So far, so agreeable, but…
> If you have thoughts, they come out sharper and faster.
I can’t help but wonder whether constant use of “agent” harnesses will lead to an atrophy of the software engineering (or really any field) muscles.
Actual muscles need exercise to stay in shape (let alone grow), so does the brain. Can we really be sure that thoughts, opinions, taste will still come out sharper and faster after five, ten, 20 years of using these tools almost every day?
Conversely, I also am a user of LLMs (true shocker these days, I know), and am noticing a speedup in areas I was already familiar with, and a quicker introduction to new ones. The obvious benefit cannot be denied, and doing so regardless makes you look uninformed. [0]
So what’s the ideal “middle ground” in this situation? Stoically continuing to sharpen your skills on your own, but risking being left in the dust productivity-wise? Or taking an “agent first” approach and trying to learn and improve more only on the side, as more of an afterthought?
[0] Excluding people who don’t want anything to do with LLMs out of moral principle, which curiously just like the overarching topic I also both respect and understand, but on the other hand don’t do myself.
For sure. You cannot have "only higher level thoughts" without doing lower level work.
Ironically llm themselves prove that because you cannot remove facts like 'paris is capital of france' from llm and have it just retain 'high level thoughts' like 'countries have capitals that you can look up'
What do you mean? I think people routinely think about things at a very high level with almost no understanding of the lower levels. How many people use a computer each day and reason about them at a very high level while knowing nothing of capacitors, logic gates, or programming languages?
Let's consider even the original example. > You cannot remove facts like 'paris is capital of france' from llm and have it just retain 'high level thoughts' like 'countries have capitals that you can look up'
Wouldn't the knowledge that countries have capitals precede the knowledge that Paris is the capital of France?
This says nothing about the accuracy of our own models based on these abstractions that lack the lower-level understanding.
I think the counterargument would be "you can't teach people architecture alone and get good architects".
I've observed this myself in "systems engineers" whose job is to connect boxes together without understanding how the boxes work. They, invariably, design ridiculous architectures on their own and need to basically find a domain expert to route their opinions through to come up with anything sane.
It can be fine as a “user”, but not really as a “designer”. Because discussion about possible solutions is a matter of tradeoffs and tweaking of parameters, not slinging words around. Abstractions are not appliances that are plug and play. They’re often full of parameters that dictate their usefulness and costs, and not understanding those parameters is just roleplaying.
I suspect top-down learners are who benefit massively from LLMs while bottom-up learners think they are bullshit engines.
As a top-down learner myself, Claude is really the best technology of my lifetime. I can completely understand though how a bottom-up learner could see Claude as an afront to their fundamental epistemology.
Because it is.
Spend 3 days a week writing Ruby on Rails and 2 days hand rolling x86 assembly. Every web dev I know has been doing this since long before LLMs. Ensures they can keep having high level Rails thoughts.
My largest concern comes from something tangential to this: I'm not sure we're all that good at deciding what should be learned and sticking to it.
Silly example: regex. LLMs are, as far as I know, well above the average dev when it comes to writing regex. Regex is also one of those things that for many people goes unused for months, but then you encounter the occasional perfect regex problem, and it's really easy to just lean on the LLM to write the regex for you rather than spending some time tinkering and testing. Regex can be frustrating and fickle, I think we've all been there.
But then, you just don't learn regex. So where does the intuition for what regex can do come from? Do you just become unable to write regex with no LLM? People stop writing resources for regex I guess?
My concern is that there's stuff I feel I can just chuck onto the LLM but I'm sure my judgement is not perfect. It's still probably worth it, all in all, but I'm not even sure of what I might be losing along the way and that's an uneasy feel.
I am very glad that I can now just ask claude for a regex to achieve my intent.
Does it mean I'll never master regex? Yes it does, but decades has shown that was unlikely to ever happen anyway.
Sometimes the patterns are dirt-simple, like `/\bfoo/` to find things that start with foo, or a `/foo.?bar/i` to try to find multiple variations of FooBar and fooBar and foo-bar and FOO_BAR. Other contexts outside an IDE include browser dev-tools, exploratory SQL queries on the database, searching centralized logs, and occasionally turning lines of text into spreadsheets via CSV.
None of that guarantees deep expertise of the weirder tricks, but it absolutely keeps the general knowledge "warm."
I’m wondering if by the time you have been explicit enough in your prompt you might as well have looked up what you needed.
The expression is always terse so it’s not as if you are saving much typing.
What does the business benefit from you handwriting regex vs a clanker?
Immersion with the codebase is not about coding and syntax (which is actually very easy for me). It’s about being able to intuit that a particular combination can result in a buggy state and provides the wrong result downstream. You lose that you stay in the dreamland of specs.
With google that mostly went to the wayside. I just needed to know something existed and I could pull together far more information faster than I was spending before to just do quick lookups
I didn’t say the job was about memorization back then. Rather I found that each of these new abstraction layers made it easier to go broad with one’s on hand knowledge instead of deep
Things like regex and CSS are not that hard if you take some time to study them. People don't though. Now I don't have that advantage anymore since AI can do them so well. Or... Maybe that could be an advantage since there will be even less people learning it.
If you don't know how to regex and you let your imaginary LLM friend write one for you, and you foist that upon your colleagues for review, you've just abused their trust. Because they, being ethical, honest engineers, must now do all the work you neglected except they're doing it at review time instead of at the time of writing.
Or maybe they're dishonest and unethical and they'll just maliciously or negligently approve the PR without actually reviewing the regex.
Either way the situation is completely rotten. You can never ethically use an LLM to do something you're not competent to do yourself.
Well, I think most neuropsychologists would agree that the answer is "yes, there will be atrophy" - if you don't use it, you lose it.
> So what’s the ideal “middle ground” in this situation?
I've been thinking a lot about this myself. My current plan is to train myself to get good at recognizing the feeling of "there's potential effort here that I want to outsource to the LLM" and occasionally choosing to not outsource it and do it by hand - especially with personal projects, where there's far less pressure to ship with velocity than work projects - but I'm not settled on this. I'll take any idea!
I see it as something like a personal gradient descent. You're working on a problem, there are solutions down there somewhere, and you can kind of feel the gradient of the tools-and-techniques ground around you. Any way you walk means you're investing time improving some skill or another. So you should go the way that personally feels to you will best get you moving in the direction that you want to go.
For some people it's obvious LLMs are competent coders, getting better, sticking around... and those people should lean into that gradient. For some people what's obvious is nearly the exact opposites of all that, and I'd encourage those people to also follow their gradient/heart/nose down the path of sharpening their personal traditional coding skills. Some people are in a relatively flat area where nothing is obvious, and need to explore and maybe just keep doing their best to hedge with a bit of both.
It will, but I'm not sure the impact of this will be all too great. We suffer from not knowing how to use an abacus because we have a calculator, and people who feel a pull to keep their low-level chops up will do so anyway.
And imagine you can't own a calculator because owning one outright requires too much hardware (or whatever).
Calculators are a simple machine that implement very basic rules.
