It's very true to me. Another effect I noticed in less tech inclined jobs (but still office jobs) is that they're still convinced "their" special job is not possible to automatize, because it's too complicated. Mind you some of them have really simple jobs, bordering simple data entry in excel. They tried chatGPT once or twice one year ago and don't realize what's coming for them. Since in tech circles we're the first ones to optimize and use new technologies we are most aware of the huge leaps that happened and the current impressive capabilities but lot are still oblivious to it.
My biggest concern isn't actually that my job can be automated (it can), but rather that the persons who need me stop valueing being right, and start valueing >being told< that they are right by an AI.
I feel my main purpose isn't that I can technically do something my clients can't, but that I know how to interpret their needs and find the solutions that fit for them.
Now they just query some AI which spits out >some< solution and they won't know that they are wrong until it can't even be tracked back to their AI.
This will result in "but the computer said" excuses all over the place.
I think we'll see an unbelievable upheaval as companies figure out that while the technical aspects of some jobs can (and should) be automated, the interpretation of those technical aspects and other higher order, but not strictly inherent processes can't. By upheaval I mean companies closing because they replace staff with AI, and lose things along the way because of it. They won't see it until it's too late, because of the pile of money they're saving.
My example comes from maintenance in machine shops. My dad's siblings were almost all employed in machining in some fashion. One uncle was maintenance for a local large company. They decided they could move maintenance from in-house specialists and contract that work to a multi-state massive company; anyone can follow a maintenance manual right? All you need are the proper certifications and everything is cool, right? The cost was less, and the number of maintenance staff on site actually went up. Great for the owners.
But what they lost was consistency and understanding in the process that eventually led to closing the company. My uncle was there every day all year long for decades and could hear and feel in his feet when something was going wrong with a machine regardless of the maintenance schedule approved by the company that manufactured the machines. The new outside company didn't pay as well, and didn't have great benefits, which meant that they cycled through staff very regularly. No one was there consistently for more than a year or two. Even though they were technically following the manufacturers recommended maintenance schedule, machines went down. Machines then stayed down for days or weeks longer than they should've, because the expertise to repair either wasn't there, or the person leading the job would quit and they would have to cycle in a new maintenance person. They lost jobs, customers took work other places. 8 years later the company closed.
There are lessons in this kind of thing, but nobody learns them.
This is a techne/metis split. You would probably like this [0] chapter about the "rightness" struggle between engineering and service at Xerox. Some paralells to A Big Little Idea called Legibility from Ribbonfarm, too.
[0] https://books.worksinprogress.co/book/maintenance-of-everyth...
Seeing Like A State is essential reading, especially for anyone working in software or operations for a corporation.
The original book that article is based on is great.
In the IT department in which I work, now the team is reduced to just me, I find myself using the phrase "Why fix a problem in 15 minutes with your own engineers, when you can take 3 weeks with a Managed-Service Provider?". Most of what I have to care for is office/internet network related.
People in other teams are slowly getting it. When I leave (or more likely "am let go"), the cold hard reality will become much more obvious during the next network problem ;)
The way this will play out will be weird, and probably not like anyone expects. Because on the one hand, you're 100% right on the value of "feel" you develop for a system when you're around it daily for a long time. But, on the other hand...
The other day I visited our old apartment and noticed the dishwasher was making loud noises. Worrying about whether it's about to fail and whether I should call in a technician, I just recorded a bit of the noise, screenshotted a spectrogram from an app, made some photos, and threw Claude at it. It proceeded to analyze the recording and match specific frequencies to components of the machine, and gave me an analysis indicating the machine is operating normally, and the noise is just the pump being louder than I'm used to due to machine being old and a bit worn out.
So we literally have that kind of capabilities in our pockets, and this will compensate for the problem of missing experience somewhat, even simply because unlike low-paid frequently-rotating staff, LLMs are not lazy and will do the detailed troubleshooting work even off a recording and "wtf?" as prompt, so if you can train your cheap, expendable staff to do just that, that's already supercharges them.
