Posted by yegg 3 hours ago
It's tough to answer because you want to hedge for both an AI enthused employer and an AI hesitant employer with limited information about who they are and how they personally use these products. I've been responding with a sort of long winded answer about how 'there is clearly a learning curve for how this technology fits into any process and how I always always always double double double check yadayadayada'
I'm probably using the chat/ask functionality on a daily basis for quick debugging / new technology learning questions but I have yet to really use the fully agent or computer-use products because I've had more bad results than good the few times I've tried them (re-factoring a big repo of decades old fortran+C code for modern compiler/OS some things started to work but ultimately I abandoned that effort).
Have you considered just answering truthfully?
Would you even want to work somewhere where you need to play a role and where they flip out when you say the wrong word you should've correctly guessed through mind reading? That sounds not like a job but a toxic relationship.
Fair enough, so if there were one “right” answer, that would be the one to give whether true or not.
But here there is no obvious right answer. If the employer is looking for a particular answer, the poster doesn’t know what it is. In that case, the best thing to say is simply the truth, particularly when the truth that the poster gives here is completely reasonable.
The attitude suggested by your response suggests you haven't lived that reality yet.
Either way, I'd rather be rejected by an employer for speaking my truth, than lie to be somewhere I'd rather not be.
"Speak the truth, even if your voice shakes"
You were replying to “The job market is much different when you're just starting out”. The past is not now, and you are not just starting out, so your comparison of their position and yours is invalid IMO.
> and will do it again.
Good for you for sticking to your guns, I'm about to do the same with a company that has all but said “dig into AI or get left behind”¹, but those starting out as freshly minted grads likely do not have the luxuries that we might have² and the jobs market is freakishly competitive for them right now³ in a way that I don't think it ever has been before.
--------
[1] time will tell if I leave of my own volition before getting kicked!
[2] experience (both actual experience and experience “talking the talk”) to help getting the next gig, a mortgage paid off so making ends meet is easier, etc.
[3] It had been heading that way for a while, the recent explosion of GenAI+agnetics has made it worse.
I certainly feigned enthusiasm when I was in high school to get an after school job in order to help my family buy food.
I bootstrapped myself from poverty to Staff software engineer, past the age of 45.
Is that privileged? Or sheer will and force of effort?
I am not unique. I am an example.
So while I agree that privilege is certainly a factor, so is what I've just said.
A lot of people here live very cushy lives that cushion them from very pointy thoughts and questions. As someone who too has to live in this world, I'd rather they didn't.
It is very sad to me that people do feel that pressure, and how the current job market is.
On topic with the article, I would love to be able to trust AI with more, but have found that I have some useful moments with it, but more because of Internet search not being how it used to be for quality.
For example I think the decision to stick to certain morals is very hard if someone has a disabled dependent, are disabled themselves, or require consistent access to healthcare. There are different lines for different people of course. Our ire shouldn't go towards individuals who make these decisions but the people in power who force others to be in a position where these decisions need to be made.
I don't want to preach martyrdom, but I am also offended by people choosing moral bankruptcy when faced with even the slightest hardships.
My truth is I don't care either way . I get the sense that's the same for parent poster. They just want a job and to say the right thing to get past the hiring filter. Even if I did have a truth its not something I would put above being remote, pay and how a company develops software. I'd rather not have a truth and not have a daily standup.
I think you'd rather have good odds at some companies and 0% at others, rather than abysmal but non-zero odds at all companies.
And as an added bonus, you might get hired at a company where you're actually a good fit, rather than one you weasled your way into, and get to pay rent, food bills, and other expenses through employment for a long time!
"Wouldn't give a straight answer on question X" isn't an instant no-hire, but it's not a positive signal.
Plus, and leaving that aside, I have my doubts that even if you did that, that that company would stay alive for very long. Reality has the habit of eventually ripping this kind of unproductively delusional people (like e.g. a boss that flips if you don't say the right word with regards to the current hype) to shreds eventually.
Look at musk's companies. They will basically never (on any near timescale...) produce GAAP profitability and yet their IPO is in the trillions. To the point that S&P refusing to suspend their GAAP profitability requirements means the index will basically never see this company in it (which I'm quite pleased about).
