Posted by mips_avatar 5 hours ago
I haven't escaped this mindset myself. I'm convinced there are a small number of places where LLMs make truly effective tools (see: generation of "must be plausible, need not be accurate" data, e.g. concept art or crowd animations in movies), a large number of places where LLMs make apparently-effective tools that have negative long-term consequences (see: anything involving learning a new skill, anything where correctness is critical), and a large number of places where LLMs are simply ineffective from the get-go but will increasingly be rammed down consumers' throats.
Accordingly I tend to be overly skeptical of AI proponents and anything touching AI. It would be nice if I was more rational, but I'm not; I want everyone working on AI and making money from AI to crash and burn hard. (See also: cryptocurrency)
I use it for throwaway prototypes and demos. And whenever I’m thrust into a language I don’t know that well, or to help me debug weird issues outside my area of expertise. But when I go deep on a problem, it’s often worse than useless.
To management (out of IC roles for long enough to lose their technical expertise), it looks perfect!
To ICs, the flaws are apparent!
So inevitably management greenlights new AI projects* and behaviors, and then everyone is in the 'This was my idea, so it can't fail' CYA scenario.
* Add in a dash of management consulting advice here, and note that management consultants' core product was already literally 'something that looks plausible enough to make execs spend money on it'
If you ask for an endpoint to a CRUD API, it'll make one. If you ask for 5, it'll repeat the same code 5 times and modify it for the use case.
A dev wouldn't do this, they would try to figure out the common parts of code, pull them out into helpers, and try to make as little duplicated code as possible.
I feel like the AI has a strong bias towards adding things, and not removing them. The most obviously wrong thing is with CSS - when I try to do some styling, it gets 90% of the way there, but there's almost always something that's not quite right.
Then I tell the AI to fix a style, since that div is getting clipped or not correctly centered etc.
It almost always keeps adding properties, and after 2-3 tries and an incredibly bloated style, I delete the thing and take a step back and think logically about how to properly lay this out with flexbox.
If your AI reliably generates working code from a detailed prompt, the prompt is now the source that needs to be maintained. There is no important reason to even look at the generated code
This is a common intuition but it's provably false.
The fact that LLMs are trained on a corpus does not mean their output represents the median skill level of the corpus.
Eighteen months ago GPT-4 was outperforming 85% of human participants in coding contests. And people who participate in coding contests are already well above the median skill level on Github.
And capability has gone way up in the last 18 months.
https://mikelovesrobots.substack.com/p/wheres-the-shovelware...
Basically, if the tools are even half as good as some proponents claim, wouldn't you expect at least a significant increase in simple games on Steam or apps in app stores over that time frame? But we're not seeing that.
Coding contests are not like my job at all.
My job is taking fuzzy human things and making code that solves it. Frankly AI isn’t good at closing open issues on open source projects either.
It does, by default. Try asking ChatGPT to implement quicksort in JavaScript, the result will be dogshit. Of course it can do better if you guide it, but that implies you recognize dogshit, or at least that you use some sort of prompting technique that will veer it off the beaten path.
But to the untrained eye the AI did everything correctly.
Its easy to recall information, but something entirely different to do something with that information. Which is what those subject ares are all about - taking something (like a theory) and applying it in a disciplined manner given the context.
Thats not to diminish what LLMs can do. But lets get real.
I'm having a dejavu of yesterday's discussion: https://news.ycombinator.com/item?id=46126988
AI is optimized to solve a problem no matter what it takes. It will try to solve one problem by creating 10 more.
I think long time/term agentic AI is just snake oil at this point. AI works best if you can segment your task into 5-10 minutes chunks, including the AI generating time, correcting time and engineer review time. To put it another way, a 10 minute sync with human is necessary, otherwise it will go astray.
Then it just makes software engineering into bothering supervisor job. Yes I typed less, but I didn’t feel the thrill of doing so.
I'm pretty sure this is the entire enthusiasm from C-level for AI in a nutshell. Until AI SWE resisted being mashed into a replaceable cog job that they don't have to think/care about. AI is the magic beans that are just tantalizingly out of reach and boy do they want it.
... except they didn't. In fact most AI tech were good for a nice demo and little else.
In some cases, really unfairly. For instance, convnet map matching doesn't work well not because it doesn't work well, but because you can't explain to humans when it won't work well. It's unpredictable, like a human. If you ask a human to map a building in heavy fog they may come back with "sorry". SLAM with lidar is "better", except no, it's a LOT worse. But when it fails it's very clear why it fails because it's a very visual algorithm. People expect of AIs that they can replace humans but that doesn't work, because people also demand AIs never say no, never fail, like the Star Trek computer (the only problem the star trek computer ever has is that it is misunderstood or follows policy too well). If you have a delivery person occasionally they will radically modify the process, or refuse to deliver. No CEO is ever going to allow an AI drone to change the process and No CEO will ever accept "no" from an AI drone. More generally, no business person seems to ever accept a 99% AI solution, and all AI solutions are 99%, or actually mostly less.
AI winters. I get the impression another one is coming, and I can feel it's going to be a cold one. But in 10 years, LLMs will be in a lot of stuff, like with every other AI winter. A lot of stuff ... but a lot less than CEOs are declaring it will be in today.
Last one at work: "Hey, here are the symptoms for a bug, they appeared in <release XYZ> - go figure out the CL range and which 10 CLs I should inspect first to see if they're the cause"
(Well suited to AI, because worst case I've looked at 10 CLs in vain, and best case it saved me from manually scanning through several 1000 CLs - the EV is net positive)
It works for code generation as well, but not in a "just do my job" way, more in a "find which haystack the needle is in, and what the rough shape of the new needle is". Blind vibecoding is a non-starter. But... it's a non-starter for greenfields too, it's just that the FO of FAFO is a bit more delayed.
But unfortunately the nuances in the former require understanding strengths and weaknesses of current AI systems, which is a conversation the industry doesn't want to have while it's still riding the froth of a hype cycle.
Aka 'any current weaknesses in AI systems are just temporary growing pains before an AGI future'
I see we've met the same product people :)
That's when I realized how far down the rabbit hole marketing to non-technical folks on this was.
Working on our mega huge code basis with lots of custom tooling and bleeding edge stuff hasn't been the best for for AI generated code compared to most companies.
I do think AI as a rubber ducky / research assistant type has been overall helpful as a SWE.
Example: you might spend less time on initial development, but more time on code review and rework. That has been my personal experience.
I find people mostly prefer what they are used to, and if your preference was so superior then how could so many people build fantastic software using the method you don't like?
AI isn't like that. AI is a bunch of people telling me this product can do wonderful things that will change society and replace workers, yet almost every time I use it, it falls far short of that promise. AI is certainly not reliable enough for me to jeopardize the quality of my work by using it heavily.
Rinse and repeat for many "one-off" tasks.
It's not going away, you need to learn how to use it. shrugs shoulders
I work as the non-software kind of engineer at an industrial plant there is starting to emerge a trend of people who just blindly trust the output of AI chat sessions without understanding what the chat bot is echoing at them which is wasteful of their time and in some cases my time.
This not not new in the past I have experienced engineers who use (abuse) statistics/regression tools etc. Without understanding what the output was telling them but it is getting worse now.
