Everything around LLMs is still magical and wishful thinking

https://dmitriid.com/everything-around-llms-is-still-magical-and-wishful-thinking

by troupo • 1 year ago

298 346 1 year ago

atemerev4• 1 year ago

"It's crypto all over again"

Crypto is a lifeline for me, as I cannot open a bank account in the country I live in, for reasons I can neither control nor fix. So I am happy if crypto is useless for you. For me and for millions like me, it is a matter of life and death.

As for LLMs — once again, magic for some, reliable deterministic instrument for others (and also magic). Just classified and sorted a few hundreds of invoices. Yes, magic.

harel2• 1 year ago

Can you elaborate on your situation? Which country are you in? How is crypto used there?

atemerev4• 1 year ago

I am a Russian immigrant in Switzerland. As of right now, all Swiss banks block all Russian bank accounts until their owners can provide a valid physical residence permit card, due to sweeping sanctions (meanwhile, Russian-owned companies continue to freely trade crude oil from here, as they use Swiss nominal directors — the hypocrisy is through the roof). My residence permit is on renewal now and the case is being dragged for 7 months already — so, no bank account.

mumbisChungo2• 1 year ago

Said this in another thread and I'll repeat it here:

It's the same problem that crypto experiences. Almost everyone is propagating lies about the technology, even if a majority of those doing so don't understand enough to realize they're lies (naivety vs malice).

I'd argue there's more intentional lying in crypto and less value to be gained, but in both cases people who might derive real benefit from the hard truth of the matter are turning away before they enter the door due to dishonesty/misrepresentation- and in both cases there are examples of people deriving real value today.

o11c• 1 year ago

I'd argue there's more intentional lying in crypto

I disagree. Crypto sounds more like intentional lying because it's primarily hyped in contexts typical for scams/gambling. Yes, there are businesses involved (anybody can start one), but they're mostly new businesses or a tiny tack-on to an existing business.

AI is largely being hyped within the existing major corporate structures, therefore its lies just get tagged as as "business as usual". That doesn't make them any less of a lie though.

tehjoker• 1 year ago

This is basically the only use case for crypto, and one for which it was explicitly designed: censorship resistance. This is why people have so much trouble finding useful things for it to do in the legal economy, it was explicitly designed to facilitate transactions the government doesn't want or can't facilitate. In some cases, there are humanitarian applications, there are also a lot of illicit applications.

foobarchu• 1 year ago

I don't think you actually disagree with the authors quip. You seem to want to use crypto as a currency, while OP was most likely referring to the grifting around crypto as an investment. If you're using it as a currency, then the people trying to pump and dump coins and use it for a money making vehicle are your adversaries. You are best served if it's stable instead of a rollercoaster of booms and busts.

atemerev4• 1 year ago

Stablecoins are a thing. But yes, I hate the current state of affairs, with "memecoins" and whatsnot. Particularly the government push from one particular country. We created crypto to be independent from governments, not to enable them.

troupo23• 1 year ago

It's a valid use case in the sea of nonsensical hype where "you are a moron if you don't believe in some true meaning of crypto".

"You had to be there to believe it" https://x.com/0xbags/status/1940774543553146956

AI craze is currently going through a similar period: any criticism is brushed away as being presented by morons who know nothing

dcre2• 1 year ago

Similar argument to https://www.baldurbjarnason.com/2025/trusting-your-own-judge..., but I like this one better because at least it doesn’t try to pull the rhetorical trick of slipping from “we can’t know whether LLMs are helping because we haven’t studied the question systematically” to “actually we do know, and they’re shit”.

troupo23• 1 year ago

Wow. Quite a conclusion from an article that actually doesn't reach for that conclusion

dcre2• 1 year ago

It’s a bad and incoherent essay. It simultaneously wants to say “we can’t know”, “we do know and it doesn’t work”, and “we know it makes engineering faster and that’s bad because most engineering is bad.” This isn’t nuance, it’s confusion. Sorry!

notphilipmoran• 1 year ago

I think that the disparity comes between people that are too in the weeds believing that their use cases apply to everyone. The reality is this world is made up of people with a wide array of different needs and AI is yet to proliferate into all usage applications.

Sure some of this comes from a lack of education.

But similar to crypto these movements only have value if the value is widely perceived. We have to work to continue to educate, continue to question, continue to understand different perspectives. All in favor of advancing the movement and coming out with better tech.

I am a supporter of both but I agree with the reference in the article to both becoming echo chambers at times. This is a setback we need to avoid.

tempodox• 1 year ago

Was this text generated by an LLM?

antonvs• 1 year ago

But similar to crypto these movements only have value if the value is widely perceived.

The difference in the AI case is that companies that are actually able to use it to boost productivity significantly will start to outcompete those who don't.

That's why, unlike crypto/blockchain, so many mainstream companies are pouring money into AI. It's not FOMO so much as fear of extinction.

blueboo• 1 year ago

Ok. Claude Code produces most code at Anthropic. Theres an enterprise code base, with acute real needs. There are real, experienced SWEs. How much babysitting and reviewing is undetermined; but the Ants seem to tremendously prefer the workflow.

Even crypto people didn’t dogfood their crypto like that, on their own critical path.

taurath• 1 year ago

Truly, how could they have the valuation they have and do anything else?

rightbyte• 1 year ago

but the Ants

Is that the official cutsie name people working there are called? Those feels so 2020 ...

peter422• 1 year ago

In my codebase in my proprietary project, it’s possible that LLMs write around half the code, but that’s just because a lot of the trivial parts of the project are quite verbose.

The really difficult and valuable parts of the codebase are very very far beyond what the current LLMs are capable of, and believe me, I’ve tried!

Writing the majority of the code is very different from creating the majority of the value.

And I really use and value LLMs, but they are not replacing me at the moment.

thisoneworks• 1 year ago

Regardless of the usefulness of llms, if you don't work at anthropic, how gullible are you to believe that claim at face value?

troupo23• 1 year ago

Ok. Claude Code produces most code at Anthropic.m

Does it? Or does their marketing tell you that? Strange that "most code is written by Claude" and they still hire for actual humans for all the positions from backend to API to desktop to mobile clients.

How much babysitting and reviewing is undetermined; but the Ants seem to tremendously prefer the workflow.

So. We know nothing about their codebase, actual flows, programming languages, depth and breadth of usage, how much babysitting is required...

jm20• 1 year ago

The best way I’ve heard this described: AI (LLMs) is probably 90% of the way to human levels of reasoning. We can probably get to about 95% optimizing current technology.

Whether or not we can get to 100% using LLMs is an open research problem and far from guaranteed. If we can’t, it’s unclear if it will ever really proliferate the way things hope. That 5% makes a big difference in most non-niche use cases…

krapp• 1 year ago

The best way I’ve heard this described: AI (LLMs) is probably 90% of the way to human levels of reasoning. We can probably get to about 95% optimizing current technology.

We don't know enough about how LLMs work or about how human reasoning works for this to be at all meaningful. These numbers quantify nothing but wishes and hype.

ath3nd3• 1 year ago

AI (LLMs) is probably 90% of the way to human levels of reasoning

Considering LLMs have 0 level of reasoning, I can't decide if it's a bad take, or a stab at the average human's level of reasoning.

In all seriousness, the actual numbers vary from 13% to 26%: https://fortune.com/2025/02/12/openai-deepresearch-humanity-...

My take is that there are fundamental limitations to try to pigeon-hole reasoning to LLMs, which are essentially a very very advanced autocomplete, and that's why those % won't jump too much too soon.

farts_mckensy5• 1 year ago

Whenever people claim that LLMs are not capable of reasoning, I put them into a category of people who are themselves not capable of reasoning.

andy99• 1 year ago

I've always looked at it as we're not making software that can think, we're (quite literally) demonstrating that vast categories of things don't need thought (for some quality level). The problem is, it's clearly not 100%, maybe it's 90-some percent, but it doesn't matter, we're only outsourcing the unimportant things that aren't definitional for a task.

This is very typical of naive automation, people assume that most of the work is X and by automating that we replace people, but the thing that's automated is almost never the real bottleneck. Pretty sure I saw an article here yesterday about how writing code is not the bottleneck in software development, and it holds everywhere.

farts_mckensy5• 1 year ago

The discussion is completely useless without defining what thought is and then demostrating that LLMs are not capable of it. And I doubt any definition you come up with will be workable.

Ferrus914• 1 year ago

The reason management thinks coding is the bottleneck is because they don't understand the first thing abiut code and neither have the ability or temprament to. Their whole professional career is about plausibly convincing other people through jargon, manipulation and popularity contests, which generally oprn up doors, solve problems and provoke seal like clapping from all involved. The idea that the core problem in many systems and software is due to their constitutonal inability to think rigorously to define requirements logically has never crossed their mind: it must be the magic spells those losers we bullied at school use and we are now tragically dependent on.

ethan_smith• 1 year ago

These percentage estimates of AI's proximity to "human reasoning" are misleading abstractions that mask fundamental qualitative differences in how LLMs and humans process information.

sherdil2022• 1 year ago

I follow Emily Bender on LinkedIn. She cuts through the AI hype and is also the author of The AI Con book - https://thecon.ai/

Of course people will either love AI or hate AI - and some don’t care. I am cautious especially when people say ‘AI is here to stay’. It takes away agency.

farts_mckensy5• 1 year ago

AI is here to stay, and you do not have agency over that. You can choose not to use it, but that has zero impact on the broader adoption rate. Just like when the automobile was introduced and society as a whole evolved.

goatlover2• 1 year ago

There was agency in policies promoting automobile adoption over rail travel. People who act like market forces are inevitable trends in nature tend to forget the political part where companies convince politicians to make policies in their favor, while bombarding the public with advertising. Like fossil fuel companies for example. Continued climate change wasn't inevitable. It was a choice humans with lots of money and power made and are still making.

jjtheblunt• 1 year ago

https://en.wikipedia.org/wiki/Clarke%27s_three_laws

includes the 3rd law, which reads, and seems on topic,

"Any sufficiently advanced technology is indistinguishable from magic."

readthenotes1• 1 year ago

And of course it says first law that applies here.