The power of already-accumulated capital is simply more powerful than things like "don't be completely pants-on-head stupid about a recent fad" "don't seig-heil in front of the world stage" "there's no point in having people come to an office just to spend all day on zoom" etc etc etc.
The market can remain irrational longer than you can remain solvent, and companies can remain irrational longer than you can go without contributing to your 401k.
The same might not be true everywhere.
It's… like… not that simple.
Update:
Every street corner has a yellow garbage bin for recycling. That is where your plastic bottles go. Seems like a better system than having elderly going through bins.
The best benefit about working in a large office is that nobody checks the basement.
This just sounds like a standard tech interview. Mind reading to find and perform the secret “signal”. Nobody flips out if you don’t find it, they just move on to one of the other 1,000 candidates for the role.
We all filter and “nudge” the truth during interviews. We all cater our responses to the person in front of us. Let’s not pretend otherwise. Your interviewers sure aren’t.
I remember the graduate recruitment days - If you told the truth you were the only candidate they saw all day that wasn't the captain of the football team, top of the class and voted most likely to succeed - aka the worst candidate they saw all day.
It's 2026, you gotta sell your soul just to get a phone screening
From the 3 people I interviewed, all of the answers are very similar which is along the lines of: Kinda, but we need to be careful of using it, privacy, hallucination, etc.
All very safe answers and doesn't say anything new to me. If they had been more specific about why and their experiences with it, I'd probably favor them more due to their experience with it. It'd also signal to me that they form their own opinion rather than simply following the crowd.
That is of course assuming that they're looking for some long-term stable team member.
A skilled interviewer smells dishonesty.
However, and to be fair, whether and how they act on it depends on the specific situation.
Most of our stuff in this world actually does work, and the reason why it does is that skilled (teams of) people that care have built it. Meaning that these people can be found in many _many_ places.
Very few jobs are looking for opinioned most are looking at people who might fit in unless you are hiring to distory from without.
To be honest, I don't think I would want to work with or hire you, based on your response here.
I'm an old hat on both sides of this type of discussion from a post-grad view.
Recommendation: use it to own the conversation and to signal mutual fit. Yes, your idea of AI lover versus hesitant matters. I recommend reframing the question to pivot to your fit to the org (and org fit to you) question. Show/concisely explain how you consider whether LLMs are fit to a task and how to tell it improves outcomes.
An outcome focus and willingness to show thought process around a common use case will be a substantially strong response.
- Any long-winded answer to a question is immediate out and has been for years.
- Not having used agents and not being able to comment on what to do and what not to do with them is immediate out since early this year.
From all the tech that we have, agents are really not that hard to learn on the job. They're also not a magical silver bullet.
I think upskilling is the right move in this environment and it is dead simple: Invest a couple of days to show initiative, learn agents yourself and be able to speak from true experience.
You’re more or less admitting that you’re playing trendy tech lottery. Which is fine, but maybe not generalizable to the whole industry.
Why?
If the winding path is actually interesting and gives you insights into how the person works, why would that be a bad thing?
I much rather prefer someone who needs 3 seconds to triage a question and tell me: "This is X, I know this, here is the solution" or "This is Y, I don't know it, but I will get back to you within 24h".
I do absolutely not want a "Well let's think jointly about this for a couple of minutes". There is no jointly with your boss. Let's do a some math of a 1:12 manager to direct report ratio. That means for every hour you have, your boss only has 5 minutes. And if you talk to your boss' boss, they have 25 seconds for every of your hours.
So in the same interest of helping post-grad job seekers, do what you've gotta do to get yourself paid, but maybe don't presume that vibe_that_works speaks for every hiring manager.
Not to disagree of course that time is limited, but in my experience, optimizing it this harshly leads to poor results, because eventually, you just get leapfrogged by reality.
Hyper-optimized systems are brittle and can't really adapt to the market changing.
But yeah, I guess they still need developers. Just doesn't sound like a fun job :D
So let me take this a step further. You want to meet your boss' boss for 10 minutes to present them something. 10 minutes of his time are an equivalent of more than 20 hours of your time. So if your initial idea was to "take maybe 1-2h" to prepare for this -> You are underprepared by at least one order of magnitude.
Which might not be ideal, because "orging for the sake of org" to my understanding consumes significant resources not going into building products/marketshare/shareholder value.