LLMs are in a different category. The parallels you can draw between LLMs and calculators.. just don't make sense.
The technology doesn't matter, you can compare to say the power loom from the 1700's. I'm comparing the reaction of humans; human's haven't changed that much. They always react the same when they feel threatened and emotionally challenged by a technology.
What matters is the reality of the thing… which is exactly what everyone is discussing.
As they should. We need more conversations like this, early. It could have saved us decades of bad practices with impossible expectations (see: agile).
In the end, the technology will get used where appropriate. And more importantly many of its weaknesses will be overcome and replaced by new challenges.
It's just a bit tiring to hear the same denouncements repeated over and over. Everyone knows them all by heart now. They're not wrong, they're just not helpful or accomplishing anything. The technology marches forward and will develop naturally. If you personally don't want to use it, then don't.
I think what it accomplishes is developing a shared understanding of the pros/cons of the tech, the risks associated, how we should mitigate those risks, where the tech is appropriate vs. not appropriate, etc.
It feels very dismissive to say that these conversations aren't worth having. Or that we're thinking about these things too much.
I get that it's tiring, especially in the HN echo chamber. But the conversations are still worth having IMO.
The debate marches forward and will develop naturally. If you personally don't want to read it, then don't.
To Aristotle and Plato. Where Plato was against writing because people will become forgetful if they don’t train their minds remembering things.
Which, ironically enough, is still a terrible idea for a lot of businesses. But it's trendy, so they go with it.
Systems aren't a single addition. They are compounded operations with sprawling complexity. What happens when you can't reason through the system? What happens when you start asking for the wrong things? What happens when saying "fix it" on loop stops working?
After 5 years, I think the thought profile every power user of the LLMs would be an LLM derived carbon copy of each other.
Prepare the world to get even more boringly uniform
Putting all this in 2nd paragraph so that you can skip it if you think 'coding' is your primary portion of your job.
I suppose I am in a mildly privileged position in a sense that my work is a weird intersection of tech, finance, and comprehension. In other words, I don't code much, but I absolutely benefit from now being able to play with various projects I would otherwise have no business touching without a bigger support team.
I don't want to invoke Accelenrando, but the muscle imagery and analogy fits. I will give an example. I recently decided to pick up Go for a project ( have experience in some other languages, but I will still be starting fresh ). I could have codex build me what I want, but I am purposefully taking it slow so that I can learn the foundation so that I can have a frame of reference ( because I assume it won't be the only go project for me ).
Otoh, most of my one off python scripts I barely even skim anymore. And honestly,that is the part that scares me more.
On the other hand, if you actually care about the output, how do you know it's right, unless you review the script? I mean, if all you care about is plausible-looking output, you could have the LLM produce that, and skip the Python script entirely...
Along with the total absence of long-term data, I think the benefit can be (weakly) denied. Maybe not in the employmemt marketplace, but certainly for myself.
If slop is fine (and sometimes it is), the benefits are undeniable. If the dev was the kind that would have produced slop anyway - again, undeniable boost.
If the quality needs to be high I think it actually can slow you down, though.
The result is a whole bunch of dysfunctional systems unnecessarily dislodging perfectly acceptable processes.
So the mediocre-dev case may be worse than "10x more mediocre code." It's more code that also skews insecure by default, and that cost shows up downstream in review and incidents, not at the PR. Throughput goes up, and so does the liability per line.
I think there are two different claims here:
- developers overestimate productivity gains, which is a solid finding in many of these studies. Skepticism of extremely large productivity gains is warranted and I flatly disbelieve "10x uplift" claims.
- LLMs give no productivity uplift at all, which is much harder to defend. A repeat of the famous METR RCT study did find evidence of improved productivity, and this seems to align with the experience of many experts I trust.
IMO the bigger problem is that ~1.5x individual dev productivity uplift seems to translate into 1.05x uplift across the team. People have been waaaaayyyyy too overconfident about this stuff.
There was an old Onion article about the monetary savings from buying clothes on sale as computed by (a) the women buying the clothes; and (b) their male relatives.
I can context switch between two or three chats, but doing so speeds up my agent use at the cost of making reviews and discovery harder, so it might come out in the wash.
I'm finding this is a skill which I'm slowwwwwwly improving but, for now, my level means I tend to miss mistakes I'd notice if just working on one task. I also had to make accommodations to the way I work to make it achievable, including telling my employer that a 500 GiB disk just isn't enough any more when I need multiple builds in parallel.
AI definitely supercharges the side project myth, but I think we’re going to see waves of developers quitting or losing their jobs because they’re chasing dreams of running a side code project that isn’t going to turn into a business.
The incentive to be more productive at work is that they get to keep their job and not be replaced by a cheaper junior. I think we’re in for a reckoning across the industry as companies realize that there’s little difference between a lazy senior armed with Claude and a halfway motivated junior who has aspirations of growing into something more. The latter costs less and might grow into a better dev.
The productivity uplift measured in some of these reports is from more tasks being done, not from developers giving themselves more idle time during the day.
If we do see job losses from AI, my feeling is that it’s going to be concentrated in people with an attitude like this one where workers try to remain anchored to their old delivery pace. It’s becoming easier than ever to replace a worker in that position with a lower paid junior equipped with Claude who has an interest in learning. There’s no reason to keep a higher paid worker around to copy and paste between Jira and Claude.
I just compressed what would take a team a few months to deliver in one or two months, the amount of gotcha I spot from the existing system we work on everyday is staggering.
How to deliver and keep these feedback is difficult, the other teams are overwhelmed by other tasks, so any tickets or communications will be forgotten in a few months.
And as an external ressources allowed to use frontier model, I have a hard time to keep up with my own pace, and all that institutional knowledge will be lost both by my employer and client.
- chat box interaction
- vendor harness user (the cursor / antigravity crowd)
- SDD Claude users
- bespoke harness creators who let it goooooooo
I mean we already know who’s using all the tokens and driving enough cost to make penny pinchers consider removing other “costs”
It’s not really hard to defend. Because when people says that productivity is uplifted, they are talking about amount of work, not the ROI. That’s why you keep hearing about LOC, amount of PR and prototypes, and the time taken is actually “time to PR” and not “time to production + time spent on bugs”.
This is how LLMs can both result in greater productivity yet still not appear to yield much more benefit than the pre-LLM era.
And there is no benefit to the developer for doing the work first then just sitting idle. That’s how you get people putting more things on your plate.
My gain is that I produce good work in a normal workday and I am absolutely not writing code to 4 am anymore...
Overall for me it is a win for me and my company.
Managers will abuse you if you expand your workday to make sure deadlines don't slip. They will see that the work is getting done and see no urgency to hire more help.
You have to surface capacity issues to management, learn to say "no" diplomatically, and/or "I have capacity for 2 things out of these 5, prioritize which ones you want first."
If I run out of tokens I’m encouraged to spend money to buy more tokens, Claude has an overage setup, and Anthropic runs occasional token sales.
So, to that example: the more often customers are hitting token limits under duress the more likely to open up the wallet and pay to finish.
Run out of your fix? (Tap tap) Buy some more!
