Perhaps_ there was a very skilled and experienced architect for all of that, who demanded comprehensive API styles and docs? But if all that was vibe coded and then dropped on me to be responsible? I'd just quit.
d_watt · 2h ago
It's potentially the opposite. If you instrument a codebase with documentation and configuration for AI agents to work well in it, then in a year, that agent will be able to do that same work just as well (or better with model progress) at adding new features.
This assumes your adding documentation, tests, instructions, and other scaffolding along the way, of course.
bigiain · 2h ago
I wonder how soon (or if it's already happening) that AI coding tools will behave like early career developers who claim all the existing code written by others is crap and go on to convince management that a ground up rewrite is required.
(And now I'm wondering how soon the standard AI-first response to bug reports will be a complete rewrite by AI using the previous prompts plus the new bug report? Are people already working on CI/CD systems that replace the CI part with whole-project AI rewrites?)
shusaku · 57m ago
I’ve got a new project I’ve been handling with Claude code. Up until now I’ve always pair coded with AIs, so I would know (and usually tweak) every bit of code generated. Now with the agent, it’s easy to miss what’s being made.
Ive been trying to resolve this with things like “make a notebook that walks through this modules functions”, etc, to make it easier for me to review.
In the spirit of literate riding though, why but have these agents spend more time (tokens…money) walking you through what they made.
Likewise, if dev A vibe codes something and leaves it to dev B to maintain, we should think about what AI workflows can get B up to speed fast. “Give me a tour of the code”
interstice · 47m ago
> The only thing you can rely on is on your ability to decipher what a highly imperfect system generated, and maybe ask explanations to that same imperfect system about your code its code (oh, and it has forgotten everything about the initial writing process by then).
This just sounds like every other story I hear about working on ossified code bases as it is. At least AI can ingest large amounts of code quickly, even if as of today it can't be trusted to actually decipher it all.
Rickasaurus · 27m ago
The flaw in this reasoning is AI can also help you understand code much more quickly than we could before. We are now in fractional bus factor territory.
appease7727 · 48m ago
LLMs aren't bad for programming in general.
LLMs are bad for bad programmers. LLMs will make a bad programmer worse and make a layperson think they're a prodigy.
Meanwhile, the truly skilled programmers are using LLMs to great success. You can get a huge amount of value and productivity from an LLM if and only if you have the skill to do it yourself in the first place.
LLMs are not a tool that magically makes anyone a good programmer. Expecting that to be the case is exactly why they don't work for you. You must already be a good programmer to use these tools effectively.
I have no idea what this will do to the rising generation of programmers and engineers. Frankly I'm terrified for them.
schlowmo · 2h ago
At least we also have LLMs to generate our status updates during outages of our SaaS products while groping around in the dark.
bigiain · 1h ago
This is why it's important to be experienced in a range of different SaaS and AI tools.
So that when one of your employers has a SaaS related outage, you can just switch to one of your other employers and keep working.
All hail the 100x AI assisted developers doing 10x jobs at 5 different companies at the same time!
krrishd · 1h ago
Related premise / HN discussion from a bit ago - “AI code is legacy code from day one”:
LLMs make it vastly easier to work on unfamiliar code bases
windows2020 · 1h ago
A team unfamiliar with a code base demoed asking questions to an LLM about it. The answers genuinely excited some. But anyone who had spent a short time in the code base knew the answers were wrong. Oh well.
ants_everywhere · 56m ago
That is one anecdote, but it doesn't really have any information in it. To debug the process we'd need to know which LLM, the developer's backgrounds, what prompts they used etc.
I've used a variety of LLMs to ask questions about probably dozens of unfamiliar code bases many of which are very complicated technically.
At this point I think LLMs are indispensable to understanding unfamiliar code bases. Nearly always much better than documentation and search engines combined.
ardillamorris · 1h ago
We’re already at bus factor of close to zero for most bank code written in cobol lol
danjl · 1h ago
The folks that think they can now suddenly program without any experience, and not need understand how their product works, are suffering from Dunning-Kruger syndrome. Actually, it is a much broader segment and includes product managers, executives, VCs and the general public.
fzeroracer · 2h ago
Unfortunately the corporate machine has been converging on a bus factor of 0. I've been part of multiple teams now where I was the only one holding knowledge over critical subsystems and whenever I attempted to train people on it, it was futile. Mainly because they would get laid off doing 'cost-savings measures'.