The people I have talked to at length about using AI tools claim that it has been a boon for productivity: a nurse, a doctor, three (old) software developers, a product manager, and a graduate student in Control Systems.

It is entirely believable that it may not, on average, help the average developer.

I'm reminded of the old joke that ends with "who are you going to believe, me or you're lying eyes?"

tasty_freeze• 1 year ago

One thing I find frustrating is that management where I work has heard of 10x productivity gains. Some of those claims even come from early adopters at my work.

But that sets expectation way too high. Partly it is due to Amdahl's law: I spend only a portion of my time coding, and far more time thinking and communicating with others that are customers of my code. Even if does make the coding 10x faster (and it doesn't most of the time) overall my productivity is 10-15% better. That is nothing to sneeze at, but it isn't 10x.

louthy• 1 year ago

overall my productivity is 10-15% better. That is nothing to sneeze at, but it isn't 10x.

It is something to sneeze at if you are 10-15% more expensive to employ due to the cost of the LLM tools. The total cost of production should always be considered, not just throughput.

bravesoul2• 1 year ago

That's a good insight be because with perfect competition it means you need to share your old salary with an LLM!

votepaunchy• 1 year ago

if you are 10-15% more expensive to employ due to the cost of the LLM tools

How is one spending anywhere close to 10% of total compensation on LLMs?

CharlesW7• 1 year ago

It is something to sneeze at if you are 10-15% more expensive to employ due to the cost of the LLM tools.

Claude Max is $200/month, or ~2% of the salary of an average software engineer.

wubrr2• 1 year ago

One thing I find frustrating is that management where I work has heard of 10x productivity gains. Some of those claims even come from early adopters at my work.

Similar situation at my work, but all of the productivity claims from internal early adopters I've seen so far are based on very narrow ways of measuring productivity, and very sketchy math, to put it mildly.

jppope2• 1 year ago

The reports from analysis of open source projects are that its something in the range of 10%-15% productivity gains... so it sounds like you're spot on

smcleod• 1 year ago

That's about right for copilots. It's much higher for agentic coding.

datpuz• 1 year ago

It's just another tech hype wave. Reality will be somewhere between total doom and boundless utopia. But probably neither of those.

The AI thing kind of reminds me of the big push to outsource software engineers in the early 2000's. There was a ton of hype among executives about it, and it all seemed plausible on paper. But most of those initiatives ended up being huge failures, and nearly all of those jobs came back to the US.

People tend to ignore a lot of the little things that glue it all together that software engineers do. AI lacks a lot of this. Foreigners don't necessarily lack it, but language barriers, time zone differences, cultural differences, and all sorts of other things led to similar issues. Code quality and maintainability took a nosedive and a lot of the stuff produced by those outsourced shops had to be thrown in the trash.

I can already see the AI slop accumulating in the codebases I work in. It's super hard to spot a lot of these things that manage to slip through code review, because they tend to look reasonable when you're looking at a diff. The problem is all the redundant code that you're not seeing, and the weird abstractions that make no sense at all when you look at it from a higher level.

2muchcoffeeman• 1 year ago

This was what I was saying to a friend the other day. I think anyone vaguely competent that is using LLMs will make the technology look far better than it is.

Management thinks the LLM is doing most of the work. Work is off shored. Oh, the quality sucks when someone without a clue is driving. We need to hire again.

mlinsey2• 1 year ago

I don't disagree with your assessment of the world today, but just 12 months ago (before the current crop of base models and coding agents like Claude Code), even that 10X improvement of writing some-of-the-code wouldn't have been true.

__loam• 1 year ago

It still isn't.

timr4• 1 year ago

I don't disagree with your assessment of the world today, but just 12 months ago (before the current crop of base models and coding agents like Claude Code), even that 10X improvement of writing some-of-the-code wouldn't have been true.

So? It sounds like you're prodding us to make an extrapolation fallacy (I don't even grant the "10x in 12 months" point, but let's just accept the premise for the sake of argument).

Honestly, 12 months ago the base models weren't substantially worse than they are right now. Some people will argue with me endlessly on this point, and maybe they're a bit better on the margin, but I think it's pretty much true. When I look at the improvements of the last year with a cold, rational eye, they've been in two major areas:

  * cost & efficiency

  * UI & integration

So how do we improve from here? Cost & efficiency are the obvious lever with historical precedent: GPUs kinda suck for inference, and costs are (currently) rapidly dropping. But, maybe this won't continue -- algorithmic complexity is what it is, and barring some revolutionary change in the architecture, LLMs are exponential algorithms.

UI and integration is where most of the rest of the recent improvement has come from, and honestly, this is pretty close to saturation. All of the various AI products already look the same, and I'm certain that they'll continue to converge to a well-accepted local maxima. After that, huge gains in productivity from UX alone will not be possible. This will happen quickly -- probably in the next year or two.

Basically, unless we see a Moore's law of GPUs, I wouldn't bet on indefinite exponential improvement in AI. My bet is that, from here out, this looks like the adoption curve of any prior technology shift (e.g. mainframe -> PC, PC -> laptop, mobile, etc.) where there's a big boom, then a long, slow adoption for the masses.

majormajor• 1 year ago

just 12 months ago (before the current crop of base models and coding agents like Claude Code), even that 10X improvement of writing some-of-the-code wouldn't have been true.

You had to paste more into your prompts back then to make the output work with the rest of your codebase, because there weren't good IDEs/"agents" for it, but you've been able to get really really good code for 90% of "most" day to day SWE since at least OpenAI releasing the ChatGPT-4 API, which was a couple years ago.

Today it's a lot easier to demo low-effort "make a whole new feature or prototype" things than doing the work to make the right API calls back then, but most day to day work isn't "one shot a new prototype web app" and probably won't ever be.

I'm personally more productive than 1 or 2 years ago now because the time required to build the prompts was slower than my personal rate of writing code for a lot of things in my domain, but hardly 10x. It usually one-shots stuff wrong, and then there's a good chance that it'll take longer to chase down the errors than it would've to just write the thing - or only use it as "better autocomplete" - in the first place.

deadbabe• 1 year ago

Wait till they hear about the productivity gains from using vim/neovim.

Your developers still push a mouse around to get work done? Fire them.

ghuntley2• 1 year ago

Canva has seen a 30% productivity uplift - https://fortune.com/2025/06/25/canva-cto-encourages-all-5000...

AI is the new uplift. Embrace and adapt, as a rift is forming (see my talk at https://ghuntley.com/six-month-recap/), in what employers seek in terms of skills from employees.

I'm happy to answer any questions folks may have. Currently AFK [2] vibecoding a brand new programming language [1].

[1] https://x.com/GeoffreyHuntley/status/1940964118565212606 [2] https://youtu.be/e7i4JEi_8sk?t=29722

CuriouslyC• 1 year ago

And that's with 50% adoption and probably a broad distribution of tool use skill.

ofjcihen2• 1 year ago

There’s something hilariously Portlandia about making outlandish claims with complete confidence and then plugging your own talk.

eviks• 1 year ago

The productivity for software engineers is at around 30%

That would be a 70% descent?

coolKid721• 1 year ago

On my personal projects it's easily 10x faster if not more in some circumstances. At work where things are planned out months in advanced and I'm working with 5 different teams to figure out the right way to do things for requirements that change 8 times during development? Even just stuff with PR review and making sure other people understand it and can access it. idk sometimes it's probably break even or that 10-15%. It just doesn't work well in some environments and what really makes it flourish (having super high quality architectural planning/designs/standardized patterns etc.) is basically just not viable at anything but the smallest startups and solo projects.

Frankly even just getting engineers to agree upon those super specificized standardized patterns is asking a ton, especially since lots of the things that help AI out are not what they are used to. As soon as you have stuff that starts deviating it can confuse the AI and makes that 10x no longer accessible. Also no one would want to review the PRs I'd make for the changes I do on my "10x" local project... Especially maintaining those standards is already hard enough on my side projects AI will naturally deviate and create noise and the challenge is constructing systems to guide that to make sure nothing deviates (since noise would lead to more noise).

I think it's mostly a rebalancing thing, if you have 1 or a couple like minded engineers who intend to do it they can get that 10x. I do not see that EVER existing in any actual corporate environment or even once you get more then like 4 people tbh.

Ai for middle management and project planning on the other hand...

abletonlive6• 1 year ago

I’m a tech lead and I have maybe 5X output now compared to everybody else under me. Quantified by scoring tickets at a team level. I also have more responsibilities outside of IC work compared to the people under me. At this point I’m asking my manager to fire people that still think llms are just toys because I’m tired of working with people with this poor mindset. A pragmatic engineer continually reevaluates what they think they know. We are at a tipping point now. I’m done arguing with people that have a poor model of reality. The rest of us are trying to compete and get shit done. This isn’t an opinion or a game. It’s business with real life consequences if you fall behind. I’ve offered to share my workflows, prompts, setup. Guess how many of these engineers have taken me up on my offer. 1-2 and the juniors or ones that are very far behind have not.

dgfitz• 1 year ago

I will thank God every day I don’t work with you or for you. How toxic.

blibble2• 1 year ago

I’m done arguing with people that have a poor model of reality.

isn't this the entire LLM experience?

mattmanser• 1 year ago

You've been doing the big I am about LLMs on HN for most of your last comments.