Also importantly, can it be a neovim? Neovim hasn't had literal trillions invested over it in the space of just 5 years, local llm's aside can those smaller productivity gains justify the huge investment put into them, and which will continue to be needed for further model development?
A plane literally goes 10x faster than a car, so a 5 hour drive becomes a 30 minutes flight. But if you have to drive to the airport, arrive early, pass check-in, security, boarding, then pick up your luggage on arrival, rent a car and drive to your destination, you may realize that your 30 minute flight took you more than 5 hours in total.
You also have to consider the Amdahl's law: a 10x speedup on 10% of the project is just a 9% speedup overall.
This is totally true for me. I'm working on more features/bugs at work and getting the coding done faster with claude but the coding part is at maximum half of the process and usually less. The rest of it is clarifying the requirements from our product team or given that we're a very legacy place, what the existing state of our infrastructure and other systems are for the new changes to fit in to, etc.
It's a 10% speedup, unless you think that a 100% speedup would reduce completion time to zero rather than cutting it in half.
Assuming that "10% of the project" is measured by time requirements, when you get a 10x speedup on that part of the project, you'll end up completing 100 units of work in 91 (formerly 100) units of time. Your work rate is then 100 / 91 = 1.099 times what it was before; that's an improvement of 10%, not 9%.
(Sanity check: suppose you have a project that will take 100 days. That project is reassigned to someone 10% faster than you. They will take 100 / 1.1 = 90.91 days to finish.
If they were 9% faster, they'd take 100 / 1.09 = 91.74 days to finish.
Does a savings of 9 out of 100 days of work look more like the project that saves 9.09 days, or the one that saves 8.26 days?)
But instead of productivity, I'm much more interested in using it to improve quality. You've got a tireless reviewer who is always ready to review your code and catch any gaps.
The latest was that we need to write in object pooling in their client code, an option that doesn't exist yet in _their_ client code. Like, yo, that is a you-problem on performance bottlenecks; we expect you to provide solutions.
When a person becomes a manager, they do or do not have enough time and expertise to review all of the code that they trust the team to produce.
Managers usually get into automated testing; unit tests, integration tests, acceptance tests, and maybe also BDD syntax
Managers and developers are responsible for setting a test coverage threshold for merge approval.
If there is 100% branch coverage test coverage for a codebase, what would coverage-guided fuzzing or property testing find? If there is 100% branch coverage test coverage for a codebase, what is the value of spending resources on formal verification?
How does the value of LLM-produced 100% branch coverage compare to no-LLM 100% branch coverage?
This is such a salient question. Sometimes (definitely not always) the test suites produced by LLMs are so trivial it's scary. Coverage can be an illusion for sure.
I wrote a tool called tert - I guess it's called an agent harness now - to run various test runners and log test output and coverage output to disk. FWIU stripping spaces from JSON does save tokens. It seems like feeding coverage lines-missing maps into the prompt results in better output, better LLM-authored tests.
"Refactor these tests for maintainability and coverage. Use fixtures, mocks, and parametrization"
Substance coverage - testing the actual logic, edge cases, etc. Not mere lines.
If there is prompt insufficiency, there is probably acceptance test insufficiency.
A more assuming agent could automatically develop a plan that includes presumptive acceptance tests and request feedback before spending tokens
And, if/where we need tests, we write the source so they are few, high value, and complementary. Like actual unit tests, not complex with stuff like mocks just to generate trivial coverage.
Working with an LLM has given me a real eye opener on unwritten requirements. It's like outsourcing. "Yes, you've given me what I wrote down, but I never expected you do to it in that way"
I haven't yet made myself learn the new swarm of concurrent agents with different specializations/agent_instructions methods yet.
Are multiple worktrees worth the cognitive burden and merge overhead?
A merge maintainer is always in code review mode
> Managers usually get into automated testing; unit tests, integration tests, acceptance tests, and maybe also BDD syntax
I can see managers getting involved into acceptance tests, but never in the other type of tests. And the verification mostly is involved into a quick manual testing/watching a demo. Code is not their concern. When there's a bug, they expect you to investigate and fix it.
And then sometimes you report a memory leak and it gets fixed by a VP and you wonder if he doesn't have something better to do.
If devops has done their job, it should be trivial for a manager to contribute to the tests and run the build on git push (or manually re-run the build with the web UI).
If a manager has deploy rights, they should be able to run the tests.
If you have a "I trust my competent team to write good enough tests and test coverage isn't my responsibility" attitude, that's what quality software you'll get back.
There are people producing good and excellent quality software with LLMs. Presumably you must discard low-quality code in order to maintain quality.
There's certainly a limit to code quality with current models. On number of lines of code per unit of time, LLM tools certainly already win.
Can costly automated code review for PRs catch most of the problems before they're under consideration for merge?
For example, the vscode repo has extensive copilot integration. Every PR gets auto code reviewed. But with their tokens or the contributors'?
If I take poor quality code (AI-assisted or not) and spend a few hundred dollars on tokens for a next gen model and agent to get to 100% coverage and review for security bugs and CWE common weaknesses, what quality code will I have without refactoring with proven patterns and type annotations and polishing docstrings?
If you're a manager now and your "team" is a bunch of coding agents, those agents are hardly junior engineers at best. It is equivalent or even worse than hiring a team of 3-5 junior engineers and letting them run rampant with your code.
As much as this sentiment is nice, it is completely divorced from reality, unless the competence is verifiably there. If you take a bunch of juniors and say "yeah I trust them to do everything well enough," you're going to have a disaster on your hands.
I use agents to code. But I remember the early days of just AI smart complete in the IDE, where as the programmer I had to be more involved with designing and implementating the solution. This kept me engaged with the implementation as it was being built out. Now with agents, I find myself trying to catch up with what the agent did and spend more time code reviewing. Maybe you end up in the same place in the end. But building the implementation, vs code reviewing, feels more rewarding and I think helps keep your mental tool sharpened.
Setting aside my moral outrage over the magic token machines. What about me, who gets so tripped up over minor factual errors, that I'm unable to let them go, and it taints the whole conversation such that I'm too wrapped up in my frustration that I can't think about it clearly? Or my innate drive for correctness that's so strong that I eval the minor errors in output, as catastrophically incompatible with my goals?
> Stoically continuing to sharpen your skills on your own, but risking being left in the dust productivity-wise?
I don't believe there's a meaningful productivity increase. Please cite your published (not preprint) peer-reviewed research that proves the productivity improvement. Until then, I'm unconvinced. (Believe me I'd like to be convinced of reality, the answer is still unresolved, and I have my opinions, but I'd rather something conclusive that I can have confidence in)
Then, even if you did show a significant productivity improvement, it wouldn't help me. I have too many qualms over the output quality that I simple can not let go, (I don't think I should, but everyone keeps trying to convince me to lower my quality standards). I don't want something fast, I have plenty of really "fast" things in my life. I exclusively want to add things that are high quality to my life. Things that don't endlessly frustrate me.