I'm also amazed at the abilities of AI when I can verify them. But, how do you know that Claude is right about your dishwasher? I mean you're not that much worse off if it's wrong (compared with a similar medical query), but isn't this a case of trusting the AI without verification, just because it looks correct and was stated confidently?
I know enough to evaluate the analysis I got; I have some general expertise and worked in fields adjacent to this, which is why I had the idea to troubleshoot a dishwasher off an audio spectrogram in the first place. And on the off chance I was wrong, this is super low risk - would cost me extra few hours of my life at a less convenient time.
My point isn't that AI can substitute for expertise in general, not just yet, but that if you're competent enough to evaluate it, AI can absolutely do hours of work in minutes even for things involving troubleshooting on factory floor (or home appliances). And, that this is a lever where personnel on the floor may have sufficient expertise to evaluate diagnostic work, but not enough motivation or space to do the diagnostic work themselves.
At that point of low risk, why not just wait till it breaks and then get a replacement?
Seems the AI analysis added no value, but took, up time...
Do you always wait for things to break completely before considering getting repair or replacement?
For my rental properties where tenants don't necessarily take good care of things, most definitely wait until the appliances break to replace them.
[dead]
I believe that the imprtatnt bit that many will come to understand ,maybe after costly mistakes is that 'tasks can be automated, jobs cannot(yet)' one still needs the discernment to know what the sequence of tasks lead up to meaningfully.
It kinda reminds me of the story of the Toyota process , while their car product lines were/are heavily automated the process still relies on the smalll teams on the line to spot/solve and iterarte over mistakes/inefficiences then add the refinements back onto the automated tasks.
Similarly althose stories of radilogists going jobless have not panned out , as interpreting slieds is a 'task' that a radiologist does.
Semantics , yeah but it does shape the way of thinkings regarding jobs/economy and not to mention the whole anthropomorphizing by the AI labs and boosters. In essence the wordings can/could introduce a healthy dose of reality/skepticism to the wildest predictions regarding job losses etc.
[dead]
The right proxy is GDP not the employment rate. The Industrial Revolution didn’t result in a decrease in employment.
This is already an issue I face both with clients and personally where I can say something authoritatively, but they’ll always go and just ask an LLM.
Sometimes it’s in my favour. LLMs estimate very high prices and long timelines for software projects.
If it’s a legal / regulatory issue it’s usually wrong.
It's still mildly amusing when I ask what kind of work is involved for a change, and it gives me some estimate of 3-6 weeks or whatever, and then I say "no I mean like what changes" and then it spits out the complete change in 30 seconds.
That has been my experience as well. One-shot time and budget estimates are always ridiculously off.
If you hook it up to legal databases like Lexus Nexus and let it ONLY search proper places for law (instead of Wikipedia) these systems are superhuman at law too. A lawyer acting as a meat proxy for Opus 5.5 will smoke a lawyer without AI systems on average these days.
Yes, well, a typical client has a free tier of ChatGPT and just voice-to-text’s questions into that.
Was anyone doubting LLms are good at searching and summarising text?
Some internal engineer is doing god's work by creating a system skill for the LLM that multiplies all estimations by five. /s
On a more serious note, the pre-LLM internet had very little useful detailed information regarding software estimation addressing effort for building something in specific domains.
Most writing on software estimation tried to tackle a general approach that could apply to all domains. So you had a "applicable to all, useful to none" situation going. And since crud apps were the most ubiquitous apps, software estimation techniques were perceived to give more accurate results with respect to those type of apps.
I'm sure internally big AI companies are accumulating vast amounts of estimation training data by documenting how long it takes their LLM to complete specific app building tasks. I don't think they would be in a rush to push that data into public models though as it would be like beating a dead horse that is the current human workforce anyway.
Your concern about higher-ups valuing being told they are correct vs actually being correct is exactly what I saw at my last job (one of the largest companies in the states) and now at another smaller firm. It was really disheartening because I do use qwen a lot for personal stuff and claude at work, and have been able to multiply my productivity by only having to understand a ticket and then code review instead of doing all the intermediate steps, but a lot of people have let it go full self driving and just tell it "implement ticket ABC-123 and raise a PR" without checking the output at all.