Everyone else who raises any doubts about LLMs is an idiot and you're 10,000x better than everyone else and all your co-workers should be fired.

But what's absent from all your comments is what you make. Can you tell us what you actually do in your >500k job?

Are you, by any chance, a front-end developer?

Also, a team-lead that can't fire their subordinates isn't a team-lead, they're a number two.

Applejinx• 1 year ago

Are you the one at Ableton responsible for it ignoring the renaming of parameter names during the setState part of a Live program? Some of us are already jumping through ridiculous hoops to cover for your… mindset. There's stuff coming up that used to work and doesn't now, like in Live 12. From your response I would guess this is a trend that will hold.

We should not be having to code special 'host is Ableton Live' cases in JUCE just to get your host to work like the others.

Can you please not fire any people who are still holding your operation together?

gabrieledarrigo2• 1 year ago

Dude, if you are a tech lead, and you measure productivity by scoring tickets, you are doing it pretty badly. I would fire you instead.

swader9993• 1 year ago

"I’ve offered to share my workflows, prompts" That should all be checked in.

nasduia• 1 year ago

A new copypasta is born.

ofjcihen2• 1 year ago

It’s funny. We fired someone with this attitude Thursday. And by this attitude I mean yours.

Not necessarily because of their attitude but because it turns out the software they were shipping was ripe with security issues. Security managed to quickly detect and handle the resulting incident. I can’t say his team were sad to see him go.

solumunus2• 1 year ago

You seem completely insufferable and incredibly cringeworthy.

TeMPOraL24• 1 year ago

Maybe it's due to a more R&D-ish nature of my current work, but for me, LLMs are delivering just as much gains in the "thinking" part as in "coding" part (I handle the "communicating" thing myself just fine for now). Using LLMs for "thinking" tasks feels similar to how mastering web search 2+ decades ago felt. Search engines enabled access to information provided you know what you're looking for; now LLMs boost that by helping you figure out what you're looking for in the first place (and then conveniently searching it for you, too). This makes trivial some tasks I previously classified as hard due to effort and uncertainty involved.

At this point I'd say about 1/3 of my web searches are done through ChatGPT o3, and I can't imagine giving it up now.

(There's also a whole psychological angle in how having LLM help sort and rubber-duck your half-baked thought makes many task seem much less daunting, and that alone makes a big difference.)

jorl173• 1 year ago

This, and if you add in a voice mode (e.g. ChatGPT's Advanced Mode), it is perfect for brainstorming.

Once I decide I want to "think a problem through with an LLM", I often start with just the voice mode. This forces me to say things out loud — which is remarkably effective (hear hear rubber duck debugging) — and it also gives me a fundamentally different way of consuming the information the LLM provides me. Instead of being delivered a massive amount of text, where some information could be wrong, I instead get a sequential system where I can stop/pause the LLM/redirect it as soon as something gets me curious or as I find problems with it said.

You would think that having this way of interacting would be limiting, as having a fast LLM output large chunks of information would let you skim through it and commit it to memory faster. Yet, for me, the combination of hearing things and, most of all, not having to consume so much potentially wrong info (what good is it to skim pointless stuff), ensures that ChatGPT's Advanced Voice mode is a great way to initially approach a problem.

After the first round with the voice mode is done, I often move to written-form brainstorming.

solumunus2• 1 year ago

I’m surprised it’s only 1/3rd. 90% of my searches for information start at Perplexity or Claude at this point.

seba_dos12• 1 year ago

From time to time I use an LLM to pretend to research a topic that I had researched recently, to check how much time it would have saved me.

So far, most of the time, my impression was "I would have been so badly mislead and wouldn't even know it until too late". It would have saved me some negative time.

The only thing LLMs can consistently help me with so far is typing out mindless boilerplate, and yet it still sometimes requires manual fixing (but I do admit that it still does save effort). Anything else is hit or miss. The kind of stuff it does help researching with is usually the stuff that's easy to research without it anyway. It can sometimes shine with a gold nugget among all the mud it produces, but it's rare. The best thing is being able to describe something and ask what it's called, so you can then search for it in traditional ways.

That said, search engines have gotten significantly worse for research in the last decade or so, so the bar is lower for LLMs to be useful.

doug_durham2• 1 year ago

How much of the communication and meetings are because traditionally code was very expensive and slow to create? How many of those meetings might be streamlined or entirely disappear in the future? In my experience there is are a lot of process around making sure that software on schedule track and that it's doing what it is supposed to do. I think that the software lifecycle is about to be reinvented.

thunky3• 1 year ago

One thing I find frustrating is that management where I work has heard of 10x productivity gains.

That may also be in part because llms are not as big of an accelerant for junior devs as they are for seniors (juniors don't know what is good and bad as well).

So if you give 1 senior dev a souped up llm workflow I wouldn't be too surprised if they are as productive as 10 pre-llm juniors. Maybe even more, because a bad dev can actually produce negative productivity (stealing from the senior), in which case it's infinityx.

Even a decent junior is mostly limited to doing the low level grunt work, which llms can already do better.

Point is, I can see how jobs could be lost, legitimately.

Loughla• 1 year ago

The item lost is pipeline of talent in all of this though.

Precision machining is going through an absolute nightmare where the journeymen or master machinists are aging out of the work force. These were people who originally learned on manual machines, and upgraded to CNC over the years. The pipeline collapsed about 1997.

Now there are no apprentice machinists to replace the skills of the retiring workforce.

This will happen to software developers. Probably faster because they tend to be financially independent WAY sooner than machinists.

ericmcer• 1 year ago

Its great when they use AI to write a small app “without coding at all” over the weekend and then come in on Monday to brag about it and act baffled that tasks take engineers any time at all.

tom_m• 1 year ago

Expectations are absolutely way too high. It's going to lead to a lot of toxicity and people being fired. It's really going to suck.

labrador4• 1 year ago

I'm a retired programmer. I can't imagine trusting code generated by probablities for anything mission critical. If it were close and just needed minor tweaks I could understand that. But I don't have experience with it.

My comment is mainly to say LLMs are amazing in areas that are not coding, like brainstorming, blue sky thinking, filling in research details, asking questions that make me reflect. I treat the LLM like a thinking partner. It does make mistakes, but those can be caught easily by checking other sources, or even having another LLM review the conclusions.

garciasn• 1 year ago

Well; I can't speak to your specific experience (current or past) but I'm telling you that while I'm skeptical as hell about EVERYTHING, it's blowing my expectations away in every conceivable way.

I built something in less than 24h that I'm sure would have taken us MONTHS to just get off the ground, let alone to the polished version it's at right now. The most impressive thing is that it can do all of the things that I absolutely can do, just faster. But the most impressive thing is that it can do all the things I cannot possibly do and would have had to hire up/contract out to accomplish--for far less money, time, and with faster iterations than if I had to communicate with another human being.

It's not perfect and it's incredibly frustrating at times (hardcoding values into the code when I have explicitly told it not to; outright lying that it made a particular fix, when it actually changed something else entirely unrelated), but it is a game changer IMO.

gyomu2• 1 year ago

I built something in less than 24h that I'm sure would have taken us MONTHS to just get off the ground, let alone to the polished version it's at right now

Would love to see it!

98eb1d0ff7fb96• 1 year ago

See, your comment is a good example of what's going wrong. The OP specifically mentioned "mission critical things" - My interpretation of that would be things that are not allowed to break, because otherwise people might die, in the worst case - and you were talking about just SOMETHING that got "done" faster. No mention about anything critical.

Of course, I was playing around with claude code, too, and I was fascinated how fun it can be and yes, you can get stuff done. But I have absolutely no clue what the code is doing and if there are some nasty mistakes. So it kinda worked, but I would not use that for anything "mission critical" (whatever this means).

svdr• 1 year ago

I've been programming for 40 years and have started using LLM's a few months ago, and it has really changed the way I work. I let it write pieces of code (pasting error messages from logs mostly result in a fix in less then a minute), but also brainstorming about architecture or new solutions. Of course I check the code it writes, but I'm still almost daily amazed at the intelligence and accuracy. (Very much unlike crypto).

labrador4• 1 year ago

That's good to know if I ever get an idea for a side project. Anything to relieve the tedius aspects of programming would be very welcome.

anon-3988• 1 year ago

There's one thing that I find LLM extremely good at: data science. Since the IO is well defined, you can easily verify that the output is correct. You can even ask it write tests for you given that you know certain properties of the data.

The problem is that the LLM needs context of what you are doing, contexts that you won't (or too lazy) to give in a chat with it ala ChatGPT. This is where Claude Code changes the game.

For example, you have PCAP file where each UDP packet contain multiple messages.

How do you filter the IP/port/protocol/time? Use LLM, check the output

How do you find the number of packets that have patterns A, AB, AAB, ABB.... Use LLM, check the output

How to create PCAPs that only contain those packets for testing? Use LLM, check the output

Etc etc

Since it can read your code, it is able to infer (because lets be honest, you work aint special) what you are trying to do at a much better rate. In any case, the fact that you can simply ask "Please write a unit test for all of the above functions" means that you can help it verify itself.

fleebee• 1 year ago

I tried the "thinking partner" approach for a while and for a moment I thought it worked well, but at some point the cracks started to show and I called the bluff. LLMs are extremely good at creating an illusion that they know things and are capable of reasoning, but they really don't do a good job of cultivating intellectual conversation.