It is not uncommon to hear something like: "Oh I investigated that problem and this particular issue we experienced was because of reasons x, y and z."
Then when you push back because what they've said sounds highly unlikely it boils down to. "I don't know that is what the AI told me".
Then if they are sufficiently optimistic they'll go back and prompt it with "please supply evidence for your conclusion" or some similar prompt and it will supply paragraphs of plausible sounding text but when you dig into what it is saying there are inconsistencies or made up citations. I've seen it say things that were straight up incorrect and went against Laws of Thermodynamics for example.
It has become the new "I threw the kitchen sink into a multivariate regression and X emerged as significant - therefore we should address x"
I'm not a complete skeptic I think AI has some value, for example if you use it as a more powerful search engine by asking it something like "What are some suggested techniques for investigating x" or "What are the limitations of Method Y" etc. It can point you to the right place assist you with research, it might find papers from other fields or similar. But it is not something you should be relying on to do all of the research for you.
There's a shit-ton of bad and inefficient code on the internet. Lots of it. And it was used to train these LLMs as much as the good code.
In other words, the LLMs are great if you're OK with mediocrity at best. Mediocrity is occasionally good enough, but it can spell death for a company when key parts of it are mediocre.
I'm afraid a lot of the executives who fantasize about replacing humans with AI are going to have to learn this the hard way.
There’s so much evidence out there of people getting real value from the tools.
Some questions you can ask yourself are “why doesn’t it work for me?” and “what can I do differently?”.
Be curious, not dogmatic. Ignore the hype, find people doing real work.
At one point you might take over, ask it for specific refactors you'd do but are too lazy to do yourself. Or even toss it away entirely and start fresh with better understanding. Yourself or again with agent.
You know where this is going. I asked Claude if audio plugins were well represented in its training data, it said yes, off I went. I can’t review the code because I lack the expertise. It’s all C++ with a lot of math and the only math I’ve needed since college is addition and calculating percentages. However, I can have intelligent discussions about design and architecture and music UX. That’s been enough to get me a functional plugin that already does more in some respects than the original. I am (we are?) making it steadily more performant. It has only crashed twice and each time I just pasted the dump into Claude and it fixed the root cause.
Long story short: if you can verify the outcome, do you need to review the code? It helps that no one dies or gets underpaid if my audio plugin crashes. But still, you can’t tell me this isn’t remarkable. I think it’s clear there will be a massive proliferation of niche software.
In other words you can’t vibe code in an environment where evaluating “does this code work” is an existential question. This is the case where 7k LOC/day becomes terrifying.
Until we get much better at automatically proving correctness of programs we will need review.
This is the game changer for me: I don’t have to evaluate tens or hundreds of market options that fit my problem. I tell the machine to solve it, and if it works, then I’m happy. If it doesn’t I throw it away. All in a few minutes and for a few cents. Code is going the way of the disposable diaper, and, if you ever washed a cloth diaper you will know, that’s a good thing.
What happens when it seems to work, and you walk away happy, but discover three months later that your circular components don't line up because the LLM-written CAD software used an over-rounded PI = 3.14? I don't work in industrial design, but I faced a somewhat similar issue where an LLM-written component looked fine to everyone until final integration forced us to rewrite it almost entirely.
If you don't know how to analyze data, and flat out refuse to invest in learning the skill, then I guess that could be really useful. Those users are likely the ones most enthusiastic about AI. But are those users close to as productive as someone who learns a mature tool? Not even close.
Lots of people appreciate an LLM to generate boiler plate code and establish frameworks for their data structures. But that's code that probably shouldn't be there in the first place. Vibe coding a game can be done impressively quick, but have you tried using a game construction kit? That's much faster still.
I think the throwaway part is important here and people are missing it, particularly for non-programmers.
There's a lot of roles in the business world that would make great use of ephemeral little apps like this to do a specific task, then throw it away. Usually just running locally on someone's machine, or at most shared with a couple other folks in your department.
Code doesn't have to be good, hell it doesn't even have to be secure, and certainly doesn't need to look pretty. It just needs to work.
There's not enough engineering staff or time to turn every manager's pet excel sheet project into a temporary app, so LLMs make perfect sense here.
I'd go as far to say more effort should be put into ephemeral apps as a use case for LLMs over focusing on trying to use them in areas where a more permanent, high quality solution is needed.
Improve them for non-developers.
It's infinitely worse when your PM / manager vibe-codes some disgusting garbage, sees that it kind of looks like a real thing that solves about half of the requirements (badly) and demands engineers ship that and "fix the few remaining bugs later".
And its tricky because I'm trying not to appeal to emotion despite being fascinated with how this tool has enabled me to do things in a short amount of time that it would have taken me weeks of grinding to get to and improves my communication with stakeholders. That feels world changing. Specifically my world and the day-to-day roll I play when it comes to getting things done.
I think it is fine that it fell short of your expectations. It often does for me as well but it's when it gets me 80% of the way there in less than a day's work, then my mind is blown. It's an imperfect tool and I'm sorry for saying this but so are we. Treat its imperfections in the same way you would with a JR developer- feedback, reframing, restrictions, and iterate.
Well… That's no longer true, is it?
My partner (IT analyst) works for a company owned by a multinational big corporation, and she got told during a meeting with her manager that use of AI is going to become mandatory next year. That's going to be a thing across the board.
And have you called a large company for any reason lately? Could be your telco provider, your bank, public transport company, whatever. You call them, because online contact means haggling with an AI chatbot first to finally give up and shunt you over to an actual person who can help, and contact forms and e-mail have been killed off. Calling is not exactly as bad, but step one nowadays is 'please describe what you're calling for', where some LLM will try to parse that, fail miserably, and then shunt you to an actual person.
AI is already unavoidable.
My multinational big corporation employer has reporting about how much each employee uses AI, with a naughty list of employees who aren't meeting their quota of AI usage.
The fact that companies have to force you to use it with quotas and threats is damning.
It's mostly a sign leadership has lost reasoning capability if it's mandatory.
But no, reporting isn't necessarily the problem. There are plenty of places that use reporting to drive a conversation on what's broken, and why it's broken for their workflow, and then use that to drive improvement.
It's only a problem if the leadership stance is "Haha! We found underpants gnome step 2! Make underpants number go up, and we are geniuses". Sadly not as rare as one would hope, but still stupid.
All of this predates LLMs (what “AI” means today) becoming a useful product. All of this happened already with previous generations of “AI”.
It was just even shittier than the version we have today.
This is what I always think of when I imagine how AI will change the world and daily life. Automation doesn't have to be better (for the customer, for the person using it, for society) in order to push out the alternatives. If the automation is cheap enough, it can be worse for everyone, and still change everything. Those are the niches in ehich I'm most certain will be here to stay— because sometimes, it hardly matters if it's any good.
If you're lucky. I've had LLMs that just repeatedly hang up on me when they obviously hit a dead end.
AI's not exactly a step down from that.
I'd argue that's not true. It's more of a stated goal. The actual goal is to achieve the desired outcome in a way that has manageable, understood side effects, and that can be maintained and built upon over time by all capable team members.