I wholeheartedly agree. To clarify my position, I don't think having these amazing models mean we should replace everyone, quite the opposite.
I do think they are close enough to the stage where people will be replaced, even though it's not the right or correct thing to do, and that we need to move on from the "denial" phase quickly enough so our societies can adapt / regulate / legislate while not too late.
This, 100 times over. And it's already happening.
Which isn't necessarily a bad thing. Smart leaders / entrepreneurs will realize this and then use human input where required. The rest will be cleared out after the computer has made enough mistakes for them.
>I know how to interpret their needs and find the solutions that fit for them.
Without additional explanation this is indistinguishable from the role of a technical product manager.
A capable developer will at some point become a capable people manager :)
[dead]
As a white-collar worker, it really does seem like my programmer friends are living in a different reality when it comes to AI. They have totally restructured their workflows around AI, and I'm still struggling to find the value in it for my legal and compliance consulting work. (And I consider myself pretty tech savvy, with some programming and web development knowledge and having spent 10 years at crypto startup.)
Even in my personal life, while I've been addicted to using Claude and ChatGPT as a first step for research/projects/etc, I'm still not convinced they are actually saving me time -- just too many errors, too confidently stated, and I have to re-do all the research myself.
I wonder if the validation piece is easier with coding?
EDIT: To be clear, when I say I'm having trouble incorporating AI into my white-collar work, I'm talking about producing GOOD work. On a few occasions I have given in to the temptation of using Claude to produce work that looked good (superficially), without verifying everything.....it worked out, and I'm sure others are doing it. But that scares me, because it feels like a ticking timebomb.
Sizing the steps matter. Disclaimer up front: we are all figuring it out as we go along, there are no hard and fast rules that fully generalize. You just need to try and develop your own feel for it.
In my case, for some types of work - which sound similar to things you mention - I started slowing AI down intentionally. Software development, if it's something simple/common and thus "in-distribution", LLMs can zero-shot complex projects just fine. But in more R&D work, I found that errors crop up quickly and accumulate, and end result of "go and just make X" is usually bad.
Now, the problem I found with usual workflows for this kind of work is, agents race ahead and very quickly, I can't keep up anymore. So I reworked my workflow to make AI work in small steps optimized strictly for maximizing my ability to validate the outcome, and minimize time I spend doing it. So now, the thing (say diagnosing a performance regression of some complex codebase run on a cluster) may take me a whole day, where "unchained", agentic swarm would do it in 30-60 minutes - but, at the end of the day, my method is near 100% guaranteed to be correct, with every intermediate step vouched and reusable for future work, whereas "agentic swarm" approach, that hour of work, 90% of the times I'd just skim and throw away and reprompt, because I couldn't even begin to guess which part is correct, which isn't, so in the end, it was just a fast way to waste lots of tokens.
I see many people have complex systems involving skills, agent swarms with roles, agent factories, 20 termux panes and whatnot. Maybe that's the way, I don't know. Nobody knows. But it's not necessary to start getting value from LLMs. I therefore recommend my "lazy person's basic approach":
- Throw a task at LLM
- Check result, see what you like or not like about it
- If you just can't keep up, ask it to do smaller steps; if you don't understand, ask it to explain
- Just do that a lot, and don't be afraid to make both absurdly large and absurdly small asks
- You'll quickly get a feel for where AI can help you, and where it's not yet worth the time handholding it
Thanks, I'll continue to experiment!
I've used LLMs for some very small coding and math tasks, and I've found it much easier to work with on "formal" types of problems.
But for my consulting and legal work, which is "fuzzier," I struggle more with verification. I feel like I have to verify every single fact/premise in the LLM's output (this problem doesn't seem as big with coding?), as well as every single logical conclusion. Plus, even with explicit instructions I feel like the LLM is not great at anticipating what certain audiences will be interested in or find compelling.
Compared to the black-and-white nature of a computer program, all the components of a legal memorandum (premises, logical structure, conclusions, etc) seem so much fuzzier and require so much specific and accumulated context to apply reliably.
Really well put. This parallels my approach and it's good to see someone advocating a more tactful use of LLMs.