I think it's dangerously easy to get misled when trying to prod LLMs for knowledge, especially if it's a field you're new to. If you were using a regular search engine, you could look at the source website to determine the trustworthiness of its contents, but LLMs don't have that. The output can really be whatever, and I don't agree it's necessarily that easy to catch the mistakes.

labrador4• 1 year ago

You don't say what LLM you are using. I'm using ChatGPT 4o. I'm getting great results, but I review the output with a skeptical eye similar to how I read Wikipedia articles. Like Wikipedia, GPT 4o is great for surfacing new topics for research and does it quickly, which makes stream of thought easier.

selfhoster115• 1 year ago

This is very model-dependent. If you use something heavy on sycophancy and low on brain cells (like GPT-4o, the default paid ChatGPT model), you'll get lots and lots of cracks because these models are optimised for engagement.

That said, don't use model output directly. Use it to extract "shibboleth" keywords and acronyms in that domain, then search those up yourself with a classical search engine (or in a follow-up LLM query). You'll access a lot of new information that way, simply because you know how to surface it now.

edg50003• 1 year ago

Mostly agree. What works better is seeing the AI as a "high level to low level converter", in the same way Java is converted to machine code when it's ran. You describe exactly what you want it to report or do, and steer it whenever there are ambiguities. It does "grunt work" for you. With the bar of what grunt work means being moved up. Grunt work used to be doing the dishes or calculating numbers by hand on a paper spreasheet. Decades ago we automated those these. Now we've automated searching for information, summarization, implementing fully specified technical desings for software, the list goes on.

shash• 1 year ago

I say this as an LLM skeptic.

All code, including stuff that we experienced coders write is inherently probabilistic. That’s why we have code reviews, unit tests, pair programming, guidelines and guardrails in any critical project. If you’re using LLM output uncritically, you’re doing it wrong, but if you’re using _human_ output uncritically you’re doing it wrong too.

That said, they are not magic, and my fear is that people use copilots and agentic models and all the rest to hide poor engineering practice, building more and more boilerplate instead of refactoring or redesigning for efficiency or safety or any of the things that matter in the long run.

danielbln3• 1 year ago

LLMs turn foot guns into foot bazookas. Still, if you learn to aim them away from your body they do make a bigger better boom.

edg50003• 1 year ago

Treating it like a thinking partner is good. When programming, I don't treat it like a person, but rather a very high programming language. Imagine exactly how you want the code to be. Then find a way to express that unambiguously in natural language; the idea is that you'll still have a bit of work to do writing things out, but it will be a lot quicker than typing out all code by hand. Combined with iterations of feedback as well as having the AI build and run your program (at least as a sanity check); and asking the AI to check the program behaviour in the same way you would, gets you quite far.

A limitation is the lack of memory. If you steer it from style A to B using multiple points of feedback, if this is not written down, the next AI session you'll have to reexplain this all over.

Deepseek is about 1TB in weights; maybe that is why LLMs don't remember things across sessions yet. I think everybody can have their personal AI (hosted remote unless you own lots of compute); it should remember what happened yesterday; in particular the feedback it was given when developing. As an AI layman I do think this is the next step.

danielbln3• 1 year ago

You don't trust the code coming out of the probabilistic machine. You build a validation cage around it with hard interfaces and you also review the output.

yahoozoo• 1 year ago

The thing is, the questions such as “are they an expert in the domain” … “are they good at coding to being with” … and so on only really apply to the folks claiming positive results from LLMs. On the flip side, someone not getting much value - or dare I say, a skeptic - pushes back because they _can see_ what the LLM gave them is wrong. I’m not providing any revelatory comment here, but the simple truth is: people who are shit to begin with think this is all amazing/magic/the future.

hotpotat8• 1 year ago

I have to say I’m in the exact camp the author is complaining about. I’ve shipped non trivial greenfield products which I started back when it was only ChatGPT and it was shitty. I started using Claude with copying and pasting back and forth between the web chat and XCode. Then I discovered Cursor. It left me with a lot of annoying build errors, but my productivity was still at least 3x. Now that agents are better and claude 4 is out, I barely ever write code, and I don’t mind. I’ve leaned into the Architect/Manager role and direct the agent with my specialized knowledge if I need to.

I started a job at a demanding startup and it’s been several months and I have still not written a single line of code by hand. I audit everything myself before making PRs and test rigorously, but Cursor + Sonnet is just insane with their codebase. I’m convinced I’m their most productive employee and that’s not by measuring lines of code, which don’t matter; people who are experts in the codebase ask me for help with niche bugs I can narrow in on in 5-30 minutes as someone whose fresh to their domain. I had to lay off taking work away from the front end dev (which I’ve avoided my whole career) because I was stepping on his toes, fixing little problems as I saw them thanks to Claude. It’s not vibe coding - there’s a process of research and planning and perusing in careful steps, and I set the agent up for success. Domain knowledge is necessary. But I’m just so floored how anyone could not be extracting the same utility from it. It feels like there’s two articles like this every week now.

the__alchemist2• 1 year ago

Web dev CRUD in node?

hotpotat8• 1 year ago

Multi platform web+native consumer application with lots of moving parts and integration. I think to call it a CRUD app would be oversimplifying it.

gyomu2• 1 year ago

I’ve shipped non trivial greenfield products

Links please

hotpotat8• 1 year ago

I’d like to, but purposefully am using a throwaway account. It’s an iOS app rated 4.5 stars on the app store and has a nice community. Mild userbase, in the hundreds.

larve2• 1 year ago

Here's maybe the most impressive thing I've vibecoded, where I wanted to track a file write/read race condition in a vscode extension: https://github.com/go-go-golems/go-go-labs/tree/main/cmd/exp...

This is _far_ from web crud.

Otherwise, 99% of my code these days is LLM generated, there's a fair amount of visible commits from my opensource on my profile https://github.com/wesen .

A lot of it is more on the system side of things, although there are a fair amount of one-off webapps, now that I can do frontends that don't suck.

0x696C69613• 1 year ago

I find that the code quality LLMs output is pretty bad. I end up going through so many iterations that it ends up being faster to do it myself. What I find agents actually useful for is doing large scale mechanical refractors. Instead of trying to figure out the perfect vim macro or AST rewrite script, I'll throw an agent at it.

CharlesW7• 1 year ago

I find that the code quality LLMs output is pretty bad.

That was my experience with Cursor, but Claude Code is a different world. What specific product/models brought you to this generalization?

AnotherGoodName• 1 year ago

I disagree strongly at this point. The code is generally good if the prompt was reasonable at this point but also every test possible is now being written, every ui element has the all required traits, every function has the correct documentation attached, the million little refactors to improve the codebase are being done, etc.

Someone told me ‘ai makes all the little things trivial to do’ and i agree strongly with that. Those many little things are things that together make a strong statement about quality. Our codebase has gone up in quality significantly with ai whereas we’d let the little things slide due to understaffing before.

the__alchemist2• 1 year ago

What sort of mechanical refactors?

bamboozled2• 1 year ago

I use Claude code for hours a day, it’s a liar, trust what it does at your own risk.

I personally think you’re sugar coating the experience.

CharlesW7• 1 year ago

I use Claude code for hours a day, it’s a liar, trust what it does at your own risk.

The person you're responding to literally said, "I audit everything myself before making PRs and test rigorously".

swader9993• 1 year ago

It lies with such enthusiasm though.

troupo23• 1 year ago

Please re-read the article. Especially the first list of things we don't know about you, your projects etc.

Your specific experience cannot be generalized. And speaking as the author, and who is (as written in the article) literally using these tools everyday.

But I’m just so floored how anyone could not be extracting the same utility from it. It feels like there’s two articles like this every week now.

This is where we learn that you haven't actually read the article. Because it is very clearly stating, with links, that I am extracting value from these tools.

And the article is also very clearly not about extracting or not extracting value.

hotpotat8• 1 year ago

I did read the entire article before commenting and acknowledge that you are using them to some affect, but the line about 50% of the time it works 50% of the time is where I lost faith in the claims you’re making. I agree it’s very context dependent but, in the same way, you did not outline your approaches and practices in how you use AI in your workflow. The same lack of context exists on the other side of the argument.

mccoyb• 1 year ago

Same experience here, probably in a slightly different way of work (PhD student). Was extremely skeptical of LLMs, Claude Code has completely transformed the way I work.

It doesn't take away the requirements of _curation_ - that remains firmly in my camp (partially what a PhD is supposed to teach you! to be precise and reflective about why you are doing X, what do you hope to show with Y, etc -- breakdown every single step, explain those steps to someone else -- this is a tremendous soft skill, and it's even more important now because these agents do not have persistent world models / immediately forget the goal of a sequence of interactions, even with clever compaction).

If I'm on my game with precise communication, I can use CC to organize computation in a way which has never been possible before.