The difference between what business folks see as the "output" of software developers (code) and what (good) software developers actually deliver over time is significant. AI can definitely do the former. The latter is less clear. This is one of the fundamental disconnects in discussions about AI in software development.
I'm going to say this next thing as someone with a lot of negative bias about corporations. I was laid off from Twitter when Elon bought the company and at a second company that was hemorrhaging users.
Our job isn't to write code, it's to make the machine do the thing. All the effort for clean, manageable, etc is purely in the interest of the programmer but at the end of the day, launching the feature that pulls in money is the point.
If everyone on your team is doing that, it's not long before huge chunks of your codebase are conceptually like stuff that was written a long time ago by people who left the company. Except those people may have actually known what they were doing. The AI chatbots are generating stuff that seems to plausibly work well enough based on however they were prompted.
There are intangible parts of software development that are difficult to measure but incredibly valuable beyond the code itself.
> Our job isn't to write code, it's to make the machine do the thing. All the effort for clean, manageable, etc is purely in the interest of the programmer but at the end of the day, launching the feature that pulls in money is the point.
This could be the vibe coder mantra. And it's true on day one. Once you've got reasonably complex software being maintained by one or more teams of developers who all need to be able to fix bugs and add features without breaking things, it's not quite as simple as "make the machine do the thing."
Maybe I'm not understanding you're point, but this is the kind of thing that happens in software teams all the time and is one of those "that's why they call it work" realities of the job.
If something "seems right/passed review/fell apart" then that's the reviewer's fault right? Which happens, all the time! Reviewers tend to fall back to tropes and "is there tests ok great" and whatever their hobbyhorses tend to be, ignoring others. It's ok because "at least it's getting reviewed" and the sausage gets made.
If AI slashed the amount of time to get a solution past review, it buys you time to retroactively fix too, and a good attitude when you tell it that PR 1234 is why we're in this mess.
I mean this in sincerity, and not at all snarky, but - have you considered that you haven't used the tools correctly or effectively? I find that I can get what I need from chatbots (and refuse to call them AI until we have general AI just to be contrary) if I spend a couple of minutes considering constraints and being careful with my prompt language.
When I've come across people in my real life who say they get no value from chatbots, it's because they're asking poorly formed questions, or haven't thought through the problem entirely. Working with chatbots is like working with a very bright lab puppy. They're willing to do whatever you want, but they'll definitely piss on the floor unless you tell them not to.
Or am I entirely off base with your experience?
I prefer to use LLM as a sock puppet to filter out implausible options in my problem space and to help me recall how to do boilerplate things. Like you, I think, I also tend to write multi-paragraph prompts repeating myself and calling back on every aspect to continuously hone in on the true subject I am interested in.
I don't trust LLM's enough to operate on my behalf agentically yet. And, LLM is uncreative and hallucinatory as heck whenever it strays into novel territory, which makes it a dangerous tool.
The problem is that this comes off just as tone-deaf as "you're holding it wrong." In my experience, when people promote AI, its sold as just having a regular conversation and then the AI does thing. And when that doesn't work, the promoter goes into system prompts, MCP, agent files, etc and entire workflows that are required to get it to do the correct thing. It ends up feeling like you're being lied to, even if there's some benefit out there.
There's also the fact that all programming workflows are not the same. I've found some areas where AI works well, but a lot of my work it does not. Usually things that wouldn't show up in a simple Google search back before it was enshittified are pretty spotty.
Then there’s people like me, who you’d probably term as an old soul, who looks at all that and says, “I have to change my workflow, my environment, and babysit it? It is faster to simply just do the work.” My relationship with tech is I like using as little as possible, and what I use needs to be predictable and do something for me. AI doesn’t always work for me.
This is almost the complete opposite of my experience. I hear expressions about improvements and optimism for the future, but almost all of the discussion from active people productivly using AI is about identifying the limits and seeing what benefits you can find within those limits.
They are not useless and they are also not a panacea. It feels like a lot of people consider those the only available options.
It can't reason from first principles and there isn't training data for a lot of state-of-the-art computer science and code implementations. Nothing you can prompt will make it produce non-naive output because it doesn't have that capability.
AI works for a lot of things because, if we are honest, AI generated slop is replacing human generated slop. But not all software is slop and there are software domains where slop is not even an option.
More precisely:
In one side, it's the "tools that build up critical mass" philosophy. AI firmly resides here.
On the other, it's the "all you need is brain and plain text" philosophy. We don't see much AI in this camp.
One thing I learned is that you should never underestimate the "all you need is brain and plain text" camp. That philosophy survived many, many "fatal blows" and has come up on top several times. It has one unique feature: resilience to bloat, something that the current smart tools camp is obviously overlooking.
I think the feeling stems from the exaggeration of the value it provides combined with a large number of internal corporate LLMs being absolute trash.
The overvaluation is seen in effect everywhere from the stock market, the price of RAM, the cost of energy as well as IP theft issues etc etc. AI has taken over and yet it still feels like just a really good fuzzy search. Like yeah I can search something 10x faster than before but might get a bad answer every now and then.
Yeah its been useful (so have many other things). No it's not worth building trillion dollar data centers for. I would be happier if the spend went towards manufacturing or semiconductor fabs.
It’s because they know it works better every day and the people controlling it are gleefully fucking over the rest of the world because they can.
The plainly stated goal is TO ELIMINATE ALL HUMAN EMPLOYEES, with no plan for how those people will feed, clothe, or house themselves.
The reactions the author was getting was the reaction of a horse talking to someone happily working for the glue factory.
... but maybe not in the way that these CEOs had hoped.[0]
Part of the AI fatigue is that busy, competent devs are getting swarmed with massive amounts of slop from not-very-good developers. Or product managers getting 5 paragraphs of GenAI bug reports instead of a clear and concise explanation.
I have high hopes for AI and think generative tooling is extremely useful in the right hands. But it is extremely concerning that AI is allowing some of the worst, least competent people to generate an order of magnitude more "content" with little awareness of how bad it is.
I don't recognize that because it isn't true. I try the LLMs every now and then, and they still make the same stupid hallucinations that ChatGPT did on day 1. AI hype proponents love to make claims that the tech has improved a ton, but based on my experience trying to use it those claims are completely baseless.
One of the tests I sometimes do of LLMs is a geometry puzzle:
You're on the equator facing south. You move forward 10,000 km along the surface of the Earth. You are rotate 90° clockwise. You move another 10,000 km forward along the surface of the earth. Rotate another 90° clockwise, then move another 10,000 km forward along the surface of the Earth.
Where are you now, and what direction are you facing?
Anecdotes are of course a bad way to study this kind of thing.
Unfortunately, so are the benchmarks, because the models have quickly saturated most of them, including traditional IQ tests (on the plus side, this has demonstrated that IQ tests are definitely a learnable skill, as LLMs loose 40-50 IQ points when going from public to private IQ tests) and stuff like the maths olympiad.
Right now, AFAICT the only open benchmarks are the METR time horizon metric, the ARC-AGI family of tests, and the "make me an SVG of ${…}" stuff inspired by Simon Willison's pelican on a bike.
