Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Don't be a meat proxy

gruhn.me

Post reply on HN
Print view View original

751–760 of 787 posts

Re: Don't be a meat proxy

#751

Post by skinwill »

natsucks wrote: post #573
binary0010 wrote: post #371

Earlier quoted context omitted.

Whenever people compare ai to a one real world tool always surprises me tbh. The only equivalent tool in wood working would be a robot that can build an entire table from scratch and you just sit there and tell it what to do and what kinda table you want and watch it autonomously build everything. 1 saw type is not really even remotely equivalent to what ai is to software.

Yeah, AI is a general purpose technology, on a par with the introduction of the computer itself. Analogies to tools seems inappropriate.

My comment wasn't about the tool, it was how people are treating the new tool. Much as they would when they learn other new tools. They tend to have a preference and apply it where they maybe shouldn't. Do people really need to ask AI everything like simple math problems or what goes best with this dress? Is that really a good use for early generation, water hungry, huge datacenter AI?

Re: Don't be a meat proxy

#752

Post by jurgenburgen »

marliechiller wrote: post #599
jurgenburgen wrote: post #274

Earlier quoted context omitted.

> Is it not more akin to asking a handyman to go and do the work for you? Does your handyman go off the rails and build pipes around the neighbors house because they didn’t know you already have an incoming water pipe? Does your handyman require you to have expertise in plumbing to make sure they don’t use bubble gum to seal the pipes? It’s not taking credit if the “handyman” is a gibbering idiot who needs more handh…

That seems like an extreme interpretation if I'm honest. As ever, its more likely somewhere in between your hammer analogy and the handyman. In any case, there is almost zero downside to listing your sources when answering a question as it provides the person asking the question with more context as to the origin of the answer

If you provide sources, provide the sources the LLM found for you after checking them yourself. Otherwise the recipient will have to vet them.

Slight rant: I’ve started to get a bit fed up with people antrhopomorphizing these things, it’s unhealthy. It’s definitely not a handyman and it really is just a tool. I know they’re designed to seem conscious and that just makes them dangerous to use. If my colleague starts treating his hammer as a person, it’s definitely not normal.

Re: Don't be a meat proxy

#753

Post by jagenabler2 »

wuliwong wrote: post #610
gregsadetsky wrote: post #243

On social media, I saw the much more vulgar “Learned engineering just to become the condom between Claude Code and prod” And that (re)framing helped as well to think about the “what are we even (left) doing” as an industry

It's interesting to me, I have never once felt this. I have basically just been excited by using AI for coding. I've been a software engineer for 15 years and before that a physics PH.D. student (and 1 year professional scientist) who focused largely on computational work. I love programming and I've felt reinvigorated by the introduction of AI and developing systems around it. I tend to think it's just different per…

I’m noticing a big divide between engineers who work in computational/library work and product work. And maybe a lot of this is coming from how we have to interact with people and how requirements are formed.

I’m the latter, have been working in product my whole career. My role has been architecture/tech lead for a while. Managing teams has become significantly more difficult. A lot of the implicit forces that hold engineering orgs together are falling apart.

Well-described greenfield work has always been non-existent. Plans starts ambiguously, and then they evolve as the team’s understanding of the problem gets deeper. Now everyone I work with has no interest in the deeper understanding, I feel like I have to “force” people to think. This is a losing battle and I’ve mostly given up on building good product. Customer asks are just fed into Claude and a PR is born. Cohesiveness, maintenance, and complexity are an afterthought.

For example, a customer recently asked for more information on a data table in our product. An engineer fed the request into Claude and generated a stack of 3 PRs totaling 1,200 lines with new endpoints and UI components. All the table needed was a tooltip.

The easiness of getting to a “final product” has totally eroded team collaboration. Before, the engineer would probably ask someone with more understanding of the domain and learn _why_ the tooltip is all they needed. This creates two domain experts. Now, we are ensuring we have none.

Re: Don't be a meat proxy

#754

Post by zumtrotz »

kooi wrote: post #504
jvanderbot wrote: post #429

Earlier quoted context omitted.

Sure, but even worse, is the state-of-the-art: they use a cli agent and the "copy-paste, try to run, paste error, try to run" loop is called "Agentic engineering." Now we have to believe they're 10/10? I see no difference there except speed at the cost of whatever little understanding may have been gained by the manual inspection between steps.

Any competent developer knows there are limits to prompting. Personally I've found LLM's suck at multi-threaded applications. (Because I've been tempted by the ~agentic loop~ and been burned. Then I hand code the core logic and all is well). Woe to the developer who tries to prompt their way through this.

They suck at multi-threading, they suck at maintaining a coherent goal through multiple files, they suck at basic math (claude could NOT figure out how to get the hypotenuse of a triangle given angle and side lengths...).