The output of agent swarms will of course grab attention, just like when you saw those same Microsoft Frontpage themes in the first few years it came out. But there's a deeper value in using the tool in harmony with your own experience and value additions.
Opus 5.5 is superhuman at most white collar work. At this point, not using it (including for health/law) is basically malpractice.
"Superhuman" in what sense? With respect to legal work.
Is it just faster? More persuasive? Better at coming up with novel arguments?
Speed is great from a client perspective, but not as important as accuracy on important stuff. I haven't seen LLMs produce arguments that are more persuasive than those crafted by good lawyers. Finding the "perfect case" in Westlaw is helpful, but rarely the deciding factor.
I could see asking an LLM to reproduce your work AFTER you've done a first pass, to see if you missed anything -- that seems like a real value add. But I truly can't understand how not using it would be "malpractice." In fact, I still think committing malpractice would be MUCH more likely with AI. (Now if by "malpractice", you meant pissing off clients with expensive bills, that's a different story...)
I think the parent post meant that it's very good at writing code and operating the computer. I think we're also biased because we've used older generation models and the latest ones are orders of magnitude better than what we had 6 months ago.
For example, for your legal work, it can:
Develop a machine learning model (CRF, running on CPU) for clause extraction from scratch, Find training documents and training data and train the model for various provisions (you might help it here by pointing it to the directory with annotated documents),
Build an app where you upload your legal documents and it extracts the provisions you want from all documents.
Integrate semantic search and an AI agent into your app which lets it answer any question about your documents with exact evidence from the documents.
All of this from one well crafted prompt, a few skills and a workflow. It could probably complete this in one night, while you sleep.
Or you can point it to your saas of choice and tell it - 'create me a copy of this app'.
All you need to know is how to steer it through the planning/research phase.
So it's not that it's 'accurate' in legal terms, it's that it can build you any tool you want from scratch so that you can become more accurate.
I'm not in the legal world, so my experience is limited from a coding perspective. I don't think AI could ever truly come up with novel arguments, at least not yet, but their value for you may be to sharpen your arguments or pull on every thread that you may not even have thought about.
For example, theres a very popular coding skill set call Matt Pocock's Grill Me where it just asks you a relentless amount of questions about the feature you're trying to build. You may be able to tailor this somehow to act as the opposing counsel and have it poke holes in your argument. I suppose the closest analog would be a sparring partner.
Its helpful for me and my ADHD brain because I can go down all sorts of rabbit holes and it can easily keep track of all the ideas i have while grilling and it can send sub-agents out to investigate other claims to add more context to the grilling session.
I've found it depends a lot on how Claude is setup, and spending some real time on the default Private Instructions to focus on accuracy, identifying/flagging uncertainty, citing sources, checking itself, and first surfacing all assumptions and asking questions dramatically pays off in terms of getting concise and solid usable results.
Edit: add: I've found this to the degree that a relative who works at a top biglaw firm and has their internal AI (MS-based) and used to find ChatGPT and Claude actively time-wasteful, when I posed a question they'd spent a day+ fruitlessly researching (multiple areas of law and international), Claude came up with a very useful answer in ~10min — they were quite impressed.
The last year has made a huge difference
The frontier labs themselves are still hiring hordes of human "account associates", "android engineers", and "data scientists" (https://openai.com/careers/search/), when they would be better positioned than anyone to leverage state-of-the art AI capabilities.
Yeah, and I keep wondering why.
They need top talent to operate the current generation of models, that's at the mediocre-talent level, in order to teach the next generation how to operate at their level, and eventually above. After which, they won't be needed anymore.
Because LLMs are a tool.
They are best placed to not fall for their own marketing claim that their tool can replace engineers.
Because either:
1) They are knowingly lying about AI being at human replacement level rather than a useful developer tool
AND/OR
2) They think they are special, and that their software developers ("AI researchers") are doing harder stuff than those most other companies
They can hire more people and still be massively augmented by AI.