FWIW, Claude Opus 4.5 gets this right for me, assuming that is the intended answer. On request, it also gave me a Mathematica program which (after I fixed some trivial exceptions due to errors in units) informs me that using the ITRF00 datum the actual answer is 0.0177593 degrees north and 0.168379 west of where you started (about 11.7 miles away from the starting point) and your rotation is 89.98 degrees rather than 90.
(ChatGPT 5.1 Thinking, for me, get the wrong answer because it correctly gets near the South Pole and then follows a line of latitude 200 times round the South Pole for the second leg, which strikes me as a flatly incorrect interpretation of the words "move forward along the surface of the earth". Was that the "usual error they all used to make"?)
How does that bridge get built? I can provide tangible real life examples but I've found push back from that in other online conversations.
To my mind, the 'only just started' argument is wearing off. It's software, it moves fast anyway, and all the giants of the tech world have been feverishly throwing money at AI for the last couple of years. I don't buy that we're still just at the beginning of some huge exponential improvement.
no, they are currently losing money on inference too
Simple. The company providing the tool needs actual earning suddenly. Therefore, they need to raise the prices. They also need users to spend more tokens, so they will make the tool respond in a way that requires more refinement. After all, the latter is exactly what happened with google search.
At this point, that is pretty normal software cycle - try to attract crowd by being free or cheap, then lock features behind paywall. Then simultaneously raise prices more and more while making the product worst.
This literally NEEDS to happen, because these companies do not have any other path to profitability. So, it will happen at some point.
It’s going to definitely be crappy, remember Google in 2003 with relevant results and no endless SEO , or Amazon reviews being reliable, or Uber being simple and cheap, etc. once growth phase ends monetization begins and experience declines but this is guard railed by the fact that there are many players.
This will cause bankruptcies and huge job losses. The argument for and against AI doesn't really matter in the end, because the finances don't make a lick of sense.
With AI I can... generate slop. Sometimes that is helpful, but it isn't yet at the point where it's replacing anything for me aside from making google searches take a bit less time on things that I don't need a definitive answer for.
It's popping up in my music streams now and then, and I generally hate it. Mushy-mouthed fake vocals over fake instruments. It pops up online and aside from the occasional meme I hate it there too. It pops up all over blogs and emails and I profoundly hate it there, given that it encourages the actual author to silence themselves and replaces their thoughts with bland drivel.
Every single software product I use begs me to use their AI integration, and instead of "no" I'm given the option of "not now", despite me not needing it, and so I'm constantly being pestered about it by something.
It has, thus far, made nearly everything worse.
I think this is probably the disconnect, this seems so wildly different from my experience. Not only that, I’ll grant that there are a ton of limitations still but surely you’d concede that there has been an incredible amount of progress in a very short time? Like I can’t imagine someone who sits down with Claude like I do and gets up and says “this is crap and a fad and won’t go anywhere”.
As for generated content, I again agree with you and you’d be surprised to learn that _execs_ agree with you but look at models from 1, 2, 3 years ago and tell me you don’t see a frightening progression of quality. If you want to say “I’ll believe it when I see it” that’s fine but my god just look at the trajectory.
For AI slop text, once again agree, once again I think we all have to figure out how to use it, but it is great for e.g. helping me rewrite a wordy message quickly, making a paper or a doc more readable, combining my notes into something polished, etc, and it’s getting better and better and better.
So I disagree it has made everything worse but I definitely agree that it has made a lot of things worse and we have a lot of Pets.com ideas that are totally not viable today, but the point I think people are maybe missing (?) is that it’s not about where we are it’s about the velocity and the future. You may be terrified and nauseated by $1T in capex on AI infra, fine but what that tells you is the scale is going to grow even further _in addition_ to the methodological / algorithmic improvements to tackle things like continual learning, robustness, higher quality multimodal generation with e.g. true narrative consistency, etc etc etc. in 5 years I don’t think many people will think of “slop” so negatively
A similar thing played out a bit with IoT and voice controlled systems like Alexa. They've got their places, but nobody needs or wants the Amazon Dash buttons, or for Alexa to do your shopping for you.
Setting an alarm or adding a note to a list is fine, remote monitoring is fine, but when it comes to things that really matter like spending money autonomously, it completely falls flat.
Long story short, I see a fad that will fall into the background of what people actually do, rather than becoming the medium that they do it by.
In e.g. the US, it's a huge net negative because kids aren't probably taught these values and the required discipline. So the overwhelming majority does use it to cheat the learning process.
I can't tell you if this is the same inside e.g. China. I'm fairly sure it's not nearly as bad though as kids there derive much less benefit from cheating on homework/the learning process, as they're more singularly judged on standardized tests where AI is not available.
It's a sad commentary on the state of search results and the Internet now that ChatGPT is superior, particularly since pre-knowledge-panel/AI-overview Google was superior in several ways (not hallucinating, for one, and being able to triangulate multiple sources to tell the truth).
But the shockwave will cause a huge recession and all those investors that put up trillions will not take their losses. Rich people never get poorer. One way or another us consumers will end up paying for their mistakes. Either by huge inflation, job losses, energy costs, service enshittification whatever. We're already seeing the memory crisis having huge knock on effects with next year's phones being much more expensive. That's one of the ways we are going to be paying for this circus.
I really see value in it too, sure. But the amount of investment that goes into it is insane. It's not that valuable by far. LLMs are not good for everything and the next big thing is still a big question mark. AI is dragged in by the hair into usecases where it doesn't belong. The same shit we saw with blockchains, but now on a world crashing scale. It's very scary seeing so much insanity.
But anyway whatever I think doesn't matter. Whatever happens will happen.
Yup. Ai is so fickle it’ll do anything to accomplish the task. But ai is just a tool it’s all about what you allow it to do. Can’t blame ai really.
These were (formerly) not the kinds of humans who regularly made these kinds of mistakes.
Those slops already existed, but AI scales them by an order of magnitude.
I guess the same can be said of any technology, but AI is just a more powerful tool overall. Using languages as an example - lets say duck typing allowed a 10% productivity boost, but also introduced 5% more mistakes/problems. AI (claims to) allow a 10x productivity boost, but also ~10x mistakes/problems.
I see how the use of AI is useful, but I feel that the practitioners of AI-as-coding-agent are running away from the real work. How can you tell me about the system you say you have created if you don't have the patience to make it or think about it deeply in the first place?
You don’t get to fix bugs in code by simply pouring it through a filter.
-206dev
This article assumes that AI is the centre of the universe, failing to understand that that assumption is exactly what's causing the attitude they're pointing to.
There's a dichotomy in the software world between real products (which have customers and use cases and make money by giving people things they need) and hype products (which exist to get investors excited, so they'll fork over more money). This isn't a strict dichotomy; often companies with real products will mix in tidbits of hype, such as Microsoft's "pivot to AI" which is discussed in the article. But moving toward one pole moves you away from the other.
I think many engineers want to stay as far from hype-driven tech as they can. LLMs are a more substantive technology than blockchain ever was, but like blockchain, their potential has been greatly overstated. I'd rather spend my time delivering value to customers than performing "big potential" to investors.
So, no. I don't think "engineers don't try because they think they can't." I think engineers KNOW they CAN and resent being asked to look pretty and do nothing of value.