It's not easier than programming (if you care about quality!), but it is different, and it comes with different idioms.

exe342• 1 year ago

but my productivity was still at least 3x

How do you measure this?

hotpotat8• 1 year ago

Mean time to shipping features of various estimated difficulty. It’s subjective and not perfect, but generally speaking I need to work way less. I’ll be honest, one thing I think I could have done faster without AI was to implement CRDT-based cloud sync for a project I have going. I think I’ve tried to utilize AI too much for this. It’s good at implementing vector clock implementations, but not at preventing race conditions.

xoralkindi• 1 year ago

I audit everything myself before making PRs and test rigorously

How do you audit code from an untrusted source that quickly, LLMs do not have the whole project in their heads and are proned to hallucinate.

On average how long are your prompts and does the LLM also write the unit tests?

hotpotat8• 1 year ago

The auditing is not quick. I prefer cursor to claude code because I can review its changes while it’s going more easily and stop and redirect it if it starts to veer off course (which is often, but the cost of doing business). Over time I still gain an understanding of the codebase that I can use to inform my prompts or redirection, so it’s not like I’m blindly asking it to do things. Yes, I do ask it to write unit tests a lot of the time. But I don’t have it spin off and just iterate until the unit tests pass — that’s a recipe for it to do what it needs to do to pass them and is counterproductive. I plan what I want the set of tests to look like and have them write functions in isolation without mentioning tests, and if tests fail I go through a process of auditing the failing code and then the tests themselves to make sure nothing was missed. It’s exactly how I would treat a coworkers code that I review. My prompts range from a few sentences to a few paragraphs, and nowadays I construct a large .md file with a checklist that we iterate on for larger refactors and projects to manage context

rester3242• 1 year ago

But you just confirmed everything the blogpost claimed.

You didn't share any evidence with us even though you claim unbelievable things.

You even went as far as registering a throwavay account to hide your identity and to make verifying any of your claims impossible.

Your comment feels more like a joke to me

peteforde• 1 year ago

... this from an account with <100 karma.

Look, the person who wrote that comment doesn't need to prove anything to you just because you're hopped up after reading a blog post that has clearly given you a temporary dopamine bump.

People who understand their domains well and are excellent written communicators can craft prompts that will do what we used to spend a week spinning up. It's self-evident to anyone in that situation, and the only thing we see when people demand "evidence" is that you aren't using the tools properly.

We don't need to prove anything because if you are working on interesting problems, even the most skeptical person will prove it to themselves in a few hours.

gabrieledarrigo2• 1 year ago

I started a job at a demanding startup and it’s been several months and I have still not written a single line of code by hand

Damn, this sounds pretty boring.

hotpotat8• 1 year ago

It’s not. It’s like I used to play baseball professionally and now I’m a coach or GM building teams and yielding results. It’s a different set of skills. I’m working mostly in idea space and seeing my ideas come to life with a faster feedback loop and the toil is mostly gone

dosnem• 1 year ago

there’s a process of research and planning and perusing in careful steps, and I set the agent up for success

Are there any good articles you can share or maybe your process? I’m really trying to get good at this but I don’t find myself great at using agents and I honestly don’t know where to start. I’ve tried the memory bank in cline, tried using more thinking directives, but I find I can’t get it to do complex things and it ends up being a time sink for me.

hotpotat8• 1 year ago

https://www.lesswrong.com/posts/dxiConBZTd33sFaRC/field-note...

itsafarqueue• 1 year ago

More anecdata: +1 for “LLMs write all my production code now”. 25+ years in industry, as expert as it’s possible to be in my domain. 100% agree LLMs fail hilariously badly, often, and dangerously. And still, write ~all my code.

No agenda here, not selling anything. Just sitting here towards the later part of my career, no need to prove anything to anyone, stating the view from a grey beard.

Crypto hype was shill from grifters pumping whatever bag holding scam they could, which was precisely what the behavioral economic incentives drove. GenAI dev is something else. I’ve watched many people working with it, your mileage will vary. But in my opinion (and it’s mine, you do you), hand coding is an apocryphal skill. The only part I wonder about is how far up and down the system/design/architecture stack the power-tooling is going to go. My intuition and empirical findings incline towards a direction I think would fuel a flame war. But I’m just grey beard Internet random, and hey look, no evidence just more baseless claims. Nothing to see here.

Disclosure: I hold no direct shares in Mag 7, nor do I work for one.

pier25• 1 year ago

And you created an account just to write this unbelievable claim?

A bit suspicious, wouldn’t you agree?

DiscourseFan• 1 year ago

ChatGPT can write research papers in about 20 minutes—its the “Deep Research” tool. These are not original papers, but it can perform complex tasks that require multiple steps that would normally take a person hours. No its not a magic superintelligence, but it will transform a lot of white collar labor.

martinald2• 1 year ago

I personally don't really get this.

_So much_ work in the 'services' industries globally comes down to really a human transposing data from one Excel sheet to another (or from a CRM/emails to Excel), manually. Every (or nearly every) enterprise scale company will have hundreds if not thousands of FTEs doing this kind of work day in day out - often with a lot of it outsourced. I would guess that for every 1 software engineer there are 100 people doing this kind of 'manual data pipelining'.

So really for giant value to be created out of LLMs you do not need them to be incredible at OCaml. They just need to ~outperform humans on Excel. Where I do think MCP really helps is that you can connect all these systems together easily, and a lot of the errors in this kind of work came from trying to pass the entire 'task' in context. If you can take an email via MCP, extract some data out and put it into a CRM (again via MCP) a row at a time the hallucination rate is very low IME. I would say at least a junior overworked human level.

Perhaps this was the point of the article, but non-determinism is not an issue for these kind of use cases, given all the humans involved are not deterministic either. We can build systems and processes to help enforce quality on non deterministic (eg: human) systems.

Finally, I've followed crypto closely and also LLMs closely. They do not seem to be similar in terms of utility and adoption. The closest thing I can recall is smartphone adoption. A lot of my non technical friends didn't think/want a smartphone when the iPhone first came out. Within a few years, all of them have them. Similar with LLMs. Virtually all of my non technical friends use it now for incredibly varied use cases.

deepsquirrelnet• 1 year ago

Making a comparison to crypto is lazy criticism. It’s not even worth validating. It’s people who want to take the negative vibe from crypto and repurpose it. The two technologies have nothing to do with each other, and therefore there’s clearly no reason to make comparative technical assessments between them.

That said, the social response is a trend of tech worship that I suspect many engineers who have been around the block are weary of. It’s easy to find unrealistic claims, the worst coming from the CEOs of AI companies.

At the same time, a LOT of people are practically computer illiterate. I can only imagine how exciting it must seem to people who have very limited exposure to even basic automation. And the whole “talking computer” we’ve all become accustomed to seeing in science fiction is pretty much becoming reality.

There’s a world of takes in there. It’s wild.

I worked in ML and NLP several years before AI. What’s most striking to me is that this is way more mainstream than anything that has ever happened in the field. And with that comes a lot of inexperience in designing with statistical inference. It’s going to be the Wild West for a while — in opinions, in successful implementation, in learning how to form realistic project ideas.

Look at it this way: now your friend with a novel app idea can be told to do it themselves. That’s at least a win for everyone.

ysofunny• 1 year ago

wrong.

ultimately, crypto is information science. mathematically, cryptography, compression, and so on (data transmission) are all the "same" problem.

LLMs compress knowledge, not just data, and they do it in a lossy way.

traditional information science work is all about dealing with lossless data in a highly lossy world.

TeMPOraL24• 1 year ago

Look at it this way: now your friend with a novel app idea can be told to do it themselves. That’s at least a win for everyone.

For now, anyways. Thing is, that friend now also has a reasonable shot at succeeding in doing it themselves. It'll take some more time for people to fully internalize it. But let's not forget that there's a chunk of this industry that's basically building apps for people with "novel app ideas" that have some money but run out of friends to pester. LLMs are going to eat a chunk out of that business quite soon.

marinmania• 1 year ago

I would guess that for every 1 software engineer there are 100 people doing this kind of 'manual data pipelining'.

For what time of company is this true? I really would like someone to just do a census of 500 white collar jobs and categorize them all. Anything that is truly automatic has already been automated away.

I do think AI will cause a lot of disruption, but very skeptical of the view that most people with white collar jobs are just "email jobs" or data entry. That doesn't fit my experience at all, and I've worked at some large bureaucratic companies that people here would claim are stuck in the past.

saulpw2• 1 year ago

Each FTE doing that manual data pipelining work is also validating that work, and they have a quasi-legal responsibility to do their job correctly and on time. They may have substantial emotional investment in the company, whether survival instinct to not be fired, or ambition to overperform, or ethics and sense to report a rogue manager through alternate channels.

An LLM won't call other nodes in the organization to check when it sees that the value is unreasonable for some out-of-context reason, like yesterday was a one-time-only bank holiday and so the value should be 0. *It can be absolutely be worth an FTE salary to make sure these numbers are accurate.* And for there to be a person to blame/fire/imprison if they aren't accurate.

samrus2• 1 year ago

You are correct that review and validation should still be manual. But the actual "translation" from one format to another should be automated with llms

satyrun2• 1 year ago

People are also incredibly accurate at doing this kind of manual data piping all day.

There is also a reason that these jobs are already not automated. Many of these jobs you don't need language models. We could have automated them already but it is not worth someone to sign off on. I have been in this situation at a bank. I could have automated a process rather easily but the upside for me was a smaller team and no real gain while the downside was getting fired for a massive automated mistake if something went wrong.

TeMPOraL24• 1 year ago

An LLM won't call other nodes in the organization to check when it sees that the value is unreasonable for some out-of-context reason, like yesterday was a one-time-only bank holiday and so the value should be 0.