Most people think that using an LLM requires no skill: it is just talking to a machine. I think it is a soft skill that only hours of practice can make you good at. Unfortunately, most of the people I know use LLMs as a search engine: their requests are just short sentences. They get basic results and either they stop there, not impressed or they keep using it as a smart replacement for Google. So they judge what AI can do from their first weak experience. Going further than that would requires more effort: going deep on a subject they know, thinking carefully about the prompt they write, more importantly judging the answer they get and pushing back.
I'm reminded of an interaction I had with a non-technical user. They knew ChatGPT can translate text, and so, to remove spelling mistakes from their (Dutch) text, they would first ask ChatGPT to translate it to English and then ask it to translate it back to Dutch. Mind-blown when I said "you can just say 'fix the spelling and gammar', no need to translate"
This is the level that many non-technical users are operating on.
Right. A life hack carried over from when Google Translate was the best tool, and loop-back through it could be reliably used for longer-form text (similar to how "did you mean?" in Google Search was an excellent spelling corrector for shorter text).
Using LLMs well with human language requires precision.
There are only a few professions that are intrinsically precise with language.
No surprise we're seeing adoption in software programming and mathematics.
If you work with spec files maybe.
For regular work, since we use modern reasoning models, you might as well type your requests with the wrong keyboard layout selected.
It still decodes your intent and completes the requested task. You can work from your phone with voice dictation. It requires no precision at all.
You are essentialy interacting with a SaaS that can be modified and altered at any point by its creator. Treating the LLM as some predictable entity that always works orders of magnitude better with precise language needs to be supported by evidence for each specific case.
Since the vast number of potential users of LLMs are, as you say, poor practitioners of precise language, it is incumbent on LLM providers to continuously iterate on providing LLMs that work better with imprecise language.
Hence, the people who claim that precise language is some 10x - 100x productivty skill seem to ignore the cycle of big LLM providers who are optimizing the models for users who lack those. (One of the early optimizations was the chatbot asking follow-up questions to the user for example).
So I think the people who think that their precise language skills are going to allow them to avoid the jobocalypse are also in a separate bubble of theit own. Precise language skills are going to matter less and less as LLMs increase their hand-holding skills for less equipped users.
This presupposes that the llm will be able to pick up the slack for the users of imprecise language as to me that is a proxy for domain expertise. If so then the human is not needed which does seem like a stretch.
I think the argument that precise language doesn't matter is even harder than that.
It requires that LLMs are equally effective with imprecise language as precise language.
And while I'm sure the frontier labs are continually iterating on 'garbage in, something vaguely useful out', it's a high bar to get that output to be as useful as if it were more precisely prompted.
My view is that the human is needed for verifying the source of truth, not for providing precise language which the LLM is continuously trained to adapt better without.
I have not come across a rigorously argued technical analysis that has claimed that using precise language can in any way eliminate the possibility of hallucinations or unaligned behavior. Short of providing a proof for the elimination of hallucinations and confabulations, merely reducing the behavior by a random percentage is not an acceptable status quo for completely removing the human.
> I think it is a soft skill that only hours of practice can make you good at.
So this, I think, is the crux of it.
When most people, at least in the world of SWE, think of what it means to be skilled at something, they think of knowledge and experience acquired over a much longer timeframe - months to years even. So from that pov, something you can get reasonably good at within just a few hours or days might not seem like much of a "skills moat" at all.
I'm currently seeing a lot of developers and AI industry insiders that do not distinguished between the capabilities of their LLM Agent and the added capabilities of the agent harness. They think the capabilities of the agent harness are within the LLM, and in that confusion they are trying to get their agent to make them an agent harness with all these capabilities. But LLM agent harnesses are brand spanking new, how to make these harnesses are not in the training data of their agent, these developers are more or less vibe begging their agents to make them a harness, so it hallucinates the harness architecture. I am seeing a "agent harnesses" that are delusions, the these developers are not strong and experienced enough to realize they are literally doing the worst work they possibly could... And these idiots are the AI leads.
> think it is a soft skill that only hours of practice can make you good at Yikes
It's shocking to me the amount of people that confused "not automatable" with "not cost effective to automate" (which is effectively everything), and the further step which is - the reduction in quality people will be forced to bear to get massively lower costs.