A lot of us tried it and just said, "huh, that's interesting" and then went back to work. We hear AI advocates say that their workflow is amazing, but we watch videos of their workflow, and it doesn't look that great. We hear AI advocates say "the next release is about to change everything!", but this knowledge isn't actionable or even accurate.
There's just not much value in chasing the endless AI news cycle, constantly believing that I'll fall behind if I don't read the latest details of Gemini 3.1 and ChatGPT 6.Y (Game Of The Year Edition). The engineers I know who use AI don't seem to have any particular insights about it aside from an encyclopedic knowledge of product details, all of which are changing on a monthly basis anyway.
New products that use gen AI are — by default — uninteresting to me because I know that under the hood, they're just sending text and getting text back, and the thing they're sending to is the same thing that everyone is sending to. Sure, the wrapper is nice, but I'm not paying an overhead fee for that.
"Engineers don't try" doesn’t refer to trying out AI in the article. It refers to trying to do something constructive and useful outside the usual corporate churn, but having given up on that because management is single-mindedly focused on AI.
One way to summarize the article is: The AI engineers are doing hype-driven AI stuff, and the other engineers have lost all ambition for anything else, because AI is the only thing that gets attention and helps the career; and they hate it.
Worse, they've lost all funding for anything else.
Then things don't turn out as they expected and you have to deal with a dude thinking his engineers are messing with him.
It's just boring.
But now, to your point: they can vibe-code their own "mockups" and that brings us back to that problem
There are many valid critiques of AI, but “there’s not much there” isn’t one of them.
To me, any software engineer who tries an LLM, shrugs and says “huh, that’s interesting” and then “gets back to work” is completely failing at their actual job, which is using technology to solve problems. Maybe AI isn’t the right tool for the job, but that kind of shallow dismissal indicates a closed mind, or perhaps a fear-based reaction. Either way, the market is going to punish them accordingly.
"There's not much there" is a totally valid critique of a lot of the current AI ecosystem. How many startups are simple prompt wrappers on top of ChatGPT? How many AI features in products are just "click here to ask Rovo/Dingo/Kingo/CutesyAnthropomorphizedNameOfAI" text boxes that end up spitting out wrong information?
There's certainly potential but a lot of the market is hot air right now.
> Either way, the market is going to punish them accordingly.
I doubt this, simply because the market has never really punished people for being less efficient at their jobs, especially software development. If it did, people proficient in vim would have been getting paid more than anyone else for the past 40 years.
The skeptics are the ones that have tried AI coding agents and come away unimpressed because it can’t do what they do. If you’re proudly proclaiming that AI can replace your work then you’re telling on yourself.
In fact, it tends to be the opposite. You being more efficient just means you get "rewarded" with more work, typically without an appropriate increase in pay to match the additional work either.
Especially true in large, non-tech companies/bureaucratic enterprises where you are much better off not making waves, and being deliberately mediocre (assuming you're not a ladder climber and aren't trying to get promoted out of an IC role).
In a big team/org, your personal efficiency is irrelevant. The work can only move as fast as the slowest part of the system.
I've been around long enough that I have seen four hype cycles around AI like coding environments. If you think this is new you should have been there in the 80's (Mimer, anybody?), when the 'fourth generation' languages were going to solve all of our coding problems. Or in the 60's (which I did not personally witness on account of being a toddler), when COBOL, the language for managers was all the rage.
In between there was LISP, the AI language (and a couple of others).
I've done a bit more than looking at this and saying 'huh, that's interesting'. It is interesting. It is mostly interesting in the same way that when you hand an expert a very sharp tool they can probably carve wood better than with a blunt one. But that's not what is happening. Experts are already pretty productive and they might be a little bit more productive but the AI has it's own envelope of expertise and the closer you are to the top of the field the smaller your returns in that particular setting will be.
In the hands of a beginner there will be blood all over the workshop and it will take an expert to sort it all out again, quite possibly resulting in a net negative ROI.
Where I do get use out of it: to quickly look up some verifiable fact, to tell me what a particular acronym stands for in some context, to be slightly more functional than wikipedia for a quick overview of some subfield (but you better check that for gross errors). So yes, it is useful. But it is not so useful that competent engineers that are not using AI are failing at their job, and it is at best - for me - a very mild accelerator in some use cases. I've seen enough AI driven coding projects strand hopelessly by now to know that there are downsides to that golden acorn that you are seeing.
The few times that I challenged the likes of ChatGPT with an actual engineering problem to which I already knew the answer by way of verification the answers were so laughably incorrect that it was embarrassing.
And for the better. I've honestly not had this much fun programming applications (as opposed to students stuff and inner loops) in years.
I'm happy that it works out for you, and probably this is a reflection of the kind of work that I do, I wouldn't know how to begin to solve a problem like designing a braille wheel or a windmill using AI tools even though there is plenty of coding along the way. Maybe I could use it to make me faster at using OpenSCAD but I am never limited by my typing speed, much more so by thinking about what it is that I actually want to make.
I would argue that the "actual job" is simply to solve problems. The client / customer ultimately do not care what technology you use. Hell, they don't really care if there's technology at all.
And a lot of software engineers have found that using an LLM doesn't actually help solve problems, or the problems it does solve are offset by the new problems it creates.
Or, and stay with me on this, it’s a reaction to the actual experience they had.
I’ve experimented with AI a bunch. When I’m doing something utterly formulaic it delivers (straightforward CRUD type stuff, or making a web page to display some data). But when I try to use it with the core parts of my job that actually require my specialist knowledge they fall apart. I spend more time correcting them than if I just write it myself.
Maybe you haven’t had that experience with work you do. But I have, and others have. So please don’t dismiss our reaction as “fear based” or whatever.
AI is terrible at anything it hasn’t seen 1000 times before on GitHub. It’s bad at complex algorithmic work. Ask it to implement an order statistic tree with internal run length encoding and it will barely be able to get off the starting line. And if it does, the code will be so broken that it’s faster to start from scratch. It’s bad at writing rust. ChatGPT just can’t get its head around lifetimes. It can’t deal with really big projects - there’s just not enough context. And its code is always a bit amateurish. I have 10+ years of experience in JS/TS. It writes code like someone with about 6-24 months experience in the language. For anything more complex than a react component, I just wouldn’t ship what it writes.
I use it sometimes. You clearly use it a lot. For some jobs it adds a lot of value. For others it’s worse than useless. If some people think it’s a waste of time for them, it’s possible they haven’t really tried it. It’s also possible their job is a bit different from your job and it doesn’t help them.
I don't understand why people seem so impatient about AI adoption.
AI is the future, but many AI products aren't fully mature yet. That lack of maturity is probably what is dampening the adoption curve. To unseat incumbent tools and practices you either need to do so seamlessly OR be 5-10x better (Only true for a subset of tasks). In areas where either of these cases apply, you'll see some really impressive AI adoption. In areas where AI's value requires more effort, you'll see far less adoption. This seems perfectly natural to me and isn't some conspiracy - AI needs to be a better product and good products take time.
I have solved more problems with tools like sed and awk, you know, actual tools, more than I’ve entered tokens into an LLM.