The question about where the middle ground is a rhetorically dishonest question. You'd first have to prove/convince me, that there IS a middle ground. Instead of what I believe where that middle ground belongs is quality, and everything emitted by an LLM moves reality in the wrong direction.
Are any of these absolutes? nah, hence my request/demand for peer-review research. All the productivity claims and quality assertions (mine included) are still *exclusively* vibes. But exactly none of them are pristine, (especially not any of the LLM output.)
While I can see that you feel very passionately about this, the reality is that it's the majority experience that will dictate adoption.
There may never be published research on productivity -- blinding in this instance is impossible, so I don't know how you'd ever do fully-controlled behavioural studies that carried any weight. It doesn't matter. If enough of us decide that LLMs are useful to us, then this form of coding will become the norm.
If that ends up causing more harm than good, then eventually there will be a course correction. But for now, for enough people that matter, LLMs are at least giving the perception of productivity increases. And our decisions and choices come down to the perception of reality, not reality itself (for better or worse).
So I think it's far more useful to take a pragmatic approach, as per TFA. Accept that LLMs have issues, but also bring advantages, and that LLM use in coding is here to stay. What we can do is remain aware of the bad, and make better use of the good.
As for you, personally ... if you mentally cannot deal with LLM output, then I think you have two choices. You can either learn to author system prompts, so that LLM output better fits your needs and no longer triggers you; or you can sit more and more on the outer, raging against the machine while the world changes around you.
Eventually, you'll be like a master craftsman in an era of mass-production. But that's potentially highly valuable in niche markets (consider a watchmaker working in Glashutte, for example), so you may yet win from this. Remember that every day, LLMs are making your own coding skills and knowledge more elite and therefore lucrative, sit back, and smile.
Either all of them are wrong, or you will be left behind. Whatever the outcome, it will be on you, and you may make your peace with that.
I really appreciate your comment, I think I've discovered part of why I've never felt any anxiety over being replaced by an LLM. I've never tried or wanted to emit low quality code at high velocity, my standards and goals have always listed quality first. Feels more like asking a chef if they're afraid of llms. Why would they be, they're not doing the same thing.
I'm with you on LLMs being in a different league than good developers curating a fine design with attention to detail, soundness and maintainability. It's just that 1- not all software/components within require that same amount of care: many capabilities are already commodified (as library code or boilerplate, or frameworks, or prior art in the training data that LLMs could easily parrot), and 2- till now, the financial calculus was leaning strongly towards "new developments are expensive, let's use restraint" (irrespective of the actual effort/ratio of commodified code).
So long story short, I don't see LLMs taking out skilled and caring developers: I'll still be reaching out to them for the critical/strategic projects, but on the other hand, I appreciate that the "ancillary" or "fancy" stuff can be cobbled together with the LLM glue gun and remove the impression to my management that engaging with software development is necessarily a costly/risky/exposing endeavour. Some software is to be treated like pets and cherished, and the new reality is to accept that we might have cattle software too, with swaths of prototypes, one-off's, very specific and niche and disposable, and it's fine because this no longer represents an unsustainable financial burden (my bean counting management already likes to kill useful software on a pure spec-sheet matching exercise and shovel everything into large "do it all" systems like ERPs: the fewer things the cheaper, right? Except for those underrepresented users thrown under the bus and their effective productivity).
I don't know if I've ever seen a study that attempted to demonstrate productivity that was ironclad to the point that nobody could find any methodology holes to pick in it.
I think that's a far better and more enjoyable way of working than letting the LLM write it and letting users review it. Partially because writing is more enjoyable than reviewing, and I think people are inherently bad at reviewing while LLMs with their infinite patience are much better at it. But also because the writing itself shapes our thinking and we learn much more from it; it makes us better in ways reviewing the LLM's work does not.
That said, I do let AI write a lot for me, because I'm tackling issues that would otherwise be way too big for a single person. Is this a good idea? Maybe not in general, but iny particular context I think it is.
I'm not sure to what extent I'm fooling myself about that last bit.
Compiler evolutions really harmed how well software engineers understand or how often they have to drop down to assembly language.
Is that a problem for the 99% of developers around? Probably not.
I view LLMs as the next evolution. Some people will still need to care about the layer below, the shape of the code that is being written. But over time, just as it was with the transition from crafted ASM to higher level languages, the compilers became better, more efficient and trustworthy and I think the same will happen to LLMs, and we probably won't have to check the generated code as much, at least for most of the code around.
Is that a problem? Yes, for code that is intended to interface with humans (most of it still). The quality will probably become better and it won't be much of an issue.
Only problem is that, like a LLM, you don't retain anything.
Hobby projects may become a lot more important now because if you do them without LLMs you may retain a brain cell or two.
This will and is 100% happening. I have a friend who hasn't written code by hand in around a year, but uses LLMs every day, and he tells me he can't remember how to write code by hand anymore. He has been a developer for 10 years. But he's not working for anyone at the moment, so I imagine if he was in a workplace the circumstances would be different and they probably wouldn't settle for this.
I think that as a result of this it likely also atrophies the problem solving and architecture building skills that writing the code manually gives you. It just ends up degrading into a loop of tell the agent to do X and assume it knows what it is doing.
What's the risk here? Left behind by who or what?
> Or taking an “agent first” approach and trying to learn and improve more only on the side, as more of an afterthought?
This reads like anxiety resulting from FOMO.
Here's my take: I don't care about LLMs or AI in the sense that I don't feel any need or want to use them. I've only ever tinkered with the free ChatGPT. Never opened an account with any LLM vendor and never even considered it. I program by hand for the joy of it and sometimes for work. Still by hand as I have been doing. MY work gives me that luxury. For now.
Am I obsolete? Am I no longer of any value to society? Of course not. That thinking is just implanted by a group of money hungry individuals who don't give a fuck about me, you or society as a whole. So why would or should I care about LLMs?
The related team investigated, found nothing; They escalated to other teams, found nothing; Gemini, claude and whatnot were involved, found nothing;
Somebody raised to topic to me, solution has been found
See my point ? Like GP, I do not use AI. If there is only one person thinking left, being that person does not looks like a bad idea;
Mixing code and "productivity" is trying to remake coder as bluecollars, and most of them probably are (not thinking, replaceable cogs that are overpaid and shall be decommissionned), they will not be missing. The only coder's output is its idea, everything else is a mean to that end.
That doesn't seem to be the case, at least currently. The credible numbers from from companies that are using LLMs extensively are in the 15-30% productivity improvements for some developers. Those figures don't factor in how that impacts end-to-end SDLC productivity and many of the published results ignore quality which impacts maintenance costs.
Before someone hops in with "Future models will do better", that's also by no means guaranteed. I think the poster you're replying too is safe for a while.
Get some hobby projects.
I would still travel much less distance.
And just like cars, LLMs will reshape the world to the point that our brains could not even get us to the supermarket because soon it will be 5 miles away and require a car ( or at least a local LLM bike )
The point being that cars also reshape your perception of the world. METR had trouble replicating their dev productivity study because devs have already become much whinier about doing things "manually." Things that were slightly tedious in 2022 have become impossibly difficult without AI in 2026.