There were times where I was close to getting fed up and just quitting during some of the high profile ops I had to deal with which would've left the entire system inoperable for an extended period of time. And frankly from talking to a lot of other engineers, it sounds like a lot of companies operate in this manner.
I fully expect a lot of these issues to come home to roost as AI compounds loss of institutional knowledge and leads to rapid system decay.
bigiain · 1h ago
My guess? The AI companies will keep the free and $20/month plans to entice developers and their managers. They will have $200/month plans with bigger context windows to allow effective work on larger that toy codebases. But sooner or later companies with large scale projects are going to need a much larger context window, that _that_ will suddenly become a $200k/year/developer subscription. There's a lot of correlation between "institutional knowledge" and context window I think.
zzzeek · 1h ago
just not even ten years ago the discussion here was all about "software engineering" trying to be more legitimized as a formal engineering practice, if there should be licensing, if there should be certifications, lots of threads about formal methods to prove algorithms work, and look where we are now. Arguing if humans should even care look at the code we are producing and shipping. crazy shit man
rerdavies · 20s ago
[delayed]
oompydoompy74 · 2h ago
By that same logic, if a project is documented so thoroughly that an agent could handle all the work, then the bus factor effectively becomes infinite.
getnormality · 1h ago
This guy thinks bus factors of zero started with ChatGPT. Hahahahahaha. Adorable.
How many of you have asked about a process and been told that nobody knows how it works because the person who developed it left the company?
There was a blog post at the top of HN about this, like, yesterday.
I hate the current AI hype hothouse and everything it seems to be doing to the industry... but I couldn't help but laugh.
The post is great. Bus factor of zero is a great coinage.
windows2020 · 1h ago
The difference is this isn't some legacy system that still exists a decade later. It's brand new with the tag still on. And it wasn't designed by a conscious being but by probability.
I've seen from beautiful to crazy legacy systems in various domains. But when I encounter something off, there appears to always be a story. Not so much with LLMs.
It got hit by a bus before it was born.
jonny_eh · 2h ago
Everything you interact with on a daily basis is either natural, or designed by a human. Until now.
Where llms excel is in situations like:
* I have <special snowflake pile of existing data structures> that I want to apply <well known algorithm> to - bam, half a days work done in 2 minutes.
* I want to set up test data and the bones of unit tests for <complicated thing with lots of dependencies> - bam, half a days work done in 2 minutes (note I said to use the llms for a starting point - don't generate your actual test cases with it, at least not without very careful review - I've seen a lot of really dumb ai generated unit tests).
* I want a visual web editor for <special snowflake pile of existing data structures> that saves to an sqlite db and has a separate backend api, bam 3 days work done in 2 minutes.
* I want to apply some repetitive change across a large codebase that's just too complicated for a clever regex, bam work you literally would have never bothered to do before done in 2 minutes.
You don't need to solve hard problems to massively increase your productivity with llms, you just need to shave yaks. Even when it's not a time save, it still lets you focus mental effort on interesting problems rather than burning out on endless chores.
You would naively think that, as did I, but I've tested it against several big name models and they are all eventually "lazy", sometimes make unrelated changes, and worse as the context fills up.
On a small toy example they will do it flawlessly, but as you scale up to more and more code that requires repetitive changes the errors compound.
Agentic loops help the situation, but now you aren't getting it done in 2 minutes because you have to review to find out it wasn't done and then tell it to do it again N times until it's done.
Having the LLM write a program to make the changes is much more reliable.
I ended up doing this when switching our 50k-LOC codebase to pnpm workspaces, and it was such a good experience. It still took me a day or two of moulding that script to get it to handle the dozens of edge cases, but it would have taken me far longer to split things up by hand.
I still feel like I am under-using the ability of LLMs to spit out custom scripts to handle one-off use-cases.
And I would never want to use a piece of software written by you ever.
If you think that writing the code was the hard part, your code was probably always shite.