Most people are wrapped in plastic instead of cotton.
>"not cost effective to automate" (which is effectively everything)
There still seems to be a wide gulf in automation capabilities and things that require human dexterity, such as doing things for the infirm. Even just cooking and laundry seem quite a bit away. I guess the original tweet does mention this:
>Capability peaks in domains that are digital, verifiable and economically valuable.
Driving is the most impressive one so far, in my opinion.
Or maybe AI is particularly suited to programming and a few other niches.
It is very good at programming. Because programming is a lot of language, and these tools were also made by programmers and post trained to be good at it.
But I assure you it's even "better" (not really but the task is so much easier) at making your spreadsheet look nice and compare actuals vs forecast for your budget than a complicated programming task. It will just one shot it, and it will be objectively correct, all in a very short time. But these non tech people don't know they can use an agent, they have asked one question in the chat interface with 0 context about what they're trying to do, and conclude they are not good enough.
I'm not saying everything is replaceable by AI - and by the way I don't want anything to be replaced by it, but I believe it will happen - just that some people don't realize yet how "trivial" their job is to modern models.
This. The main blocker in AI proliferation I see is that people don't even consider it can do more than they thing it can.
Just yesterday, my wife recounted to me a conversation with her friend, where the friend mentioned they felt lazy and had some piece of copywriting done by AI this time. To which my wife said, "TeMPOraL would've also had AI upload it to the website afterwards".
Yes, I would have, and I would've also had it install the tools needed to upload it too, because why not. The only difference between me and my wife's friend here is that I know I can ask AI that, and I know it will work.
People have gotten very used to breaking down tasks into separate things they know how to do on a computer. Asking an LLM to do each of those individual tasks is easy but feels both lazy and also ‘so what?’ - I knew how to do that task.
The thing people are slow in realizing is that you don’t need to break the task down, and you don’t need to know how to do the task. You don’t even need to have the software installed on your computer to do the task, or even know that software to do the task exists - a sufficiently motivated LLM can pull down the needed open source tools and use them, or even build its own.
Exactly. It takes some... willingness to break through your own mindset (I don't want to call this "courage", but at times feels like it), to just tell the LLM, "I need <this and that> to be <like this>, make it so".
Two years ago, that was unthinkable. A year ago, it was a quick way to set a lot of tokens on fire. Today, well, I'm yet to find a thing Opus 5.5 doesn't do at least passable (and usually stellar) job on.
This is why hosted agents (muse, dots, grok bot) will succeed. Organizations won't relax their PC security policies, so they'll need a platform where the agent actually has permissions to develop and install software.
[flagged]
It is particularly suited to programming. But most business processes are basically a program that is executed. Especially in modern times, where office clerks have less and less decision-making options.
Tasks can be automated, jobs are much more difficult because a job is not just a sum of tasks. Add in needing the context of all of reality, and I am still unconvinced this will replace the most basic of jobs.
Tasks, oh yeah 100%.
I have a friend with a job like that and he is not being replaced not because it is Impossible but because it would be too expensive atm. But they replaced his whole team with workers south east Asia.
The prices are dropping fast. At some point one of the labs will decide they have a model that is good enough and they will burn its weights into silicon which will make the model much cheaper still.
I feel like there are 3 groups of people I regularly encounter: The first believes in LLMs, the second uses free Chat GPT, the third uses Gemini.
I hear stories from someone working in basically government accounting at the county level. Which amounts to low scale manual data ETL and basic reporting.
They do rely on something that sounds like a COBOL mainframe from 40 years ago to me. They see people making the same basic clerical mistakes every year when towns and government organizations report their numbers. Like literally not being able to format numbers according to certain rules on the first try and such. Sending excel file in emails is the level of automation they rely on. They are all union jobs.
She was recently preparing for an obscure accounting certification. There's very little publicly available information to use and at most one textbook. She was asking her LLM (Gemini?) to formulate exam-level questions instead and explain the answers. She told me it worked spectacularly well.
If that were true their jobs would be getting automated. But they're not. So...
[flagged]