But then again, I'm no hiring manager in such a structure, so this is probably just an uninformed take.
But why?
Most of my most fulfilling experiences in tech have come out sitting down and hashing out a problem with someone else (including with managers/leaders).
It sounds like a miserable org if I am not expected/allowed to have an actual back and forth conversation with my boss. If I'm employed to be on a team working on an aligned common goal, why would I not use that collective skill and experience to my fullest advantage?
You're describing a coding sweatshop. What is the point of any discussion at all then? If the "boss" can't carve out enough time, that's their own problem. Letting that stress propagate to the team is plain bad leadership.
I know you might think some of these candidates don't have other much better choices to find work, but they absolutely do.
But that sounds more like "evasive" is the problematic attribute and not "long winding".
Which does show up at the same time often, true. But not always.
want a Flutter developer who is unusually strong at directing AI-driven software delivery. This is not a traditional "write the code yourself" role.
Having been in academia in the past and now in software I can say with a lot of certainty that this will take a lot more upfront work than otherwise.
Academic code does not have a lot of structure. And usually lacks a lot in terms of tests. While AI is best when it can mimic patterns as well as there are tests to target.
So you will probably need to budget a few weeks to establish good patters, docs as well as testing patterns before you can seriously make it really do what you want it to do.
Even with 3 weeks I'm just not the Fortran/C programmer to get that job done so I moved on to other things.
The analogy I've had for myself is that it feels like using a bulldozer to dig rather than a shovel. If you use it to dig archaeological artifacts, it can make things worse than you started. A lot of the work however, is just moving dirt around, so you are wasting time by using a shovel.
That's probably not going to be enough for AI maxxers, but it probably won't be too much of a turn off for anyone but the most extreme AI minners, and everyone in between will probably be fine with it.
Frankly I plan to steer well clear of any "the majority of our code is AI generated" shops for the foreseeable future. Seems like disasters waiting to happen and I'd rather let other people step on those rakes
Look at the uptime and incident rate of all the big tech companies that have gone all in on AI generated code
As of 2023, 27% of American working-age adults were at a PIAAC Literacy Level of 1 or below, out of a total of 5 levels. This has gotten drastically worse in the past 10 years as, in 2013, Level 1 and below was only 17%.
Full scores for 2023 are: % Level 1 or below: 27% Level 2: 29% Level 3: 31% Level 4/5: 13%
For reference, Level 1 means someone can't really handle a full page of text, and can sort of handle simple 1-page web pages. Level 2 is the point where someone can start to handle a few pages of straightforward text, but still nothing particularly complicated.
(Both of those descriptions undersell just how bad it really is, but I'll leave it at that, for the sake of brevity.)
People that aren't using AI at all often aren't using it because they effectively can't. On a fundamental level.
Source: https://nces.ed.gov/surveys/piaac/2023/national_results.asp
In that case, it's way better to simply write the code yourself.
IMHO the best of both worlds option is agents working with deterministic CLIs. Where the agent does the reasoning (and text generation) but uses CLIs to carry out all of the actions (issuing refunds, unblocking accounts, or whatever).
It's possible to get very reliable and consistent work out of agents when they're using well written prompts with well designed CLIs.
Although you can certainly do a better-and-worse job of preventing these kinds of issues.
Some people might use skill-based scripts, MCPs, or some kind of raw access to a database. My point is that well designed CLIs are the optimal programmatic interface, for many reasons.
Wait raw access to the database? That’s one of the options for issuing a refund?
At Big Tech Company I Work At the LLM is quite happy to make raw API calls. If it thinks the data is big, then it'll write a Python tool to do it.
The reason crafted backing CLIs are useful is you can guide the LLM towards stuff that is immediately useful rather than hoping the nondetermism can separate the wheat from the chaff.
Take CI: is it interesting to know which tests passed? Maybe, but probably not. What is really interesting is what failed. Instead of having the LLM go out and talk directly to the CI system, write an intermediate CLI that filters out less actionable stuff by default, and have a flag that'll deliver the full dump if necessary.
It's a skill to do this stuff, and it's a lot of hard won experience than something I think is easily teachable. You kind of have to feel out your model and how it "thinks" about solving problems.
And then a new model version comes out and you have to learn it all again!