Nobody seemed to give a fuck as long as the problem was solved.
This it getting out of hand.
I personally use it, I find it helpful at times, but I also find that it gets in my way, so much so it can be a hindrance (think losing a day or so because it's taken a wrong turn and you have to undo everything)
FTR The market is currently punishing people that DO use it (CVs are routinely being dumped at the merest hint of AI being used in its construction/presentation, interviewers dumping anyone that they think is using AI for "help", code reviewers dumping any take home assignments that have even COMMENTS massaged by AI)
There's a lot of disconnected-from-reality hustling (a.k.a lying) going on. For instance, that's practically Elon Musk's entire job, when he's actually doing it. A lot of people see those examples, think it's normal, and emulate it. There are a lot of unearned superlatives getting thrown around automatically to describe tech.
If you haven’t had a mind blown moment with AI yet, you aren’t doing it right or are anchoring in what you know vs discovering new tech.
I’m not making any case for anything, but it’s just not that hard to get excited for something that sure does seem like magic sometimes.
Edit: lol this forum :)
I AM very impressed, and I DO use it and enjoy the results.
The problem is the inconsistency. When it works it works great, but it is very noticeable that it is just a machine from how it behaves.
Again, I am VERY impressed by what was achieved. I even enjoy Google AI summaries to some of the questions I now enter instead of search terms. This is definitely a huge step up in tier compared to pre-AI.
But I'm already done getting used to what is possible now. Changes after that have been incremental, nice to have and I take them. I found a place for the tool, but if it wanted to match the hype another equally large step in actual intelligence is necessary, for the tool to truly be able to replace humans.
So, I think the reason you don't see more glowing reviews and praise is that the technical people have found out what it can do and can't, and are already using it where appropriate. It's just a tool though. One that has to be watched over when you use it, requiring attention. And it does not learn - I can teach a newbie and they will learn and improve, I can only tweak the AI with prompts, with varying success.
I think that by now I have developed a pretty good feel for what is possible. Changing my entire workflow to using it is simply not useful.
I am actually one of those not enjoying coding as such, but wanting "solutions", probably also because I now work for an IT-using normal company, not for one making an IT product, and my focus most days is on actually accomplishing business tasks.
I do enjoy being able to do some higher level descriptions and getting code for stuff without having to take care of all the gritty details. But this functionality is rudimentary. It IS a huge step, but still not nearly good enough to really be able to reliably delegate to the AI to the degree I want.
In the end you can save like 90% of the development effort on a small one-off project, and like 5% of the development effort on a large complex one.
I think too many managers have been absolutely blown away by canned AI demos and toy projects and have not been properly disappointed when attempting to use the tools on something that is not trivial.
The first 90 percent of the code accounts for the first 90 percent of the development time. The remaining 10 percent of the code accounts for the other 90 percent of the development time.
It feels like a gigantic win when it carves through that first 90%… like, “wow, I’m almost done and I just started!”. And it is a genuine win! But for me it’s dramatically less useful after that. The things that trip up experienced developers really trip up LLMs and sometimes trying to break the task down into teeny weeny pieces and cajole it into doing the thing is worse than not having it.
So great with the backhoe tasks but mediocre-to-counterproductive with the shovel tasks. I have a feeling a lot of the impressiveness depends on which kind of tasks take up most of your dev time.
When you work with a large codebase which have a very high complexity level, then the bugs put in there by AI will not worth the cost of the easily added features.
I know I don't. I have never been paid to write anything beyond a short script.
I actually can't even picture what a professional software engineer actually works on day to day.
From my perspective, it is completely mind blowing to write my own audio synth in python with Librosa. A library I didn't know existed before LLMs and now I have a full blown audio mangling tool that I would have never been able to figure out on my own.
It seems to me professional software engineering must be at least as different to vibe coding as my audio noodlings are to being a professional concert pianist. Both are audio and music related but really two different activities entirely.
Or your job isn't what AI is good at?
AI seems really good at greenfield projects in well known languages or adding features.
It's been pretty awful, IME, at working with less well-known languages, or deep troubleshooting/tweaking of complex codebases.
This is precisely my experience.
Having the AI work on a large mono repo with a front-end that uses a fairly obscure templating system? Not great.
Spinning up a greenfield React/Vite/ShadCN proof-of-concept for a sales demo? Magic.
Well, there’s your problem. You should have selected React while you had the chance.
There are some exceptions where AI is genuinely useful, but I have employees who try to use AI all the time for everything and their work is embarrassingly bad.
Yes, this is better phrased.
Results are stochastic. Some people the first time they use it will get the best possible results by chance. They will attribute their good outcome to their skill in using the thing. Others will try it and will get the worst possible response, and they will attribute their bad outcome to the machine being terrible. Either way, whether it's amazing or terrible is kind of an illusion. It's both.
Much of this boils down to people simply not understanding what’s really happening. Most people, including most software developers, don’t have the ability to understand these tools, their implications, or how they relate to their own intelligence.
> Edit: lol this forum :)
Indeed.
This takes all the joy away, even traditional maintenance projects of big corps seems attractive nowadays.
PC, Web and Smartphone hype was based on "we can now do [thing] never done before".
This time out it feels more like "we can do existing [thing], but reduce the cost of doing it by not employing people"
It all feels much more like a wealth grab for the corporations than a promise of improving a standard of living for end customers. Much closer to a Cloud or Server (replacing Mainframes) cycle.
I was doing RPA (robotic process automation) 8 years ago. Nobody wanted it in their departments. Whenever we would do presentations, we were told to never, ever, ever talk about this technology replacing people - it only removes the mundane work so teams can focus more on the bigger scope stuff. In the end, we did dozens and dozens of presentations and only two teams asked us to do some automation work for them.
The other leaders had no desire to use this technology because they were not only fearful of it replacing people on their teams, they were fearful it would impact their budgets negatively so they just quietly turned us down.
Unfortunately, you're right because as soon as this stuff gets automated and you find out 1/3rd of your team is doing those mundane tasks, you learn very quickly you can indeed remove those people since there won't be enough "big" initiatives to keep everybody busy enough.
The caveat was even on some of the biggest automations we did, you still needed a subset of people on the team you were working with to make sure the automations were running correctly and not breaking down. And when they did crash, since a lot of these were moving time sensitive data, it was like someone just stole the crown jewels and suddenly you need two war rooms and now you're ordering in for lunch.
Or hiring a mathematician to calculate what is now done in a spreadsheet.
"You should be using AI in your day to day job or you won't get promoted" is the 2025 equivalent of being forced to train the team that your job is being outsourced to.
I think there is a broader dichotomy between the people-persuation-plane, and the real-world-facts plane. In the people-persuation plane, it is all about convincing someone of something, and hype plays here, and marketing, religion and political persuation too. In the real world plane, it is all about tangible outcomes, and working code or results play here, and gravity and electromagnetism too. Sometimes there is a reflex loop between the two. I chose the engineering career because, what i produce is tangible, but I realize that a lot of my work is in the people-plane.
This right here is the real thing which AI is deployed to upset.
The Enlightenment values which brought us the Industrial Revolution imply that the disparity between the people-persuasion-plane and the real-world-facts-plane should naturally decrease.