I've done this successfully on multiple projects in the 10-20k LOC, ~100 file area - fully LLM generated w/ tons of my personal feedback - and it works just fine. 4/5 features I implement it gets pretty close to nailing from the spec I provide and the work is largely refactoring. But the times it doesn't get it right, it is a slog that could eat the better part of a day. On the whole though it probably is a 3-5x speedup.
I'm a little less confident about doing this on projects that are much bigger... then breaking things up into modules begins to look more attractive.
> bam, x days work done in 2 minutes
This is a bit of a misrepresentation, since those two minutes don’t account for the reviewing time needed (nor prorperly, which vastly exceeds that time. Otherwise you end up in the situation of “doing it wrong” described in your first paragraph.
If you have an LLM transform a big pile of structs, you plug them into your program and it will either compile or it won't.
All programmers write countless one-off throwaway scripts. I can't tell you how many times I've written scripts to generate boring boilerplate code.
How many hours do you spend reviewing such tools and their output? I'll bet anything it's just about zero.
I've also spent countless hours debugging throwaway scripts I wrote myself and which don't work exactly like I intended when I try them on test data.
My job right now depending on the week is to either lead large projects dealing with code I don’t write or smaller “full stack” POCs- design, cloud infrastructure (IAC), database, backend code and ETL jobs and rarely front end code. Even before LLMs if I had to look at a project I did it took me time to ramp up.
I want to create a new project using framework XYZ. I already know how to do it, but I don't remember how to set up it since I only do that once, or I don't know how to set up a class that inherits from the framework because I usually just copy the other from another class in the same project. I can simply tell the bot to write the starting code and take it from there.
The sad thing is for a LOT of use cases an LLM is completely unnecessary. Like why do I even need an LLM for something like this? Why can't I just download a database of code examples, plug it into a search engine that apppears in the sidebar, and then just type "new project XYZ" or "new class XYZ.foo" to find the necessary snippet? A lot of NPM frameworks have a set up script to get you started with a new project, but after that you are practically forced to use Google.
It's crazy that a problem that could be solved so easily with a local file search has been ignored for so long and the only solution has been something impossibly inefficient for the problem it's supposed to solve.
But in the long term, when experienced developers actually feel comfortable letting LLMs write large swats of code, or when the machine no longer needs to generate human readable code, then we will start forgetting how it works.
[0]: https://news.ycombinator.com/item?id=43909111
- Before LLMs, provided that your team did some of their due diligence, you could always expect to have some help when tackling new code-bases.
Has the author never worked on legacy code before?
- (oh, and it has forgotten everything about the initial writing process by then).
Does the author not think this can ever be solved?
- Because of the situation of a Bus Factor of zero that it creates, vibe coding is fundamentally flawed. That is, only until there is an AI that can generate 100% accurate code 100% of the time, and it is fed 100% accurate prompts.
Why does it require 100% accuracy 100% of the time? Humans are not 100% accurate 100% of the time and we seem to trust them with our code.
I agree with your first point, maybe AI will close some of those gaps with future advances, but I think a large part of the damage will have been done by then.
Regarding the memory of reasoning from LLMs, I think the issue is that even if you can solve it in the future, you already have code for which you've lost the artifacts associated with the original generation. Overall I find there's a lot of talks (especially in the mainstream media) about AI "always learning" when they don't actually learn new anything until a new model is released.
> Why does it require 100% accuracy 100% of the time? Humans are not 100% accurate 100% of the time and we seem to trust them with our code.
Correct, but humans writing code don't lead to a Bus Factor of 0, so it's easier to go back, understand what is wrong and address it.
If the other gaps mentioned above are addressed, then I agree that this also partially goes away.
At least having a partner explain some aspects is a huge unlock for me. Maybe the OP shadowboxing a world with no humans, but no humans is already a status quo issue that my team and I face sometimes.
I've said this before, but I'd say it again: anti-AI people, instead of AI users, are usually those who expect AI to be magical panacea.
The vibe reminds me of some people who are against static typing because "it can't catch logical error anyway."
Shameful plug, i've been writing a book on this with my retrospective as a CTO building like this. I just updated it so you can choose your price (even $0) to make this a less shameful plug on HN: https://ctoretrospective.gumroad.com/l/own-your-system
I dont think anyone has the perfect answer, yet, but LLM-built systems arent that different from having the system built by 10 diff people on eLance/Upwork/Fiverr...so the principles are the same.