Why not? LLMs are the first kind of technology that can take this kind of global view. We're not making much use of it in this way just yet, but considering "out-of-context reasons" and taking a wider perspective is pretty much the defining aspect of LLMs as general-purpose AI tools. In time, I expect them to match humans on this (at least humans that care; it's not hard to match those who don't).

I do agree on the liability angle. This increasingly seems to be the main value a human brings to the table. It's not a new trend, though. See e.g. medicine, architecture, civil engineering - licensed professionals aren't doing the bulk of the work, but they're in the loop and well-compensated for verifying and signing off on the work done by less-paid technicians.

lottin• 1 year ago

You're vastly underestimating the complexity of these types of jobs.

tiahura2• 1 year ago

Everything? As a lawyer, I’m producing 2x - with fewer errors. Admittedly, law is a field that mostly involves shuffling words around so it may be the best case scenario, but much of the skepticism comes off as cope.

KerrAvon• 1 year ago

https://www.technologyreview.com/2025/05/20/1116823/how-ai-i...

https://hai.stanford.edu/news/hallucinating-law-legal-mistak...

https://www.reuters.com/technology/artificial-intelligence/a...

There are more of these stories every week. Are you using AI in a way that doesn’t allow you to be entrapped by this sort of thing?

tiahura2• 1 year ago

ChatGPT links to the actual text in a case now. Also, take the output from Claude, and put it into Gemini and tell it to verify holdings. Furthermore, spot checking 10 cases doesn’t take long.

logsr• 1 year ago

computer programming and law are very similar. computer code is called code because it is the law that dictates the behavior of the computer. law is a bit different because it is a program that runs on people, who aren't as deterministic as machines, but in theory law and the interpretation of law are also supposed to be completely logical, and you can translate back and forth directly from the logic of law to a logical expression in a computer program.

i specialize in programming, and LLMs are very good right now, if you set them up with the right tooling, feedback based learning methods, and efficient ways of capturing human input (review/approve/suggest/correct/etc).

with programming you have compilers and other static analysis tools that you can use to verify output. for law you need similar static analysis tooling, to verify things like citations, procedural scheduling, electronic filing, etc, but if you loop that tooling in with an llm, the llm will be able to correct errors automatically, and you will get to an agent that can take a statement of fact, find a cause of action, and file a pro se lawsuit for someone.

courts are going to be flooded with lawsuits, on a scale of 10-100X current case loads.

criminal defendants will be able to use a smart phone app to represent themselves, with an AI handling all of the filings and motions, monitoring the trial in real time, giving advice to the defendant on when to make motions and what to say, maximizing delay and cost for the state with maximum efficiency.

with 98% of convictions coming from guilty pleas (https://www.npr.org/2023/02/22/1158356619/plea-bargains-crim...) which are largely driven by not being able to afford the cost of legal services the number of criminal defendants electing to go to full jury trial could easily explode 10-20X or more very quickly.

fun times!

larve2• 1 year ago

Software methodologies and workflows are not engineering either, yet we spend a fair amount of time iterating and refining those. You can very much become better at prompt engineering. There is a huge differential between individuals, for example.

The code coming out of LLMs is just as deterministic as code coming out of humans, and despite humans being feckle beings, we still talk of software engineering.

As for LLMs, they are and will forever be "unknowable". The human mind just can't comprehend what a billion parameters trained on trillions of tokens under different regimes for months corresponds to. While science has to do microscopic steps towards understanding the brain, we still have methods to teach, learn, be creative, be rigorous, communicate that do work despite it being this "magical" organ.

With LLMs, you can be pretty rigorous. Benchmarks, evals, and just the vibes of day to day usage if you are a programmer, are not "wishful thinking", they are reasonably effective methods and the best we have.

sureglymop2• 1 year ago

Loosely related, but I find the use of AGI (and sometimes even AI) as terms annoying lately. Especially in scientific papers, where I would imagine everything to be well defined. If at least in how it is used in that paper.

So, why can't we just come up with some definition for what AGI is? We could then, say, logically prove that some AI fits that definition. Even if this doesn't seem practically useful, it's theoretically much more useful than just using that term with no meaning.

Instead it kind of feels like it's an escape hatch. On wikipedia we have "a type of ai that would match or surpass human capabilities across virtually all cognitive tasks". How could we measure that? What good is this if we can't prove that a system has this property?

Bit of a rant but I hope it's somewhat legible still.

AlienRobot• 1 year ago

We have a definition.

"AI is whatever hasn't been done yet."[1]

1. https://en.wikipedia.org/wiki/AI_effect

CharlesW7• 1 year ago

"AI" and "AGI" are very different things.

selfhoster115• 1 year ago

You don't need consensus on the meaning across the board. I maintain my own, more permissive milestone for what constitutes "AGI", but I have no expectations that others will share it. Much like "crypto" to me is still cryptography, not cryptocurrency - sometimes the mainstream will just have a different opinion.

sureglymop2• 1 year ago

My point is, it's not about the mainstream or marketing. In science, some rigor is expected. There doesn't need to be consensus if a definition is established within some context. It's perfectly fine to redefine something as needed for the research but only if that definition is declared.

AbrahamParangi• 1 year ago

This reads like the author is mad about imprecision in the discourse which is real but to be quite frank more rampant amongst detractors than promoters, who often have to deal with the flaws and limitations on a day to day basis.

The conclusion that everything around LLMs is magical thinking seems to be fairly hubristic to me given that in the last 5 years a set of previously borderline intractable problems have become completely or near completely solved, translation, transcription, and code generation (up to some scale), for instance.

troupo23• 1 year ago

but to be quite frank more rampant amongst detractors than promoters, who often have to deal with the flaws and limitations on a day to day basis.

"detractors" usually point to actual flaws. "promoters" usually uncritically hail LLMs as miracles capable of solving any problem in one go, without giving any specific details.

DavidPiper• 1 year ago

Translation, transcription, and code generation (up to some scale) were borderline intractable problems?

Google Translate, Whisper and Code Generators (up to some scale) have existed for quite some time without using LLMs.

famouswaffles• 1 year ago

Google Translate just spits out nonsense for distant language pairs (English<->Korean etc) and doesn't compare to Sota LLMs, Whisper is a Transformer (Architecture used for LLMs) and Code Generators have nothing on LLMs.

arendtio2• 1 year ago

I think it is more like googling: when the search engine appeared, everybody had to learn how to write a good query, even though the expectation was that everybody could use them.

With LLMs, it's quite similar: you have to learn how to use them. Yes, they are non-deterministic, but if you know how to use them, you can increase your chances of getting a good result dramatically. Often, this not only means articulating a task, but also looking at the bigger picture and asking yourself what tasks you should assign in the first place.

For example, I can ask the LLM to write software directly, or I can ask it to write user stories or prototypes and then take a multi-step approach to develop the software. This can make a huge difference in reliability.

And to be clear, I don't mean that every bad result is caused by not correctly handling the LLM (some models are simply poor at specific tasks), but rather that it is a significant factor to consider when evaluating results.

leptons4• 1 year ago

Yes, they are non-deterministic

The LLM is more like a Ouija board than a reliable tool.

I can ask it to write user stories or prototypes

By the time I write enough to explain thoroughly to an LLM what to write in "user stories" or "prototypes", I could have just written it myself, without the middleman(bot), and without the LLM hallucinating.

If half the time I spend with an LLM is telling it what to do, and then another half is correcting what it did, then I'm not really saving any time at all by using it.

selfhoster115• 1 year ago

Writing the prompt takes time, but in many cases it's like rubber-duck debugging - it helps you to figure out the actual requirements, flow, or even incorrect/missing assumptions. Is it worth the time? Subjective.

arendtio2• 1 year ago

For me, it is a different experience. Yes, it takes time to provide the LLM with the necessary context, but overall it can save me 3-4 times the time I would have invested. Maybe I am just slower at software development myself ;-)

However, the work you do is indeed more that of a product owner than that of a developer. To avoid hallucinations, providing extensive documentation and allowing the LLM to perform test-driven development can be a game-changer.

When you do that, generation time is highly correlated (negatively) with code quality. So when the AI solves the tasks quickly and easily, you have a good chance of it generating good code. As soon as the AI has to try and try again to build something working, you should be very skeptical of the result.

Over the past months and years, I have used this method, and because it is somewhat reproducible, you can see the progress that LLMs are making. Sessions where the model does stupid things are becoming fewer, and sessions where the model finds a good solution become more frequent.

alganet16• 1 year ago

LLM tech probably will find some legitimate use, but by then, everything will be filled with people misusing it.

Millions of beginner developers running with scissors in their hands, millions of investment in the garbage.

I don't think this can be reversed anymore, companies are all-in and pot commited.

Sherveen2• 1 year ago

This is completely incoherent. 3 reasons:

1. he talks about what he's shipped, and yet compares it to crypto – already, you're in a contradiction as to your relative comparison – you straight up shouldn't blog if you can't conceive that these two are opposing thoughts

2. this whole refrain from people of like, "SHOW ME your enterprise codebase that includes lots of LLM code" – HELLO, people who work at private companies CANNOT just reveal their codebase to you for internet points

3. anyone who has actually used these tools has now integrated them into their daily life on the order of millions of people and billions of dollars – unless you think all CEOs are in a grand conspiracy, lying about their teams adopting AI

infp_arborist• 1 year ago

Re. 2 - I find it fascinating that people using coding agents have seemingly no problem to reveal their full codebases to AI companies.

Sherveen2• 1 year ago

Can you clearly articulate me why they should have a problem?