Sometimes I mess around with prompts just to see what I get back, more often than not, it's mostly a waste of time. God forbid you use it to refactor anything.

Re: Don't be a meat proxy

#755

Post by remich »

defgeneric wrote: post #721
danolivo wrote: post #471

A while back, I noticed that using Claude all day left me feeling mentally tired. After thinking it over, I realised the problem was Claude’s complex language. I’ve used English daily for years, so that shouldn’t be an issue. To compare, I switched Claude to my native language and had the same problem the author described. The text still needed to be literally deciphered before I could understand it. After talking wi…

I've seen this sentiment so much in the past month that it's starting to make me worried that the labs will silently make this the default output to satisfy users. I do recognize the outputs can be so information-dense that it strains reading for many but they absolutely aren't nonsense or mere affectation. What people are asking for now amounts to asking for less intelligence per token only a few months after everyo…

Yeah I worry about this too. I have a remote agent that writes PRs for my job, and at the request of other devs I implemented a simplified writing skill that does the technical english thing mentioned elsewhere in these comments. It's definitely easier to read and the other devs seem to like it, so I'm okay with keeping it, but I haven't changed my personal claude settings because I never had a problem with the density of the output myself. Probably a side effect from having a multi-disciplinary background.

Re: Don't be a meat proxy

#756

Post by javcasas »

DenisM wrote: post #702
javcasas wrote: post #566

Earlier quoted context omitted.

I am that person on the blast radius. The most effective response I have found is dumping the documents into the AI, and asking it to review them, then posting back the AI response. Any resistance to this is met with a "But I'm also using AI as the board demands!"

Diabolical. I’m not sure if I’m frightened or fascinated… , I’m thinking of where can this lead if applied iteratively. Like, slop supernova followed by by a slop brown dwarf?

I call it "surviving by fighting fire with fire".

Re: Don't be a meat proxy

#757

Post by spencerflem »

mattgreenrocks wrote: post #678
jvanderbot wrote: post #429

Earlier quoted context omitted.

Sure, but even worse, is the state-of-the-art: they use a cli agent and the "copy-paste, try to run, paste error, try to run" loop is called "Agentic engineering." Now we have to believe they're 10/10? I see no difference there except speed at the cost of whatever little understanding may have been gained by the manual inspection between steps.

The industry has always found it strangely taboo to actually talk about the wide variance in developer skill.

Is it? I hear about 10xers all the time

Re: Don't be a meat proxy

#758

Post by sourcecodeplz »

eddythompson80 wrote: post #8

I deal with this all day long at work and it’s exhausting. People almost acting like no one has thought of it “I asked Claude what happened, and it spit out this 300 line response. Can you read it for me and see if it’s right?” What kills me is you might expect this from a busy high level manager that doesn’t really understand the technical details and they just point the AI to an error they got. They don’t know how…

its okay because it worked for them once, they actually got valuable info from Claude and passed it along.

the person that got it, thought it was actually a good idea and they went with it.

that confidence now translates into behavior.

Re: Don't be a meat proxy

#759

Post by sourcecodeplz »

impulsivepuppet wrote: post #97
jasonkester wrote: post #61

Earlier quoted context omitted.

I just can’t understand why this is not a fireable offense. Like, I can see getting a strong warning the first time somebody notices you do this, then if you persist you’d have to be let go. Just like anybody else who is clearly not doing their job. Why would you tolerate this sort of behavior at your company?

what is often overlooked in corporate culture: if you take a step back, two coworkers refusing to cooperate, regardless of the reason, is a typical "corporate kindergarten" scenario for the manager to take care of. To avoid rushed or incorrect decisions, conflicts intentionally move through gradual escalation. If someone behaves unprofessionally, for example by submitting a code review they do not understand, the fir…

this is a great comment.

(name doesnt check out though)

Re: Don't be a meat proxy

#760

Post by biztos »

NorthSouthNorth wrote: post #670
biztos wrote: post #608

Earlier quoted context omitted.

That's a great idea. I've been thinking about how to avoid having AI-isms creep into my own voice, because I see that way too much in my tech-world friends. It's natural enough: if you spend a lot of time talking to somebody, you usually end up mirroring some of their speech patterns. One idea I had was to start doing my Claude sessions in German. (That happens to be my strongest second language, no other reason.) Th…

+1 on this being a great idea. I also need to learn how to write in German (I can speak at B2 and read C1, but I find writing difficult as I've never bothered tbh). The added benefit of avoiding Claudisms is interesting. I think I'll join you on this.

FWIW I did a day of German-language software design with Claude and it was great, a bit strenuous of course but I also wonder if it might have found some bugs it would have overlooked in English.

If you're doing it for writing practice, one of the good/bad things is that it's quite typo-tolerant (in any language). So you can have the chat with less effort, at the risk of practicing bad habits.

Post reply on HN