What TFA is arguing is that never before we had a trend towards Bus Factor zero. Before, the worst was often 1 (occasionally zero, of course, but now TFA argues we're aiming for zero whether we're aware or not).
Spot on. 1 might as well be zero. Totally unfair to the worker also, who now cannot take time off.
I considered it a success when I realized a company doesn’t need me anymore and I can move on and talk about what I did at my next interview in STAR format.
I solve this by sufficient compartmentalization with good inter-component interfaces. Worst case, you excise part of your system and rebuild. Possibly you can take the schema and docs and rebuild with an LLM :-)
I talk about this in my upcoming book on the topic (link above.) Most good systems are rebuilt 3 or 4 times anyway.
What surprised me was how useful AI was. It helped me not only understand the code but also infer the probable intent behind it, which made debugging much faster. I started generating documentation directly from the code itself.
For me, this was a big win. Code is the source of truth. Developer documentation and even shared knowledge are often full of bias, selective memory, or the “Chinese whispers” problem where the story shifts every time it’s retold and never documented. Code doesn’t lie, it just needs interpretation. Using AI to cut through the noise and let the code explain itself felt like a net positive.
I've been trying to use LLMs to help code more to mixed success, honestly, but it's clear that they're very good at some things, and pretty bad at others. One of the things they good at obviously is producing lots of text, two important others are that they can be very persistent and thorough.
Producing a lot of code can be a liability, but an LLM won't get annoyed at you if you ask it for thorough comments and updates to docs, READMEs, and ADRs. It'll "happily" document what it just did and "why" - to the degree of accuracy that they're able, of course.
So it's conceivable to me at least, that with the right guidance and structure an LLM-generated codebase might be easier to come into cold years later, for both humans and future LLMs, because it could have excellent documentation.
So even if you aren't even vibe coding and are trying to review every change, your skills are atrophying. We see this all the time as engineers enter management, they become super competent at the new skills the role requires but quickly become pretty useless at solving technical problems.
It's similar to why it's so hard to go from level2 to level5 in driving automation. We're not really designed to be partially involved in a process - we quickly loose attention, become lazy, and blindly trust the machine. Which is maybe fine if the machine is 100% reliable but we know that isn't the case.
No comments yet
A company acceptable with a bus factor of zero is a company that is not willing to pay the economic advantage to the expertise required to do the work.
The economic demand, of humanity competing with AI is zero because AI does things its good at with an order of magnitude difference in cost, and the deception and lies surrounding the marketing coupled with communications channel jamming lead to predictable outcomes. What happens when that demand and thus economic benefit go to zero? Any investment in the knowledge in the first place has no return. No one goes into it, no one learns, and that's quite dangerous in economies based in money-printing.
So while there may not be a problem right now, there will no doubt be problems in the next proverbial quarter. Career development pipelines are years in the making. They are sequential pipelines. Zero into a sequential pipeline is zero out with the time-lag between the two (which is years).
~2 years without economic incentive is when you lose your best and brightest. From there is a slow march to a 10 year deadline after which catastrophic losses occur.
I had a chance to have a interesting discussion with a local community college Dean. Apparently they have had to lower the number of Computer Science related program sections because of lack of demand. In intro courses, for 18 sections there were 13 people who had declared for the major, most students when asked citing AI concerns and lack of career development pipeline.
What happens when you have no expertise that you can pay at any price to fix the processes involved because you put them all out of business using what amounts to a semi-thinking non-sentient slave.
Without supply, there can be no demand. Where supply becomes infinite because of external parties, there can be no demand. There can be need, but demand is not need.
So this all started in 2022. Best and brightest are re-skilling. There's a glut of lower competency talent too. Bad things happen when you mess with the foundations of economics, and they happen on a time lag, where you can't react fast enough after-the-fact.
What will it take? At some point there will be a crisis where they will have to treat it as triage on the battlefield. The people in charge didn't want to hear about this when it could have made a difference.
Humans (and I strongly suspect LLMs, since they're statistical synthesis of human production) are fairly predictable.