It is striking to reread Lord of the Rings: so much of the story is spent walking in and observing the Middle-Earth wilderness, and there is a distinct "pantheism" absent from the movies. JRR Tolkien, being one of the first modern fantasy writers, was one of the last who remembered a life without cars.
Used to think so, but they actually can also be used to train and strengthen skills, and learn new ones.
I had a coding interview, where they kindly sent a brief beforehand to help prepare, presenting a list of topics and concepts that might be useful during the interview, the tech stack, what kind of expectations they would have, and what they’d be paying attention to.
Obviously, it’s not exact list, and there are probably other evaluation dimensions.
But since I was out of practice on some of those, I had Claude generate a dozen sample projects, with each a list of tasks in one document and the solutions in another, and got to it.
Midway, I thought of using codex to role play as an interviewer, to tell it my train of thought and ideas as I went, get feedback, question my choices, etc.
Sure I only went through two and half, maybe three of those projects… but it’s the first time I actually enjoyed prepping for an interview. And I actually learned some things in the process.
Hardest part was probably stopping the LLM from doing the tasks, but nothing unsolvable given a bit more time, clearer instructions, and separation.
>> You would probably be surprised how few actually have the agency to do even this small step.
I speculate the gains must be heavily context dependent (no pun intended). Different teams, tech stacks, product types, and company policies perhaps have an outsized effect on what gains LLMs give. We are allowed to be pretty independent and don't have much of a restriction on token usage.
Love to see worker unity in action.
A spec is NOT CODE.
Someone still needs to understand what the code means.
C optimizers are not the same as the radical departure of an LLM.
This is not true, though. Technically I don't do C but I used to do Assembly. Yet in Java code paths, to have a decent proof (and satisfactory results) of the optimizations made by JIT, PrintAssembly still works.
Assembly is not just the raw instructions that trigger microcode or knowing the latency of each one of them, or the (un)expected cache misses, it's the understanding the hardware, itself.
https://arxiv.org/pdf/2604.04721
For what it is worth, my colleagues who use these tools more liberally see a much higher regression rate and are left refactoring more as well. While they definitely put out more PRs, I can confidently say that mine are more valuable and better stand the test of time. Whether that matters or not is up to debate as I guess you could argue there may be some point where the productivity overcomes their regression rate.
My position is, if the technology will be so good and progressing so fast to the point that I don't have to write code anymore and I'm just dictating our needs and what I want architecturally, then I don't really feel the need to put in the time to learn the tools as they are now. That seems way easier to learn than anything computer science related that I had to learn to build robust systems.
E.g.
> A 1.6GB Docker image for a Node app. Every code change rebuilds npm install from scratch. CI takes 8 minutes for a one-line fix.
> The team's proposal: add more CI runners.
Would it be surprising if that team makes a React app which does 10x more requests than necessary, has 10x more JS code than necessary, flickers, etc. People just don't care, and it has nothing to do with AI. "Engineering" is too big of a word for this kind of activity.
On the other hand, I think AI can amplify capabilities of those people who still care, and it's a good thing.
It occurred to me recently how because the LLMs have to potential to lead to intellectual/skill atrophy, they are a factor which could lead to an Idiocracy-esque [0] situation.
And while it's a well-established Fermi Paradox [1] hypothesis that AI could lead to the violent destruction of a civilization, I think it's also worth considering that they could non-violently lead to their decline due to intellectual atrophy.
But then there are definitely cases where LLMs are slower and less reliable than someone using tooling by hand, especially like deterministic refactoring operations can be done close to instantly if you already know what needs to be done and how to do it.
And it's not like we can not use LLM because others are using it in a craze to get the next promotion so you have to use it and maximize your output even if your skills are atrophing
- job displacement
- ethics
- environmental
- skill atrophy
The impact is that the internet has a fraction of the value to improve people's lives as it should have. It is a very poor free market, incredibly poor competition because of lack of standards and protocols and interoperability. People's minds are ground down by social media, search engines don't work well any more and so on.
So yes - every new technology deserves many criticisms, so they can be addressed, and as a society we can gain the benefits of that technology and minimise the disadvantages.
The printing press lead to copyright, public libraries, universal literacy... All things which are now widely celebrated. They took centuries to work out, and are all government and regulatory intervention to fix problems critics noticed and campaigned about.
AI is the same, only it is at risk of moving much faster and having a much large negative impact before society reacts.
So no, most of the criticism of LLMs are not wrong - they are correct, as are the people saying the technology of LLMs is useful to people and the economy.
Critics are friends of a new technology - without responding to every criticism in a significant way, AI will rapidly lead to a Butlerian jihad. If you like AI, you should love criticism of AI even more.
Job displacement: When the Internet first arrived in a significant number of non-technical lives, there were concerns about research jobs and library jobs. In fact it increased the need for that sort of thing so increased that sort of job, and even if it hadn't the access to information for the average person would have been worth that cost. As remote work grew initially, then massively in 2020, there were major concerns about that sort of thing again (people working in Manchester on London wages, going into the office very little, it being even easier to outsource jobs to countries with cheaper workforces, etc.). These concerns are still being debated today, and that is quite valid.
Ethics: Hell yes. And again, it is good that we are still refining what we think is acceptable to this very day. There have been ethical concerns almost from the start of the Internet in the public mind, and they have only increased over time as new parts of it take shape.
Environmental: That is a lot less relevant, as the Internet itself didn't directly have the sort of impact that LLMs are having. There have been local concerns though, over the building of data-centers and the power infrastructure to support them, disturbing land to lay cables, and so forth, and again these concerns are still relevant (though have somewhat folded into just being a concern about LLMs & related ATM as they are the main reason for the current rapid growth).
Skill Atrophy: The Internet never really threatened that, except maybe a small amount of concern wrt certain research roles, it increased the base level of those skills in the general population much more than it had any negative effects (but people did raise the concerns). It gave better access to training and knowledge, but wasn't actually doing things for us in the same way LLMs and the agent systems linked to them are being used, you still needed to learn from that information/training in order to complete your task.
kinda shih people say after their Load Bearing Claims (thank you Opus 4.6) turned out wrong
Can you clarify what this is supposed to mean?
At least that's what I thought of when I read that; because answering in place of other people is another thing I find annoying. So I'm speaking for myself, but with some optimism I may have guessed the intent of OP correctly :)
Master craftsmen didn't take on apprentices to give them chores.
It will be interesting how to get the financial incentives to train juniors in knowledge work to line up with the fact that every new generation needs to be trained.
Is today opposite day?
I don't want to be too flippant here, but senior engineers generally gain very little from training juniors, not least because promo processes suck at basically every tech firm, and so the moment you train up a junior, they are going to leap to another company rather than go through promo...
Yeah, that's the thing for me. LLMs have made my work easier and faster, and they've made my side projects easier and faster. I think there are very sensible and valid critiques but so far the tool works for me.
Saying agents produce shitty code is a bad argument though. They produce shitty codebase organization, but at a micro level their code is solid if not elegant. If you let them turn your codebase into a spaghetti mess, that's on you.