If you don't want training on your codebase, many AI companies offer this option. What's the issue?

Ferrus914• 1 year ago

Some even integrating in such a way as to professionally humiliate himself or herself: https://www.hcrlaw.com/news-and-insights/in-house-lawyers-be...

ibaikov3• 1 year ago

Crypto and NFT situation happened because of our society, media and vc/startup landscape who hype things up a lot for their own reasons. We treat massive technologies as new brands of bottled water. Or, actually, as a new hype toy as fidget spinners or pop it toys. This tech is massively more complex and you have to invest time to learn about its abilities, limitations and potential developments. Almost nobody actually does this, it's easier to follow hype train and put money into something that grows up and looks cool without obvious cons. Crypto is cool for some stuff. On the other hand, where's your Stepn (and move to earn in general), decentraland cities, Apes that will make a multimedia universe? Where's "you'll be paying using crypto for everything"?

Same for LLMs and AI: it is awesome for some things and absolutely sucks for other things. Curiously tho, it feels like UX was solved by making chats, but it actually still sucks enormously, as with crypto. It is mostly sufficient for doing basic stuff. It is difficult to predict where we'll land on the curve of difficult (or expensive) vs abilities. I'd bet AI will get way more capable, but even now you can't really deny its usefulness.

CharlesW7• 1 year ago

It makes no sense to compare the current AI hype to the tulip mania of crypto/NFTs. A much better parallel is to cloud computing hype in 2009.

ibaikov3• 1 year ago

I think it does. Startups, VCs and media still provoke hype trains. People cherry pick and extrapolate everything in AI as well.

ares623• 1 year ago

I was just joining the workforce at that time, so I was unaware of any hype. From my perspective, cloud computing was just "there". Was the hype really comparable to today?

afiodorov• 1 year ago

We've been visited by alien intelligence that is simultaneously fascinating and underwhelming.

The real issue isn't the technology itself, but our complete inability to predict its competence. Our intuition for what should be hard or easy simply shatters. It can display superhuman breadth of knowledge, yet fail with a confident absurdity that, in a person, we'd label as malicious or delusional.

The discourse is stuck because we're trying to map a familiar psychology onto a system that has none. We haven't just built a new tool; we've built a new kind of intellectual blindness for ourselves.

localghost3000• 1 year ago

I've developed the following methodology with LLM's and "agentic" (what a dumb fucking word...) workflows:

I will use an LLM/agent if

- I need to get a bunch of coding done and I keep getting booked into meetings. I'll give it a task on my todo list and see how it did when I get done with said meeting(s). Maybe 40% of the time it will have done something I'll keep or just need to do a few tweaks to. YMMV though.

- I need to write up a bunch of dumb boilerplatey code. I've got my rules tuned so that it generally gets this kind of thing right.

- I need a stupid one off script or a little application to help me with a specific problem and I don't care about code quality or maintainability.

- Stack overflow replacement.

- I need to do something annoying but well understood. An XML serializer in Java for example.

- Unit tests. I'm questioning if this ones a good idea though outside of maybe doing some of the setup work though. I find I generally come to understand my code better through the exercise of writing up tests. Sometimes you're in a hurry though so...<shrug>

With any of the above, if it doesn't get me close to what I want within 2 or 3 tries, I just back off and do the work. I also avoid building things I don't fully understand. I'm not going to waste 3 hours to save 1 hour of coding.

I will not use an LLM if I need to do anything involving business logic and/or need to solve a novel problem. I also don't bother if I am working with novel tech. You'll get way more usable answers asking about Python then you will asking about Elm.

TL;DR - use your brain. Understand how this tech works, its limitations, AND its strengths.

hamilyon2• 1 year ago

I am impressed by speed of the sound goalpost movement.

Few days ago Google released very competent summary generator, interpreter between 10-s of languages, gpt-3 class general purpose assistant. Working locally on modest hardware. On 5 years old laptop, no discrete GPU.

It alone potentially saves so much toil, so much stupid work.

We also finally “solved computer vision”. Read from PDF, read diagrams and tables.

Local vision models are much less impressive and need some care to use. Give it 2 years.

I don't know if we can overhype it when it archives holy grail level on some important tasks.

assuagering4• 1 year ago

I have had SOTA models stray from factual content in documents I provided them with within 2-3 prompts.

They haven't solved anything. They are just fast and look good doing what we ask them to do. But they corrupt data with a passion and to that hype just responds: "just give us 10x as much money and compute".

ls-a• 1 year ago

AI helped me find a bug that was going undetected and also helped me fix it. I was debugging a completely different bug, and decided to get a bit creative with what context i should share. AI pointed out the other bug. What I'm trying to say is people should stop talking about AI making them faster, that is not a good goal, unless your manager sucks. People should talk about concrete cases of how AI helped them instead. Another example I can give is AI helped me understand a part of a complicated protocol without reading the spec. It explained just the part I needed at the moment. AI made me dread my work less. I hope founders start getting creative with AI tools instead of copying each other.

3cats-in-a-coat• 1 year ago

These polarized opinions that say AI is everything or nothing only reveal the emotional state of the writer, rather than any deep insights.

kgeist2• 1 year ago

We recently started using LLMs at our company, and the first job I had was to transcribe 20k customer calls and extract the following info:

1) what products we're usually compared to

2) what problems users have with our software

3) what use cases users mention most often

What used to take weeks of research took just a couple of hours. It helped us form a new strategy and brought real business value.

I see LLMs as just a natural language processing engine, and they're great at that. Some people overhype it, sure, but that doesn't change the fact that it's been genuinely useful for our cases. Not sure what's up with all those "LLM bad" articles. If it doesn't work for you, just move on. Why should anyone have to prove anything to anyone? It's just a tool.

hx8• 1 year ago

I think you are underestimating the negative impacts that overhype cause. It's distorting the market, causing over investment, preemptively slashing departments, and creating an expectation that will never be meet. These articles are important for cooling expectations. When people sell LLMs, they usually aren't talking about summarizing customer support calls, they are trying to sell the idea of firing customer support staff.

DebtDeflation2• 1 year ago

If anything, the LLM overhype is starting to die down........to make way for the AI Agent hype which is on trajectory to be 1000X worse. People are writing articles and making videos about how AI Agents will replace SaaS. What?

herbst2• 1 year ago

This very much. People who claim there is no real use of LLM never faced the problem of processing a lot of data in kinda reliably way.

For years most of the translations on the web didn't have context. Now they can have.

standardUser2• 1 year ago

Like most skeptics and critics, I use these tools daily. And 50% of the time they work 50% of the time.

I use LLMs nearly every day for my job as of about a year ago and they solve my issues about 90% of the time. I have a very hard time deciphering if these types of complaints about AI/LLMs should be taken seriously, or written off as irrational use patterns by some users. For example, I have never fed an LLM a codebase and expected it to work magic. I ask direct, specific questions at the edge of my understanding (not beyond it) and apply the solutions in a deliberate and testable manner.

if you're taking a different approach and complaining about LLMs, I'm inclined to think you're doing it wrong. And missing out on the actual magic, which is small, useful and fairly consistent.

leptons4• 1 year ago

Your comment is no better than the comment in the article that the author is calling out.

"90%" also seems a bit suspect.

richardw3• 1 year ago

I just went through the last 10 chat titles and all of them were spot on for me. Maybe the person you’re responding to has a different experience than you do and calling their perspective “suspect” is somewhat uncharitable.

(There are times I do other kinds of work and it fails terribly. My main point stands.)

standardUser2• 1 year ago

It either helps me find a solution or it doesn't. About 90% of the time, or less formally I would just say "almost all of the time", it does. Keep in mind that I, the user, decide which questions to ask in the first place. If my batting average seems unbelievably high, perhaps my skill is in knowing when to use an LLM and when not to.

geuis• 1 year ago

Hmm. Ok so you're basically quoting the line from The Weatherman "60% of the time, it works all of the time."

I also use gpt and Claude daily via cursor.

Gpt o3 is kinda good for general knowledge searches. Claude falls down all the time, but I've noticed that while it's spending tokens to jerk itself off, quite often it happens on the actual issue going on with out recognizing it.

Models are dumb and more idiot than idiot savant, but sometimes they hit on relevant items. As long as you personally have an idea of what you need to happen and treat LLMs like rat terriers in a farm field, you can utilize them properly

awkwabear2• 1 year ago

This seems like a really bad take.

I do PhD research for superconducting materials and right I've been adapting and scaling an existing segmentation model from a research paper for image processing to run multithreaded and took the training runtime per image from 55min to 2min. Yeah it was low hanging fruit but honestly its the type of thing that is just tedious and easy to make mistakes and spend forever debugging.

Like sure I could have done it myself but it would have taken me days to figure out and I would have had to test and read a ton of docs. Claude got it working in like half an hour and generated every data plot I could need. If I wanted to test out different strategies and optimizations, I could iterate through various strategies rapidly.

I don't really like to rely on AI a bunch but it indisputably is incredibly good at certain things. If I am just trying to get something done and don't need to worry about vulnerabilities as it is just data collection code that runs once, it saves a tremendous amount of time. I don't think it will outright replace developers but there is some room for it to expand the effectiveness of individual devs so long as they are actually providing oversight and not just letting it do stuff unchecked.

I think the larger issue is more how economically viable it is for businesses to spend a ton on electricity and compute for me to be able to use it like this for 20 bucks a month. There will be an inevitable enshittification of services once a lot of the spaces investors are dumping money are figured out to be dead ends and people start calling for returns on their investment.