Some systems do support issuing refunds, among many other actions, by creating an appropriate row in a database.
But that's not worth trillions of dollars...
We are slowly waking up to the fact, which was always true, that “coding” is just a fanciful preparatory task in order to appease the spirits properly so that we may invoke the spirit of what we are actually after: a live, running process that does useful things. Code is completely useless when separated from that fact.
Typing it is a complete waste of time unless getting up close and personal with it will result in some kind of useful and actionable improvement in you or your understanding. Knowing when it does and when it does not have this property is a skill of its own.
I believe this is the general belief about basically every human skill, that if you stop doing the technical fundamentals you get worse at understanding the activity. The question is whether coding is like sailing a square-rigged wooden ship, which became completely useless knowledge after the invention of the steam engine, or if it's like playing an instrument, which while technically unnecessary after the advent of MIDI and other tools, absolutely hurts your ability to arrange, compose and perform if the skill is neglected.
For my money: I think the AI scenario is more like the latter, but "humans are worse at coding" isn't the consequence I see coming. I worry that in ten years we will be awash in software that's impossible to understand. I don't think that's happened in any human industry ever. Someone has always understood how the machines are built, even if they're very remote from the users of the machine.
Like, perhaps, understanding that it is free of security and functionality bugs.
If you find yourself writing repetitive code you should consider adding a layer of abstraction. If your language isn't powerful enough you can write a code generator.
Code is obscenely low level.
No one has ever needed to do that for something that is new. And if it’s not new, you want to do it repeatedly with some guarantee of reliability. Not just in an uncontrolled manner.
That is why we have snippet systems, macros and code generators. And the best with code is to solve problem once and reuse the solution. Which we have done with libraries, frameworks and supporting software.
I would argue that this is nearly always the case. I don't think people really understand programs that they've only read at more than a very superficial level. This is why I tend to make (temporary) small changes, printlns, etc. when exploring a new code base: it aids greatly in understanding how a program actually works.
And it's even worse (in my experience) with LLM generated code, as it tends not to result in particularly understandable code. It is a lot like LLM generated prose: it often looks entirely reasonable at a surface level, but has a of weirdness/incorrectness hidden beneath the surface. But that surface level makes it very hard to avoid glossing over the details when reviewing the code. For this reason, I personally find it's much more effort to carefully review code than it is to write it.
Humans make mistakes all the time, but their code tends to naturally be structured for human understanding (to some degree based on skill/experience) because they themselves needed to understand it to write it.
I think LLMs are very useful tools, but after quite a lot of experience using them, I think it's generally better to use them as a sounding board, or to help you get unstuck or remove points of friction. Using them to write all of your code (at least for me) seems like a net negative.
I also think it's extremely easy to overestimate how much time they save. It feels like they're a productivity boost because it takes less intense focus to implement something. But I've experienced several instances where actually writing the code myself would have been both quicker and have resulted in better code.
All that being said, it can also be really hard to not write all of your code with agents once you get used to it. There's also a kind of slot-machine-like effect where you write a prompt, excited for the result, and when it doesn't quite come out right, you think "ah just one more prompt and it'll be good." It's hard to see when you're actually doing it though.
It's also weird to me how much people think typing is what the LLM is replacing. Typing was never the hard part. It's the translation of the high-level idea into an unambiguous process that's hard. That's also the valuable part, that requires thinking through the edge cases and consequences of decisions, and that just gets glossed over when using an LLM unless you rigorously review what the LLM has done.
At the end of the day there's a real tradeoff to be made, and it's worth being conscious of what's being given up.
That is one of the things code does. It also communicates the developer's thoughts about how that process should work to others. If the latter is neglected, the code becomes very difficult to collaborate on. Very few lines of code that are written are "write once". Mostly they're changed, repeatedly, over time by many people. The live, running process is a very temporary entity by comparison. Yes, it needs to exist and do useful work. No, it is absolutely not the only thing that matters.
Democratizing access to code generation so that people can craft little personal utility scripts and quickly draft mocks/prototypes is great, but a lot of people on sites like this work on much more sophisticated projects and understand that developing a thorough enough specification for AI to be able to run with isn't meaningfully different in effort than what we've already been doing for the last 50 years. And in fact, for those fluent with their traditional tools and workflows, trying to craft those specifications as an English prose conversation for AI is often much more work that takes longer and is less reliable.