Ding ding ding. This is my biggest gripe with AI. Even the SEO blogspam, the fluff in front of every recipe, yarnwork or DIY instruction, it all was clearly written by a human. Someone had invested time (and money) in getting something in front of my eyes.
But now, it's all just slop. Everywhere. And hell I'm tired because the onslaught breaks my trust filters.
Maybe I think this is an age thing. Boomers? They trust everything written down somewhere. No matter what, and no matter if they didn't spend half my childhood to "never trust what people write on the Internet", and now they fall for scams left and right. My generation as said grew up with this "never trust, always verify" thing. And the younger generation? They DGAF about anything any more, all they care about is trying to survive.
> And b), the teaching, aka “How do we teach new people?”: previously, there was this balance aka “the junior does some pretty mundane tasks, but for this the senior reviews it together with him and helps him to grow”.
GOD YES YES YES THIS x1000.
There is barely anything more rewarding than teaching someone something, to watch the other person grow - and eventually surpassing your own abilities. That is when you know you did right and well. My wife is the best example, she started out at "can you help me with Excel", and these days, she pulls off stuff that would make more than a few finance people blush.
I think many junior devs (or aspiring junior devs) look for exactly this experience. This is a matching problem we haven't solved yet. Is Open Source the solution ? I really think it has to be solved if we want truely reliable software in the future.
The entire economy is broken due to the focus on short term quarterly result instead of the health of a company in 5, 10 or even 20 years.
It's really not incongruent to use LLMs and be in awe of their frankly incredible capabilities while at the same time recognize the risks and frankly real damage we are already seeing to junior training and hiring, open source communities and (in my opinion) very soon the entire fabric of our society.
I respect that people don't want to use agents themselves for whatever personal reason.
I respect maintainers not accepting AI-authored contributions. It's a tradeoff between progress, growing new contributors and maintainer sanity. Though I do feel that categoric opposition to anything AI will likely be futile in the mid-term.
I respect people pushing for regulation of AI or a global pause or whatever.
I don't particularly respect people dismissing everything AI authored as slop. Categorically refusing to read an article because it contains em-dashes or the term "load-bearing" is silly. While this is slowly changing now, many people are still in complete denial as to what the frontier AI is capable of.
Love it, hate it - I don't care, but at least respect it, goddamit.
Even this article has some cognitive dissonance in it. What it really comes down to is how much you trust your own verification process. The branches of questions an LLM generates are still trapped within the biases of its training data. Of course, the authority to craft that initial prompt, the very first question, comes from human experience and learning.
But I think thought itself is the easiest resource to outsource. People say the human did the thinking and the LLM just amplified it, but the truth is, the LLM outsources the thinking. Otherwise, when the result is good, people say "human thought was present," and when it's bad, they say "human thought was absent." But a part of the actual thinking really is outsourced. The alternatives, the counterexamples, the sentence structure. In programming terms, the reader's experience gets outsourced. When you write a blog post, you find yourself thinking about how to make something you understand easy for someone else to understand. With an LLM, that part gets outsourced.
But at the same time, I don't get the argument that you shouldn't use it at all. We don't "think" about everything. We have limited cognitive resources. So we study deeply the things we care about, but for the things we don't need, we mostly leave them to "common sense" or prejudice. We just skim the surface.
I think of "common sense" as "the largest collection of prejudice." Because what we call common sense usually just amounts to surface level knowledge, the kind of thing we know just enough about to get by.
That's why I think LLMs are good. The reason is simple. I don't think deeply about everything in the world anyway. For everything else, I'm buried in some kind of bias. You see it on HN all the time, right? People fight over some technology, but they often don't think about its internal structure or why it works the way it does. They just treat it as an identity. They fight over a particular language, a framework, an operating system, but they rarely check how that technology actually works internally or why it was designed that way. Why use MVC, why a different architecture might be better for my case, it's easier to just go with what's popular. Put more elegantly, "job mobility" gets bundled in there too. I use Windows. In my country, if it's not Windows, you literally can't do anything. You can't even do basic online banking. From regional context like that all the way down to personal interests, people are bound to be different. So I'm just going to use LLMs. The most common excuse you hear around this is the whole "reinventing the wheel" thing.
So yeah, I'm going to use LLMs. Because I recognize that I bias myself toward only thinking about what I want to think about. And I know that bias isn't cognitively healthy. But on the flip side, I think what the world values, whether it's knowing a lot or knowing one thing deeply, is going to change.
Honestly, I don't know what's right. I think both the advocates and the critics are making valid points. I respect the people who don't use it, and the people who do just have their own workflow. There's really no reason to fight over whose workflow is superior.
The tech world does not care about woke ideology, german technical illiteracy and self importance.
LLMs are useful and here to stay.
It might, in fact, be that you are more successful with your approach. I have no idea. Congrats if so.
But I think it's odd for you to now say it doesn't matter whether or not his AI token use is leading to success or not. I personally think it's actually really important to learn whether/when this approach is smart.
But you could use tons of autonomous agents for all kinds of things. Probably shouldn't, but you can.
Im spending about 2-3K a week on personal projects and 5K a week on corporate stuff.
It's nice to see a person who actually acknowledges tradeoff.
For the curious, this seems to be his company: https://www.umh.app
It's not because LLMs are that efficient, it's because humans are that inefficient.
I'm saying it in jest, but it's also a bit true. Not necessarily because we use it any differently. But because my use of AI saves me time. But their use of AI adds more to my plate, no matter if it's slop or not.
I have coworkers who routinely post PRs with comments stating "all tests green, ready to review" and then I actually start the application, test it manually and it doesn't work. And don't get me started on reviewers who post the raw AI output of their "reviewing skill" and then I have to read through 30 paragraphs of "findings" most of which are irrelevant (just today I got the "valuable" feedback of "View is @api_view(["POST"]); no test asserts GET/PUT/PATCH/DELETE → 405")...
The big thing people used to call AI was that it was a stochastic parrot and all it did was summarize things. Clearly. None of this is/was true anymore. And very likely all the current criticism will be eliminated soon and we have to find new excuses about AI that makes us feel we are superior.
The status quo is about to change. Every 6 months. And you will always think of yourself as superior to LLMs. Your current criticisms will evolve as most of them will be rendered not true pretty soon.
Yeah, of course AI code must always be reviewed. All code must be reviewed.
If you review all code, your output will be slow. If you review less code, your output will be faster at the cost of more bugs in production. Bug rate will never go down to zero whether you use AI or not.
That is the trade off, if you review everything then output is really slow. If you start only reviewing certain types of code like model changes, database changes. Or only backend code and not frontend code, you hit a sweet spot of speed and reliability.
As LLMs improve the need for reviews becomes less and less. Companies who don't adjust are just slowing themselves down. That is the trend.
But code that never gets looked at and only tested? Yes. And as AI gets better this will be happening in all companies... more and more.
def plus(a, b):
return a + b
Testing is a statistical sample. You sample a small subset out of the full domain of possible tests and that sample you hope is somewhat representative of the actual population of tests. To put it in a sentence: If my suite of tests is correct then it is more likely that the entire full universe of every single possible test for my program can all be correct.