Right now the cash is flowing cause business people don't fully understand what its good at or not but that's not gonna last forever.

phyzome• 1 year ago

I don't think you got the author's point.

They didn't say "AI is bad". Take another look.

awkwabear2• 1 year ago

Ah I'm sorry you're totally right, I was irresponsible and just skimmed the article last time.

I retract my last statement about it being a bad take

weare138• 1 year ago

Just wait until we get AI running on the blockchain in the metaverse. The future is here.

scrubs• 1 year ago

Op? Take a standing ovation and a victory lap. Well written.

geetee• 1 year ago

I'm working with product managers that are almost certainly using LLMs to generate product requirement docs, complete with code samples, data type definitions, and diagrams. Everything looks good to the untrained eye, but it's complete and utter bullshit. LLM abuse is going to be the end of so many tech companies.

bwfan123• 1 year ago

Everything looks good to the untrained eye

thats the trick with bullshit in general.

LLM abuse is going to be the end of so many tech companies

and is also going to provide a lot of opportunities for experienced engineers to cleanup the mess.

djoldman• 1 year ago

Many well-trusted and reasonable tech folks who are known for sober takes on subjects have reported substantial improvements in their programming work by using various forms of generative AI.

What does substantial mean? Somewhere between 5% and 100%. Something NOT insignificant.

At a minimum, it is safe to say that GenAI is or could be a significantly beneficial tool for a significant number of people.

It's not required that folks disclose how many CPUs, lines of code, numbers of bytes processed, or other details for the above to be a reasonable take.

troupo23• 1 year ago

"People claim to have productivity increases anywhere between a random number I invented to solve other number I invented. We should believe these claims uncritically"

orbital-decay4• 1 year ago

The point about non-determinism is moot if you understand how it works. An accurate LLM always gives the same result where the same result is needed, no matter how many times you ask it. Try asking any LLM what is 2x2 on a temperature it's designed for, what are the chances to get 5 in a reply?

In reality, modern LLMs trained with RL have terrible variance and mainly learn 1:1 mapping of ideas to ideas, which is a big issue for creative writing and parallel inference/majority voting techniques, so there's even less meaningful "non-determinism" available than you might think. It's usually either able or not able to give the correct answer, rerolling it doesn't work well. I think even a human has more non-determinism than a modern LLM (it's impossible to measure though).

troupo23• 1 year ago

Are those LLMs in the room with us now? ;)

Actually, I did try asking ollama running locally. That should've reduced the amount of non-determinism and whatever layers providers add, and the uncertainty of computer availability.

I asked it for a list of Javascript keywords ordered alphabetically. Within 5 minutes it produced a slightly different list

Asking a model for 2x2 is moot because 2x2=5 is statically highly unlikely. Anything more complex though?

orbital-decay4• 1 year ago

I asked it for a list of Javascript keywords ordered alphabetically. Within 5 minutes it produced a slightly different list

That's not my point, the keyword here is "meaningful". How many of those lists are correct? (ignoring the fact prompting a LLM for lists is a bad idea, let alone local ones)

If you spend some time with SotA LLMs, you'll see that on rerolling they express pretty much the same ideas in different ways, most of the time.

khazhoux4• 1 year ago

I'm honestly started to get pissed seeing these articles on HN every day. Aren't we technologists? I see these articles and it's just people who can't accept the future, or feel threatened by it, or are pissed they are not contributing to it.

I'm an expert in my field with decades of experience, and a damn good programmer, and Cursor is like zipping down the street on a powered bike. There is no wishful thinking here!

Can I use it for everything? No, and that applies to every tool. But ffs I have to read here every day on HN some self-important blogger telling me I'm imagining things.

troupo23• 1 year ago

I'm honestly started to get pissed seeing these articles on HN every day. Aren't we technologists?

Funnily enough, that's exactly what the article asks.

I'm an expert in my field with decades of experience, and a damn good programmer, and Cursor is like zipping down the street on a powered bike.

Oh look, an unverified claim. Go through the first list in the article, and ask "what we, as technologists, know about this claim".

You can also actually read the article and see that I actually use these tools every day.

khazhoux4• 1 year ago

Yeah, I read your article. It's the latest in the stream of "everyone is full of crap" about LLMs.

You say you've used it to build multiple projects, but then title it "everything is wishful thinking". That makes no sense. You're either click-baiting or have an inconsistent stance.

This is crypto all over again

No. The fundamental vision for BTC is that it grows to be a mainstream currency for a huge variety of use cases. If BTC continues to be used just for pirate sites, black market, etc, then it will be considered to have failed. With LLMs the success is immediate. We don't need LLMs to solve every software problem, for them to be massive accelerator of daily coding.

taylorallred• 1 year ago

This is why I’d love to see some more examples of people sitting down and using these tools in day-to-day work rather than just hearing “they’re great!” or “they suck!”

DeepYogurt• 1 year ago

Nuance. You love to see it

eachro• 1 year ago

"And 50% of the time they work 50% of the time."

I think this is still an incredible outcome given how many dice rolls you can take in parallel with multiple claude/o3/gemini attempts at a problem with slightly different prompts. Granted, each rollout does not come for free given the babysitting you need to do but the cost is much lower than going down the path yourself/having junior colleagues make the attempt.

Narciss• 1 year ago

Ok, I’ll take the bait and answer the questions you’ve set in the article.

Who am I? Senior engineer with 5 years of experience, working within the AI team.

Project: social platform with millions of MAU

Codebase: Typescript, NextJS frontend, express backend, Prisma ORM, monorepo . Hundreds of thousands of lines of code.

My expertise sits in the same domain, language, codebase that I apply AI to.

When do I use AI? All the time. It has significantly sped me up.

How do I use AI? Almost always in a targeted fashion. I line it up like a sniper rifle - I understand my problem, where in the codebase it should act in order to apply it, and what general structure I want. Then I create the prompt to get me there. Before, I needed to do the above anyway, minus the prompt, but then I needed to use my slow hands to code dozens or hundreds of lines of code to get there. Now, the AI does it for me, and it makes me much faster. To clarify, I do not plan everything in the prompt - I let the AI color between the lines, but I do give it the general direction. Of course, I also use AI for smaller tasks like writing SQL queries for example, and it’s saved tons of time there too.

In my side projects, I have a different approach - I vibe code my way through with less “sniper targeting”, because I care less about the beauty of the code organization and more about results. If I do see AI slop, I then ask the AI to clean it up, or redo the prompt with more targeting.

Overall, AI has significantly sped me up. I am still heavily involved as you can tell, but it is a phenomenal tool, and I can envision it taking over more and more of the code writing process, with the right rules in place (we are actively working on that at my work, with md files and rebuilding our FE from the ground up with AI in mind).

Does this sort of answer give you a better idea of how AI helps?

cedws• 1 year ago

I still see little conversation about the two fundamental limitations of LLMs right now: context size, and prompt injection.

* Computation does not scale linearly with context size, meaning the ‘memory’ of LLMs is limited and gets more expensive as it gets bigger.

* Prompt injection limits the usability of LLMs in the real world. How can you put an LLM in the driving seat if malicious actors can talk it into doing something it’s not supposed to.

Whenever I see a blog post by Anthropic or OpenAI I do a Ctrl+F for “prompt injection.” Never mentioned. They want people to forget this is a problem — because it’s a massive one.

rickvidallon• 1 year ago

In a brief nutshell, I'm personally not in favor of how AI is taking over the industry (all of industry, that is). First of all, LLMs aren't even really what I would call AI, but that's a digression. I think, mainly, people are using it for the wrong reasons. I've seen graphics designers at work use it to generate images—and I don't just mean Adobe Firefly generative fill, but all-out "create me this logo." That's just lazy, given their job titles, and has started making everything look samey. Worse than how Bootstrap, Tailwind, etc., has the tendency to make everything samey.

For someone like me, who is self-taught on essentially all of his career skills, I have particular concern for a world in which people use AI to "learn things," when that tech doesn't allow them to make mistakes. It just does things for people. For that reason alone, I don't see AI as a viable way to learn at all. If your parents never take their hand from the back of the bicycle seat, do you really know how to ride a bike without falling over? Isn't the scraped knee how we truly mastered that skill?

For SEO in particular, I'd probably defer to someone with more daily experience, like you. That said, I think I can extrapolate from what I've seen elsewhere that the sameyness may likely start to affect content itself (in fact, it already has, for so many formerly good news outlets). Google search kinda blows in recent years. The AI Overview feature means so many people aren't visiting the source website anyway.

To me, none of this looks appetizing. It looks like a snake eating its own tail.

I don't mean to sound so bummer about the topic, but I've begun to worry about my own place in this ecosystem for the next 15, 20 years until I retire. Most of the joy in development has been sucked out of the art. Today, it seems mostly about getting wrapped around the axle of countless frameworks (without even really understanding them) and manhandling those chocolates on Lucy's and Ethel's conveyor belt. It's a comedy of errors. I'll yell "Get off my lawn" with the best of them. Add the same issues to SEO, and I don't really know where we'll end up, but it doesn't look creative, to me. It looks like a sad cliché, like the rows upon rows of sad souvenir shops that all kind of sell the same thing—the tourist trap that travelers (at least, travelers like me) actually loathe and try to avoid.

Ferrus914• 1 year ago

Like so much of everything produced now. Mainstream film and video games have been non-AI slop for almost 2 decades now.

Everything around LLMs is still magical and wishful thinking

Related Stories