The implicit expectation here is that as civilization as a whole learns more about how the universe works, people would naturally become more rational, and thus more persuadable by reality-compliant arguments and less persuadable by reality-denying ones.
That's... not really what I've been seeing. That's not really what most of us have been seeing. Like, devastatingly not so.
My guess is that something became... saturated? I'd place it sometime around the 1970s, same time Bretton Woods ended, and the productivity/wages gap began to grow. Something pertaining to the shared-culture-plane. Maybe there's only so much "informed" people can become before some sort of phase shift occurs and the driving force behind decisions becomes some vague, ethically unaccountable ingroup intuition ("vibes", yo), rather than the kind of explicit, systematic reasoning which actually is available to any human, except for the weird fact how nobody seems to trust it very much any more.
My assumption detector twigged at that line. I think this is just replacing the dichotomy with a continuum between two states. But the hype proponents always hope - and in some cases they are right - that those two poles overlap. People make and lose fortunes on placing those bets and you don't necessarily have to be right or wrong in an absolute sense, just long enough that someone else will take over your load and hopefully at a higher valuation.
Engineers are not usually the ones placing the bets, which is why they're trying to stay away from hype driven tech (to them it is neutral with respect to the outcome but in case of a failure they lose their job, so better to work on things that are not hyped, it is simply safer). But as soon as engineers are placing bets they are just as irrational as every other class of investor.
I understood “they think they can’t” to refer to the engineers thinking that management won’t allow them to, not to a lack of confidence in their own abilities.
I worked building tools within the Microsoft ecosystem, both on the SQL Server side, and on the .NET and developer tooling side, and I spent some time working with the NTVS team at Microsoft many years ago, as well as attending plenty of Microsoft conferences and events, working with VSIP contacts, etc. I also know plenty of people who've worked at or partnered with Microsoft.
And to me this all reads like classic Microsoft. I mean, the article even says it: whatever you're doing, it needs to align with whatever the current key strategic priority is. Today that priority is AI, 12 years ago it was Azure, and on and on. And, yes, I'd imagine having to align everything you do to a single priority regardless of how natural that alignment is (or not) gets pretty exhausting, and I'd bet it's pretty easy to burn out on it if you're in an area of the business where this is more of a drag and doesn't seem like it delivers a lot of value. And you'll have to dogfood everything (another longtime Microsoft pattern) core to that priority even if it's crap compared with whatever else might be out there.
But I don't think it's new: it's simply part and parcel of working at Microsoft. And the thing is, as a strategy it's often served them well: Windows[0], Xbox, SQL Server, Visual Studio, Azure, Sharepoint, Office, etc. Doesn't always work, of course: Windows Phone went really badly, but it's striking that this kind of swing and a miss is relatively rare in Microsoft's history.
And so now, of course, they're doing it with AI. And, of course, they're a massive company, so there will be plenty of people there who really aren't having a good time with it. But, although it's far from a foregone conclusion, it would not be a surprise for Microsoft to come from behind and win by repeating their usual strategy... again.
[0] Don't overread this: I'm not necessarily saying I'm a huge fan. In fact I do think Windows, at is core, is a decent operating system, and has been for a very long time. On the back end it works well, and I have no complaints. But I viscerally despise Windows 11 as a desktop operating system. That's right: DESPISE. VISCERALLY. AT A MOLECULAR LEVEL.
One interesting take away from the article and the discussion is that there seem to be two kinds of engineers: those that buy into the hype and call it "AI," and those that see it for the fancy search engine it is and call it an "LLM." I'm pretty sure these days when someone mentions "AI" to me I roll my eyes. But if they say, "LLM," ok, let's have a discussion.
The wealthiest person in the world relies entirely on his ability to convince people to accept hype that surpasses all reason.
Spot. Fucking. On.
Thank you.
But despite all that, for writing, refactoring, and debugging computer code, LLM agents are still completely game changing. All of these things are true at the same time. There's no way someone that works with real code all day could spent an honest few weeks with a tool like Claude and come away calling it "hype". someone might still not prefer it, or it's not for them, but to claim it's "hype", that's not possible.
so, people with experience?
In hindsight it makes sense, I’m sure every major shift has played out the same way.
It also turns out that experience can be what enables you to not waste time on trendy stuff which will never deliver on its promises. You are simply assuming that AI is a paradigm shift rather than a waste of time. Fine, but at least have the humility to acknowledge that reasonable people can disagree on this point instead of labeling everyone who disagrees with you as some out of touch fuddy-duddy.
AI is not both of these things? There are no real AI products that have real customers and make money by giving people what they need?
> LLMs are a more substantive technology than blockchain ever was, but like blockchain, their potential has been greatly overstated.
What do you view as the potential that’s been stated?
In those cases the actual "new" technology (ie, not the underlying ai necessarily) is not as substantive and novel (to me at least) as a product whose internals are not just an (existing) llm.
(And I do want to clarify that, to me personally, this tendency towards 'thin-shell' products is kind of an inherent flaw with the current state of ai. Having a very flexible llm with broad applications means that you can just put Chatgpt in a lot of stuff and have it more or less work. With the caveat that what you get is rarely a better UX than what you'd get if you'd just prompted an llm yourself.
When someone isn't using llms, in my experience you get more bespoke engineering. The results might not be better than an llm, but obviously that bespoke code is much more interesting to me as a fellow programmer)
Way better than AI jammed into every crevice for no reason.
LLMs are not an intelligence, and people who treat them as if they are infallible Oracles of wisdom are responsible for a lot of this fatigue with AI
Please don't do this, make up your own definitions.
Pretty much anything and everything that uses neural nets is AI. Just because you don't like how the definition has been since the beginning doesn't mean you get to reframe it.
In addition, if humans are not infallible oracles of wisdom, they wouldn't be an intelligence in your definition.
I also don't understand the LLM ⊄ AI people. Nobody was whining about pathfinding in video games being called AI lol. And I have to say LLMs are a lot smarter than A*.
Also it's funny how they add (supervised) everywhere. It looks like "Full self driving (not really)"
Look I don't like the advertising of FSD, or musk himself, but we without a doubt have cars using significant amounts of AI that work quite well.
Bitcoin is probably unkillable. Even if were to crash, it won't be hard to round up enough true believers to boost it up again. But it's technically stagnant.
Many other cryptocurrencies are popular enough to be easily tradable and have features to make them work better for trade. Also, you can speculate on different cryptocurrencies than your friends do.
The values of bitcoin are:
- easy access to trading for everyone, without institutional or national barriers
- high leverage to effectively easily borrow a lot of money to trade with
- new derivative products that streamline the process and make speculation easier than ever
The blockchain plays very little part in this. If anything it makes borrowing harder.
how on earth does bitcoin have anything to do with borrowing or derivatives?
in a way that wouldn't also work for beanie babies
The whole cryptocurrency world requires evangelical buy-in. But there is no directly created functional value other than a historic record of transactions and hypothetical decentralization. It doesn’t directly create value. It’s a store of it - again, assuming enough people continue to buy into the narrative so that it doesn’t dramatically deflate when you need to recover your assets. States and other investors are helping make stability happen to maintain it as a value store, but you require the story propagating to achieve those ends.