Realistically though the universe of all possible tests is so large that it's basically impossible for any sample to be a good representation of it. A suite of test does help up confidence that things will be MORE correct, but the sample size is so small it is almost guaranteed that you missed something.
> "Yet I still write all of my texts with LLMs"
So I'm guessing the author is actually ok with the point they put in the "LLMs are bad" part of the article?
A smartphone is also a genuinely good all-around tool. Even social media is a genuinely good tool for connecting people.
Yet, I feel like we've been overly optimistic about the impact of said tools on us and our societies in the past two decades.
Smartphones are so good, in fact, in some societies, half of us are addicted to them. Billions of people world-wide.
I ask myself: Will LLMs enrich my thinking in the long run, or will they ruin it?
And what about most people? Will half of us outsource most of our thinking in a decade from now?
Given the speed and global scale that we're running these experiments with, it's fair, I think, to be a bit sceptical of the conclusion that, in the long run, LLMs will enrich our thinking.
Example, cars are good. Betting the farm on cars to the detriment of bikes, buses, and trains is clearly bad. The tool of an llm is clearly handy. Betting trillions of dollars and linking the future of the nation and globe to this tool is clearly bad.
If you want to argue that LLMs are 100x more useful than a functioning railroad system I've a bridge for you to subscribe to.
LLMs are just continuing the trend humanity has long been traveling down.
You're using broadly shared but totally unsubstantiated about past humans just to argue that things are good right now.
In stone age, you had to focus on other things than thinking to have enough food for the next day. And to survive otherwise.
This is a funny twist on the glorification of the past.
Modern humans have been getting smarter even over the relatively short time periods that we’ve been measuring.
It’s called the Flynn effect: https://en.wikipedia.org/wiki/Flynn_effect
Modern day humans still can and do survive in the wild if they so desire. It’s not some lost intelligence that was only achievable by past ancient humans.
Flynn effect research says otherwise.
If you think prehistoric humans were geniuses because They could survive in the wild, by that measure monkeys and gorillas must be geniuses too.
Flynn effect is just a trend for the past century or so and it's an international measure. For the past decade there's been a trend for the US exclusively that is not aligned with the flynn effect.
https://news.northwestern.edu/stories/2023/03/americans-iq-s...
I use AI to code tools for myself, but I don't pretend anything I make is production quality. Duct tape engineering has always been a bit sloppy, and AI just made it faster.
I use AI to troubleshoot issues and plan out strategies, but I basically consider the AI draft of anything to be "draft 0", and use it as a framework for writing my own works for a real first draft of anything I write that will be read by other people. Sometimes the AI spits out a perfect paragraph that I might copy, but I don't ever blindly trust it or let it speak for me. I also double-check everything it says that I don't have existing knowledge of, rather than trust it to be right.
AI images, video, and music are all entertaining, but I only generate these things as a form of self-entertainment and maybe online meming. I could never in good conscience pass these creations off as my own, or publish them online on a personal or business website when something non-synthetic would suffice.
And I am never personally confiding in an LLM like it were a person. I have had it help me brainstorm options for office politics stuff, but I'm not about to ask it for relationship advice or to be my friend.
I do love that it accelerates the tedious stuff, and helps me learn new things pretty quickly if used right. It has definite utility. But I am always really distrustful of it. Sometimes at work we are asked to share how we use AI, and I have actually refused before, on the grounds that I may have found a useful way to use the AI, but I am worried that others will use my same method badly (e.g., not verifying eveything the AI says first), and I would rather not share.
It's like I have a finicky gun. I might be comfortable shooting it since I know its quirks and how to keep it from accidentally discharging, but I'm not loaning it out to anyone I wouldn't want to accidentally shoot themselves with it.
That seems unlikely given the diverse nature of mutually exclusive opinions that exist out there.
Critics seem to run the gamut from LLMs being incapable of even the most basic of functions to already sentient creatures secretly plotting our destruction with steganographic messages to each other.
It's maybe a bell curve with some wacky at those tails, but there's some fairly significant differences of opinion amongst the positions that are more mainstream.
Just the difference between critics of all LLMs and crutics of all closed weights models are a pretty big gap.
Similarly for those who criticise them for over censorship vs those who criticise them for unrestricted generation.
It's sad that this can be true because if you did them alone the quality would be non-existent.
What are people actually doing with all these tokens? I use LLMs pretty heavily for development, and I'm rarely spending all the tokens that come with a $10/month OpenCode Go subscription...
I write specs and review, but the LLM writes all the code, tests, etc and only gets re-prompted when it screws up badly. That said, I'm not running that many agent loops in parallel, so maybe thats where the cost bites
I'm doing lots.
I'm building a social media site that nobody will ever use. I'm printing an ancient game to Linux and expanding the features (such as adding hw acceleration where it isn't needed). I'm troubleshooting crash logs for an Xbox game that has maybe 50 active players. Im building a distributed customer service system that provides LLMs, RAG, tool calls for lots of integrations (Gmail, calendar, sheets) (this one actually has several paying users). I'm building games, and releasing them on the Google marketplace (play store?)
But what used the most tokens was a kind of agent swarm I experimented with to implement a complex desktop application for very specific data manipulation. That.. half worked.. but it's buggy.
I made several internal tools at my company that are actually getting used by pretty much everybody. That's my most successful work I think. Oh, and I created a backup deployment in Azure for my companies product because .. we had the azure credits, so why not?
This is where I take issue. I'm in a similar boat to the author. In the last couple of months, I've been experimenting with increasing the use of local and cloud LLMs for my research code. I'll create a prototype, maybe port it to a language I don't use very much like Rust, run some tests... but at the very end when I'm very happy with it, I _need_ to go line by line and understand _everything_ that is happening. Sometimes that means using an LLM to understand it, but even when I do and there is a concept I don't get, I try to read primary resources written by experts.
The least bad thing I've found LLMs good for is ideation because it's super easy to take the good nuggets and leave the bad, but even that carries risks of shaping thought and making everyone reach for and ignore the same ideas in the way the Spotify radio or YouTube autoplay has been shaping/flattening tastes for the worse.
I'm not sure what I'll rule at the end of my experiments with LLMs, but right now I'm enjoying the rush of having prototypes that run quickly. I've always been a top-down learner, being motivated by hacking a cool demo I half understand and progressively tearing it apart.
I'm currently writing an onboarding doc for my team, encouraging LLM use for some tasks. (OK, well, I'm actually procrastinating by reading HN).
At the same time, I'm in a darkened office with tinfoil on the windows and a fan pointed at me because it's hell outside and it has been for weeks, and every year it seems to get hotter and hotter and we have longer and longer heatwaves.
This seems ... discordant, at a minimum.
Really, _should_ we be using these things to speed up, say, dependency updates if the cost is the planet? I wanted to know what the author thought about that.
Carbon footprint of commute is in KGs of CO2. A day's worth of AI, depending on model, is ~15-100g of CO2.
Heck food you eat has more than a few Kgs of footprint everyday.
