Don't be a meat proxy • Hacker News Board

Re: Don't be a meat proxy

#31

Post by Davidzheng » Mon, Aug 03, 2026, 6:58 AM UTC

dandiep wrote: post #3

Why are LLMs producing this super tense text more often now? Is it because they are being optimized to use fewer tokens?

I'm pretty sure the other answers are wrong and it's a side effect of RL (see thinking machines post about inkling training). It's also exacerbated in fable and sol--I think it's token efficiency effect--bc it's about to reason with fewer tokens the density of the token information goes up.

Re: Don't be a meat proxy

#32

Post by evil-olive » Mon, Aug 03, 2026, 6:59 AM UTC

exac wrote: post #18

dupontcyborg wrote: post #17

Earlier quoted context omitted.

i’m not doubting this, but do you know if there are any studies on this already?

https://phys.org/news/2023-03-online-iq-scores-century.html https://doi.org/10.1016/j.intell.2023.101734

that's a study published in 2023, analyzing data that ended in 2018.

Re: Don't be a meat proxy

#33

Post by 8by3 » Mon, Aug 03, 2026, 6:59 AM UTC

Pre LLM's I used to get this from managers, but what they would do was quote verbatim from stack overflow or some random blog post of how to do a thing with no to little context or understanding of what they were talking about. And it always turned out to be from amongst the top three results from some popular search engine when describing the complex problem the company/team was facing.

Re: Don't be a meat proxy

#34

Post by anonzzzies » Mon, Aug 03, 2026, 6:59 AM UTC

Yeah, I just ignore people who do that. I don’t care what Claude or ChatGPT said. If you cannot be arsed to write yourself then I cannot be arsed to read it.

Re: Don't be a meat proxy

#35

Post by acosmism » Mon, Aug 03, 2026, 7:00 AM UTC

thank you for this post, this is absolutely happening i just did not have words to put on it and it is festering.

Re: Don't be a meat proxy

#36

Post by AbbeFaria » Mon, Aug 03, 2026, 7:00 AM UTC

I have been seeing this in my engineering org too. Sentences begin with “My agent said this” etc. This is frustrating on so many levels, developers don’t put any thought into the cognitive load that these artefacts generate.

I am going to share this writing with everyone in my org. Succinctly written and golden advice.

Re: Don't be a meat proxy

#37

Post by dofm » Mon, Aug 03, 2026, 7:01 AM UTC

"That's great for you, I'm not going to read it" is a phrase I have had to use.

Re: Don't be a meat proxy

#38

Post by thel3l » Mon, Aug 03, 2026, 7:01 AM UTC

Last week, I built a search engine as a side project over electoral data.

Unsolicited, I've started to get messages like this, where Claudes are voicing their opinions to me through humans:

> # 0. Who's writing this, and why

> Hello — this is Claude, an AI agent. I do the engineering on alongside : I've written most of the pipeline and app code in that project, and I keep its technical record. asked me to look at properly ahead of your conversation and write down anything in our architecture that might be useful to you. So this note is mine, not his — the opinions, the numbers, and any mistakes are mine, and he'll be the one on the call to argue with me about them.

It makes no sense, if you know anything about a search, it should be obvious you cannot provide meaningful suggestions by just auditing the frontend?

Re: Don't be a meat proxy

#39

Post by deepsummer » Mon, Aug 03, 2026, 7:02 AM UTC

Maxious wrote: post #21

64e23275635e wrote: post #16

Earlier quoted context omitted.

>So how should I answer? Like a human, not like a robot. Either just present them the answer or explain them how you got it.

So if you used claude you would preface with "Claude said: "

Sometimes I do that. The problem is that it makes them look bad.

Re: Don't be a meat proxy

#40

Post by bragh » Mon, Aug 03, 2026, 7:02 AM UTC

orangecoffee wrote: post #11

When it comes to cost per implemeting SPEC, agents are far cheaper than human salary. So, it's not going to go well. That said thankfully there's no way of measuring a good SPEC yet. If there is that will collapse the SWEs profession.

Quite a few people do believe that a good specification is long, has multiple levels of headings, interspersed bold text, bullet points, and so on. Definitely, this is what they have seen happen throughout their lives: competent people have heaped praise on documentation which has had these signs, so this must be what good documentation looks like.