1. You were a therapy session for her. Her negativity was about the layoffs.
2. FAANG companies dramatically overhired for years and are using AI as an excuse for layoffs.
3. AI scene in Seattle is pretty good, but as with everywhere else was/is a victim of the AI hype. I see estimates of the hype being dead in a year. AI won't be dead, but throwing money at the whatever Uber-for-pets-AI-ly idea pops up won't happen.
4. I don't think people hate AI, they hate the hype.
Anyways, your app actually does sound interesting so I signed up for it.
I wonder if these feelings are what scribes and amanuenses felt when the printing press arrived.
I do enjoy programming, I like my job and take pride on it, but I actively try for it not to be the life-mean giving activity. I'm a just mercenary of my trade.
Most of it is because there's little that ties actual output to organizational outcomes. AI mandates after all are just a way to bluntly for e engineers to use AI, where if you were at a startup or smaller company you would probably organically find how much an LLM helps you where. It may not even help your actual work even if it helps your coworkers. That market feedback is sorely missing from the Big Techs and so hamfisted engineering mandates have to do in order to for e engineers to become more efficient.
In these cases I always try to remind friends that you can always leave a Big Tech. The thing is, from what I can tell, a lot of these folks have developed lifestyle inflation from working in Big Tech and some of their anger comes from feeling trapped in their Big Tech role due to this. While I understand, I'm not particularly sympathetic to this viewpoint. At the end of the day your lifestyle is in your hands.
What about the complete lack of morality some (most?) AI companies exhibit?
What about the consequences in the environment?
What about the enshitification of products?
What about the usage of water and energy?
Etc.
I feel fatigued by AI. To be more precise, this fatigue includes several factors. The first one is that a lot of people around me get excited by events in the AI world that I find distracting. These might be new FOSS library releases, news announcements from the big players, new models, new papers. As one person, I can only work on 2-3 things at a given interval in time. Ideally I would like to focus and go deep in those things. Often, I need to learn something new and that takes time, energy and focus. This constant Brownian motion of ideas gives a sense of progress and "keeping up" but, for me at least, acts as a constantly tapped brake.
Secondly, there is a sentiment that every problem has an AI solution. Why sit and think, run experiments, try to build a theoretical framework when one can just present the problem to a model. I use LLMs too but it is more satisfying, productive, insightful when one actually thinks hard and understands a topic before using LLMs.
Thirdly, I keep hearing that the "space moves fast" and "one must keep up". The fundamentals actually haven't changed that much in the last 3 years and new developments are easy to pick up. Even if they did, trying to keep up results in very shallow and broad knowledge that one can't actually use. There are a million things going on and I am completely at peace with not knowing most of them.
Lastly, there is pressure to be strategic. To guess where the tech world is going, to predict and plan, to somehow get ahead. I have no interest in that. I am confident many of us will adapt and if I can't, I'll find something else to do.
I am actually impressed with and heavily use models. The tiresome part now are some of the humans around the technology who participate in the behaviors listed above.
1) A third party app simply cannot compete with Google Maps on coverage, accuracy and being up to date. Yes, there are APIs you can use to access this, but they're expensive and limited, which leads us to the second problem:
2) You can't make money off them. Nobody will pay to use your app (because there's so much free competition), and the monetization opportunities are very limited. It's too late in the flow to sell flights, you can't compete with Booking etc for hotel search, and big ticket attractions don't pay commissions for referrals. That leaves you with referrals for tours, but people who pay for tours are not the ones trying to DIY their trip planning in the first place.
So many products are like this - it sounds good on paper to consolidate a bunch of tasks in one place but it's not without costs and the benefit is just not very high.
If they become popular they'll have to move to OSM, Google's steep charging for their Maps API at high usage that has brought companies to their knees is well known [1].
Similar to "made for everyone" social networks and video upload platforms.
But there are niches that are trip planning + there are no one solving the pain! For example Geocaching. I always dreamed about an easy way to plan Geocaching routes to travel and find interesting caches on the way. Currently you gotta filter them out and then eyeball the map what seems to be nearby, despite there, maybe, not being any real roads there, or the cache is probably maybe actually lost or has to be accessed at specific time of day.
So... No one wants apps that are already solved + boring.
> I wanted her take on Wanderfugl , the AI-powered map I've been building full-time.
I can at least give you one piece of advice. Before you decide on a company or product name, take the time to speak it out loud so you can get a sense of how it sounds.
In English, I’d pronounce it very similar to “wonderful”.
So even the creator can't decide what to call it!
Now I want to know how you pronounce words like: through, bivouac, and queue.
I personally thought it was wander _fughel_ or something.
Let alone how difficult it is to remember how to spell it and look it up on Google.
I thought ‘wanderfugl’ was a throwback to ~15 years ago when it was fashionable to use a word but leave out vowels for no reason, like Flickr/ /Tumblr/Scribd/Blendr.
Among actual people (i.e. not managers) there seems to be a bit of a generation gap - my younger friends (Gen Z) are almost disturbingly enthusiastic about entrusting their every thought and action to ChatGPT; my older friends (young millennials and up) find it odious.
Or, drive through Worth and Bridgeview in IL, where all the middle eastern people in Chicago live, and notice all the AI billboards. Not billboards for AI, just, billboards obviously made with GenAI.
I think it's just not true that non-tech people are especially opposed to AI.
One look at the dystopian billboards bragging about trying to replace humans with AI should make any sane human angry at what tech has done. Or the rising rents due to an influx of people working on mostly useless AI startups, 90% of which won't be around in 5 years. Or even how poorly many in tech behave in public and how poorly they treat service workers. That's just the tip of the iceberg, and just in SF alone.
I say all this as someone living in SF and working in tech. As a whole, we've brought the hate upon ourselves, and we deserve it.
There's maybe a slow trend towards "that's not true, you should know better than to trust AI for that sort of question" in discussions when someone says something like "I asked AI how [xyz was done]" but it's definitely not enough yet to keep anyone from going to it as their first option for answering a question.
At least, that's my wife's experience working on a contract with a state government at a big tech vendor.
It makes a lot of sense that someone casually coming in to use chatgpt for 30 minutes a week doesn't have any reason to think more deeply about what using that tool 'means' or where it came from. Honestly, they shouldn't have to think about it.
Personally, I’m in-between the opinions. I hate when I’m consuming AI-generated stuff, but can see the use for myself for work or asking bunch of not-so-important questions to get general idea of stuff.
That's the presumption. There's no data on whether this is actually true or not. Most rational examinations show that it most likely isn't. The progress of the technology is simply too slow and no exponential growth is on the horizon.
I can assure you, living in Seattle I still encounter a lot a AI boosters just as much as I encounter AI haters/skeptics
The world is not at all about fairness of benefits and impacts to all people it is about a populist mass and what amuses them and makes their life convenient, hopefully without attending the relevant funerals themselves.
Honestly I don’t really know what to say to that, other than it seems rather relevant to me. I don’t really know what to elaborate on given we disagree on such a fundamental level.
It's lovely that you care about health, but I have no idea why you think you are relevant to a society that is very much willing to risk extinction to avoid the slightest upset or delay to consumer convenience measured progress.