Those costs don’t disappear and it’s truly naive to think they don’t matter. Take security issues, they may arise because what you thinks was the input is merely a subset of the true input range. And the extra possibilities lead to unforeseen behavior.
A lot of programming is about ensuring that the input and the output are the sets defined in the specs. And the rest is that the transition/relation is the right tradeoffs of performance, correctness, and costs.
Instead of using the LLM to create deterministic tools, we are using LLMs to replace them. It's completely backwards and I don't know why people (especially high ranking people in my company at least) seem to think that this is the way forward. No, I don't want a whole CI pipeline that is just LLM prompts. Yes it's very easy, but it's expensive, slow and prone to failure in ways you can't even predict.
Same things like using LLMs for the code review process. What would have been a simple linting rule is now a pass with an LLM rather than using the LLM to create the linting rule, which it is absolutely excellent at creating.
Yes, and we're also seeing lots of companies claiming they're using "AI" and it's just deterministic under the hood.
The agent paradigm will eventually give way to experiences that are a hybrid of deterministic and non deterministic and you won’t even know the llm was involved or visible.
The issue is, they don't want to provide "better" support but "cheaper" support. Imagine a trained agent that understands the big picture. Now imagine a company investing in humans to use AI to retrieve knowledge that the human can easily identify as being relevant or not, and using that knowledge to better aid the customer.
Right now AI is being sold as a "we don't need support personells" instead of "how can we provide better service." For a lot of products, better service will probably not matter as "cheaper" products will win most of the time.
Most people don't want to pay for better. They want to pay the same for something better, which is what companies are not investing their time in figuring out how to use AI properly for I think.
A lot of people want to pay for better, but that is hard. Better is more expensive, most of the time, but being more expensive is no guarantee for being better. It feels like the correlation is very weak. Most expensive products are just expensive, not good.
If there was a reliable way to identify the "better" thing, I and a lot of other people would go for that every time we can.
Instead of refining their approach, or challenging their current knowledge base for discovery of inefficiencies or baseless assumptions, they'd rather hit an "easy" button.
I understand the desire to NOT do work. I understand the desire to spend quality time and free time with family. And I understand the idea that familiarity breeds contempt.
What I don't understand is the willingness to replace a deterministic language/framework/approach with a probabilistic slop machine.
Now that’s real value.
The AI psychosis is a real thing.
Regardless which task is handed to him, he "discusses" it first with Claude and very often comes back with like "The AI said... X"
These people just destroy their ability to read and understand the systems they're working with. I actually see it as them making themselves redundant. Because if you can't understand anything without Claude, and Claude doesn't even give the right answers, then what are you worth?
I bet lemmings are grateful they were left behind.
It beggars belief that people think that they should rush in some uncertain direction, like some drawbridge is going to be lifted the moment people work out what the right direction is. It's utter stupidity.
Of course I will do that, I get paid for doing that.
Most of the times I can convince that AI is not necessary by showing small PoC flow with AWS diagrams of data flows. This works well especially if the ask comes from technical people.
Other times the C-level interjects (CEO, CFO, sometimes even CTO) and demands that AI should be there. I literally had CEOs send me instagram reels of some AI shovel-sellers to demonstrate that I am wrong and AI is the way to go. No point arguing after that because I have no problem implementing whatever AI they want rather than losing a paying project.
These models do not have any experience. They're not sentient. And are in no way capable of being "smart", let alone becoming "smarter".
<b>Included in your plan limits until Jun 22</b> <br><br>Fable takes 2x the usage of Opus.
<b> Switch models when a message is flagged</b><k <br> When safety measures flag a message, automatically switch to a different model to keep chatting. When off, your chat will pause instead. <a href="https://support.claude.com/en/articles/153636
target="_blank" rel="noopener noreferrer" > Learn more</a>
Ok wait maybe not the next one but surely the one after!
Hasn’t happened yet and there is no evidence it will.
I’d actually love it if LLMs could skip the slow high level lanaguages entirely and just churned out some weird LLM bytecode that was closer to the metal. I don’t want to read it or understand it at all. Here’s my spec, build it and notify when done. I want to ship stuff not build or dick around with code. Basically like when I go to a shop because I want a table, I don’t care if some carpenter “crafted” it or a machine mass produced and spat it out. It’s cute, but most people just want stuff and don’t care how it’s built.