We tend to tackle the same problems the same way. So how something is solved, tells you a lot about why, who and when it was solved.
Still, it's a valid point that much of the knowledge is now obscured, but that could be said too of a high employee churn organization.
I agree, the human thought process always ends up getting embedded in a which of several possible ways any one thing might be done. But it's still a process, and a vastly inferior one to having a knowledgeable person on hand. Reverse-engineering has up until now been reserved to times when it is necessary. (I mean, we all still do it, especially on legacy codebases, but it's not good for productivity at all.)
> Humans (and I strongly suspect LLMs, since they're statistical synthesis of human production) are fairly predictable.
I disagree with the parenthetical. That's what stands out to me the most about LLM code: there are definitely intentions embedded in the code, but they're a hodgepodge of different intentions all tangled up together. You can't infer much of anything from a small snippet, because the intention behind that snippet was likely relevant to a completely different origin codebase that was piled up into a compost heap to grow an LLM out of. It's actually harder to make sense of an LLM codebase because the intentions distract and confuse the reader -- just as with text, we implicitly assume that a generated artifact means the same thing as if a human wrote it, and it doesn't.
That's very scale related.
I rarely have any trouble reading and understanding Arduino code. But that's got a hard upper limit (at least on the common/original Arduinos) of 32kB of (compiled) code.
It's many weeks or months worth of effort, or possibly impossible, for me to read and understand a platform with a hundred or so interdependent microservices written in several languages. _Perhaps_ there was a very skilled and experienced architect for all of that, who demanded comprehensive API styles and docs? But if all that was vibe coded and then dropped on me to be responsible? I'd just quit.
This assumes your adding documentation, tests, instructions, and other scaffolding along the way, of course.
(And now I'm wondering how soon the standard AI-first response to bug reports will be a complete rewrite by AI using the previous prompts plus the new bug report? Are people already working on CI/CD systems that replace the CI part with whole-project AI rewrites?)
Ive been trying to resolve this with things like “make a notebook that walks through this modules functions”, etc, to make it easier for me to review.
In the spirit of literate riding though, why but have these agents spend more time (tokens…money) walking you through what they made.
Likewise, if dev A vibe codes something and leaves it to dev B to maintain, we should think about what AI workflows can get B up to speed fast. “Give me a tour of the code”
This just sounds like every other story I hear about working on ossified code bases as it is. At least AI can ingest large amounts of code quickly, even if as of today it can't be trusted to actually decipher it all.
LLMs are bad for bad programmers. LLMs will make a bad programmer worse and make a layperson think they're a prodigy.
Meanwhile, the truly skilled programmers are using LLMs to great success. You can get a huge amount of value and productivity from an LLM if and only if you have the skill to do it yourself in the first place.
LLMs are not a tool that magically makes anyone a good programmer. Expecting that to be the case is exactly why they don't work for you. You must already be a good programmer to use these tools effectively.
I have no idea what this will do to the rising generation of programmers and engineers. Frankly I'm terrified for them.
So that when one of your employers has a SaaS related outage, you can just switch to one of your other employers and keep working.
All hail the 100x AI assisted developers doing 10x jobs at 5 different companies at the same time!
https://news.ycombinator.com/item?id=43888225
I've used a variety of LLMs to ask questions about probably dozens of unfamiliar code bases many of which are very complicated technically.
At this point I think LLMs are indispensable to understanding unfamiliar code bases. Nearly always much better than documentation and search engines combined.
There were times where I was close to getting fed up and just quitting during some of the high profile ops I had to deal with which would've left the entire system inoperable for an extended period of time. And frankly from talking to a lot of other engineers, it sounds like a lot of companies operate in this manner.
I fully expect a lot of these issues to come home to roost as AI compounds loss of institutional knowledge and leads to rapid system decay.
How many of you have asked about a process and been told that nobody knows how it works because the person who developed it left the company?
There was a blog post at the top of HN about this, like, yesterday.
I hate the current AI hype hothouse and everything it seems to be doing to the industry... but I couldn't help but laugh.
The post is great. Bus factor of zero is a great coinage.
I've seen from beautiful to crazy legacy systems in various domains. But when I encounter something off, there appears to always be a story. Not so much with LLMs.
It got hit by a bus before it was born.