What about training those models? Or the usage of all of the data centres? The projections are that by 2028 a fifth of all energy consumption in the U.S. alone will be for AI.[1]
[1] https://www.technologyreview.com/2025/05/20/1116327/ai-energ...
https://libguides.usc.edu/blogs/USC-AI-Beat/hidden-costs-of-...
Wait a sec, aren't companies also making LLM usage compulsory?
Wow it's starting to sound like they don't care about the environment!
That would be in the same order of magnitude of an average daily commute with an older car.
This is because.. sometimes an idea is just bad. I learned the hard way the other day when I filed this PR against prek: https://github.com/j178/prek/pull/2302
My idea was bad.
The implementation was meant to solve my problem. All it did was expose more problems and waste my time and the maintainer's, and god knows how many tokens from Claude and Codex. In the end the PR was useless so I closed it.
The maintainer and I would have both been better off - less time wasted - if I was forced to forever wonder if the idea was good, letting it bounce around the back of my mind.
I suspect a lot of people are in this boat. We both use it and cheer against it. But it's not cognitive dissonance, it's just a recognition that there's a useful tool being controlled by vile people.
Hypocritical garbage. Into the blacklist you go.
If you can’t even use your own words for basic communication then there’s absolutely no reason for me to take you seriously.
At least the "industry" ones used for "networking" and all that other crap. Where's the news here? And why does the author not take a step back and realize that there is none?
The only evidence you have is US here and all the frontier models from China are free to download. Pure projection.
You can read China's AI strategy here: https://ipc.court.gov.cn/zh-cn/news/view-5766.html
> all the frontier models from China are free to download
Maybe? Again, not hard to imagine that they might not be releasing their actual frontier models to the public.
Short version is this: putting export controls on AI open-source will hinder their own internal open-source ecosystem
If you are bringing in juniors with the expectation that they won’t produce any value for some time then it might be better to block them from using LLMs.
Do this and you will get a better understanding of where they are in the process and they will probably learn the ropes more quickly.
A) I think many companies no longer view juniors as an investment, they want them to be productive. They view juniors as "cheaper almost-seniors"
B) Blocking them from using LLMs would mean acknowledging that there is a downside to having your employees using LLMs. If they prevent your juniors from improving, what negative effect are they having on your seniors?
I have linked this elsewhere in this thread, but the studies on this matter suggest a worse outcome than just losing hand crafted or "artisanal" software: https://arxiv.org/pdf/2604.04721
> These findings are particularly concerning because persistence is foundational to skill acquisition and is one of the strongest predictors of long-term learning.
> We posit that persistence is reduced because AI conditions people to expect immediate answers, thereby denying them the experience of working through challenges on their own.
This issue is not that we are losing the ability to write good software by hand. It is possible we are losing our internal tools to learn new things by circumventing a necessary part of the process required by our biology.
It seems to me this here is the dangerous 'force multiplier' for destructive outcomes.
It's one thing for a kid to become obsessed with a fictional character. It's a whole different playing field when that character can talk back, remember things, and potentially surface all kinds of subject matter which cannot be predicted and could at times be harmful.
Not even remotely the same level of impact. I think we're almost at the point where LLM education (as in, teaching awareness of the basic tech and its shortcomings) needs to start in Kindergarten.
LLM's are replacing your 'brain' i.e. the deepest most amazing part of you. The potential value of this to business is astronomical!
No there won't be an explosion of new software, a minor blip perhaps and even this one of diminishing returns (each cycle of LLM improvements requiring less and less human input).
If there are 25 million software developers today then the ROI by OpenAI, Anthropic et al pays off if in a few years time only 15 million developers are needed. That 10 million is the ROI and that is likely to be you.
now i am convinced that it's the same matter as with people who forget most of what they learned after the actual exam vs people who go nuts and can't progress if they didn't understand a foundational part of a formula, fact or theory.
I can honestly say that while using LLM my thoughts are as sharp and as fast as they were before, if not sharper and faster. But my workflow is that I watch the LLM as it rips, have verbose output turned on so I also read Claude's thinking process (and lots of interesting information is hidden there).
This helps me learn new stuff I wouldn't think of. Helps me stop Claude if he's veering off track. Makes it easier to proofread what's Claude doing and why step-by-step and easier to spot any mistakes (because they often come from wrong assumptions you wouldn't see if you didn't see the thinking process).
If you just let your agent write thousands of lines of code, reviewing them is painful. If you also run multiple agents for multiple tasks at the same time, your focus is divided, so you aren't spending all your brain power on the task at hand.
Writing is thinking. If you're outsourcing your writing to an LLM, you are shortchanging yourself by skipping over much of the refinement of ideas that the process of writing provides.
I wish the author had clarified how he uses LLMs for writing. It's perfectly fine to have an LLM proofread and fix your grammatical errors--that's a mechanical task. But I think it's gross when it starts putting words and content on the page.
No it just sounds completely stupid. You spent $10k to get a program to write text that you can write yourself just as well and for free. "Dissonance" doesn't even begin to describe it. It's like suddenly everyone has turned into the listless blobs of fat that humans have turned into in Wall-E.
I just hope all that will eventually correct itself: open (one way or another) models will reach parity with proprietary ones when diminishing returns eventually make it impossible for LLMs to grow any longer, OpenAI and Anthropic will succeed in bankrupting each other, it will become possible for everyone to run their own, personal, free and open (one way or another) model on their own laptop (assuming we still own our laptops) and at that point most of the problems with LLMs, particularly the ones that come from having to pay someone else to do the things you could always do yourself for free, will go away.
I just hope.
"You spent $1k on a flight ticket from SFO to NYC when you could have just walked yourself just as well and for free"
It is. That kind of money would change my life and this guy spends it on prompts that "really force him to think".
I use LLMs. So far, 20 bucks per month has me covered. I hit the limit sometimes. I just take a walk. Forces me to think.
No, they come duller and increasingly dependent on the LLM doing them for you.
And that happens to the generations that learned to code earlier without the LLM. What would happen to the later ones, those getting in uni after LLMs emerged, that would never get a real feel for the basics to begin with?
Yes there is a dissonance, using LLMS just took their dot point notes and spread them thin. I personally would prefer just the notes and some personal anecdotes to tie the notes into a cohesive whole. This, this approach here, is not better.
Humans have been gravitating into cognitive thinking and not specialized tasks, some of us, a handful do repetitive tasks and that is ok, but, the majority of us work better in context, in guidance. LLM offers that, offers that abstraction that we were and are looking into all aspects of our lives. We once plow the fields, and harvest by hand, and was good, but suddenly became cumbersome, too much to handle and machines came along to do the job of 20, faster and better, and there humans just tweak the machine, teach the small repetitive tasks. We did that with everything and we change our lives, the way that we upskill ourselves and our communities, we let specialized human tasks as a comfort/skill/romantic/artistic outlet, like some specialized vineyards or specialized farms that do specific very specific things. But the rest we delegate and move to another areas. I believe the problem is not LLM/AI is how we are going to up skill, repetitive jobs no longer need us, so we need to upskill for context, for management, strategy