In my experience, it's a mixed bag. I wrote this comment[0], yesterday. It reflects my current work, and how I am integrating an LLM.
I have used it for two parts of my project:
1) The backend (PHP), and
2) The frontend (Swift)
It has been a huge help, in both, but #2 is a cautionary tale. It really needs adult supervision, in developing native UIKit Swift apps. I'm realizing how truly bad the code it wrote was. I mean, terrible.
That's jarring, because it did a great job with #1. It made sound, reasonable design decisions, and provided code that is better than what I would write.
With #2, it behaved exactly like an inexperienced engineer, panicking, when confronted with real-world problems. My rewrite is going to feature a much simpler, sound approach.
All that said, it has been a net positive, and has increased my productivity by a large margin.
I guess the lesson I needed to get from this, is that it is good at helping me to find problems, but maybe not so good at fixing them.
It saddens me to see that high quality content is drowned in this sea of garbage to the point of being almost impossible to find.
Then you go on to search for something, and find only results that are clearly AI generated pages and come to the conclusion that directly prompting some LLM is better than reading an AI slop page that's output by the same AI for slightly less specific prompt.
My concern is that this will only get worse over time - which is great for companies selling AI tokens and bad for society and whoever wants to interact with other humans over the internet.
Swift, not so much. It's relatively new. Looking at AI's abilities like an engineer's career span scaled about 10-20x of time makes it make a bit more sense.
It's going to be worse at newer/niche things, intuitively - which is only going to get worse as it "learns" from garbage outputted by other LLMs moving forward.
You seem to assume that autoregressive pretraining (and unfiltered behavior cloning, maybe) are the only ways to improve LLM performance.
I'm unlikely to run into many of the problems that (for example) the PornHub developers hit, several times an hour.
In that case, I benefit from folks like you, that allow me to have solutions that scale down to my level.
That's fine, for a lot of corporate applications, but not for the stuff I write. I'm anal, I know, but that's how I roll.
The classic AI Gell-Mann effect.
The diagnosis, however, is not.
Have a great day!
But I've seen Claude write crazy code in Python and JavaScript, too
PHP has huge, entire frameworks and systems, refined over years.
That's one of the things that I appreciate about the PHP that the LLM provides. It uses modern idioms that make better use of the modern language.
That's an interesting analogy as, despite the real ecological issues with it and principled arguments against meat eating, in general meat consumption has trended upward globally in country after country for decades.
But 2. For most other things, LLMs are fairly underwhelming. Research is usually mediocre. Try being rigorous and repeat your research prompt many times - then make a confusion matrix to tally up how many false positives and false negatives occur. And for the rest, be honest and ask yourself if the LLM is doing much more than a basic search engine query or trip to Wikipedia would have told you. For “normie” use cases, it’s handy-ish but far from revolutionary
How does this connect to everyone's high level ideas/thoughts about "tech", "AI" and "morals and feels" etc. These lines can start to seem a little blurry, at least for me.
For example, would we say my partner is "using AI" (for all intents and purposes), if she's frequently using Google.com throughout the day, and then ends up picking and believing the AI generated answer overview at the top of the SERPs almost every time?
Or do we feel "uses AI", is more along the lines of the vampire kids running 1000 sub-agents on a mattress floor in SF?
I kind of find the whole spectrum really interesting because even basic phone use is now stuffed with AI, whether we choose to label it or not.
"low effort and convenient" seems to consistently win over "best quality" and this is going to be a downgrade in everything, for everyone
Google has search results still? I don't use Google much anymore (thanks Kagi), but this is what ends up showing for me, I don't even see any search results anymore: https://i.imgur.com/eHIA2Df.png It seems like it's 50/50 on page reload if the LLM-reply UI expands automatically or not, which covers my entire screen. I guess Google is doing some A/B testing perhaps.
Anyone who does a search and accepts the first answer just doesn't care much or is incompetent. Anyone with any critical thinking whatsoever does way more than that if they want a correct answer.
This makes me less bearish on the AI investments that are being made, if 70% of the working age population isn't using AI then there still is a lot of growth. The future is here, it's just not evenly distributed (yet)
