YHK - yet another hacker news clone

If I ask you a question and you turn around and ask AI you get to replace me with AI for all future interactions.

Hasn't the same always applied to googling on behalf of others? That's why LMGTFY exists too...

I think we have the moral imperative to document all "out of distribution" information somewhere. By this I mean anything that is tribal knowledge or principles that just exist within our heads.

> Maybe they care about you, and want an excuse to talk to you, and free associated from their problem to you because they think of you as generally competent.

This is why I hate when people paste a response from Claude/ChatGPT to me when I ask them a question.

The author's a sucker. YT premium pays for itself many times over.

Except when the question is actually interesting and has less about it written, whatever frontier model you are using may confidently give some partially correct or completely wrong answer. Unless you're an expert who wouldn't need to ask AI in the first place, you cannot tell if it's false. "Oh but it's good enough, it's correct most of the time."

This gradual accruing of more and more falsehood will hurt us in the coming years, big time.

EDIT: I just asked Claude a very basic question on an area I'm an expert in and it gave me a partially true and misleading answer. I drilled in on the false part and it wiggled and gave more confusing, incorrect, poor justifications for why it gave me the clearly wrong information. If someone was not an expert, there is absolutely no way they could fact-check this. Instead, the falsehood would be written, published somewhere, and would serve as a base for more falsehood in the future.

We've seen this song and dance with Google already.

Clearly, for some people, "type into a search bar" is some sort of cognitive obstacle, and "go bother another person" isn't.

Now that I'm thinking about it: maybe the likes of OpenAI have the right idea when they invest into conversational models? Does it meaningfully reduce the workload of flesh proxies?

Another comment: you can just tell the person asking you that you don't know the answer. You don't have to look it up for them in any way; it's nice if you do, but you're allowed to just say "Sorry, I don't know! I'd look online, or maybe ask an LLM or something."

You don't have to start doing their research for them, or copying and pasting or transcribing Claude's answer.

I was thinking the other day as a kid I would always pester my parents with questions like "how does lightning work?" etc etc. My world view was formed with answers based on what they understood. And sometimes I'm sure they'd just make something up when they didn't know the answers.

Anyway, I guess that's so different now when kids have internet devices in their hands and now even moreso with an AI assistant that has the worlds knowledge

Fun fact ... if we continue training LLMs on human data, in the future when you ask Claude a question it will say: "Have you asked Claude?"

The "taking credit" thing rings true for sure.

I have trouble saying "I" when an LLM has done the leg work.

"I found a report with this info..." turns into "There's a report with this info". heh.

That comment on the substack is incredibly depressing. And the author liked it! That commenter should call his mom more.

It’s like when someone asks you for directions. They can easily use Google Maps, but they like your input.

This many levels of irony and insincerity and condescension just seem... not very useful? Why did you want me to read this?

You'd be surprised how many people, and I'm not talking about 80 year olds, have a hard time using search engines, or general knowledge of finding information.

If you're not careful, it is also super easy to extrapolate your own skills - and your peers skills - to the rest of the population.

Try being technical support to some non-tech people over the phone, it'll really open your eyes to the general computer knowledge of people.

I've found that many "skills" with regard to using AI become obsolete very quickly. It's really hard to stay ahead of the curve, IMO. I was using agents over 2 years ago and felt way ahead of the curve [1]. Today, Claude will spin up agents without even having been asked.

Feels a bit like a fools errand.

[1] https://www.loom.com/share/aa6e55acb18c445e9186637ac763c08d

There's another problem with people who know enough to ask an LLM to do something, don't know how to actually turn that into something useful, but want to act like they already did the heavy lifting.

"I had an idea for a new app. Here's a set of screen mockups and a requirements doc I did with Claude. Can you go ahead and finish that off now that I've done the design?"