Hacker Newsnew | past | comments | ask | show | jobs | submit | thunky's commentslogin

> And then when you come back after a few days, you realize that it has overwritten

1. Review what it writes as it writes it

2. Give it a smaller, more focused, scope to make changes

Or, keep vibing but then don't get mad about it.


Oh, if I had the ability to do these things, agents would be much nicer to use.

Sadly, at my current company, this doesn't seem to be acceptable practice.


What do you mean? Because management says it makes you too slow? Or is it unacceptable for some other reason?

> that they lack basic skills, such as knowing their own zip code

Knowing your zip code isn't a skill though.

I don't know my ATM pin. Does that make me an idiot? Turns out I just don't need to know it.


Sure, remembering basic facts about your life isn't a "skill", per se. But it's pretty crazy to think that Jensen literally doesn't know his own address.

But again, I think you're missing the point of the article. The point of him bragging that he can't remember basic stuff about his own life is both in-group signaling and a test to the interviewer. If the interviewer pushes back and asks "You really don't know your zip code? Are you insane?", the interviewer is never going to get any more time with any of these rich and powerful people. It's an easy way to filter out any critical coverage and a way for the media to show fealty.

The title of the podcast episode is "Jensen Huang vs the AI Doomers", not "Shovel seller is so rich he cannot conceive of what it's like to be a normal person".


Your comment inspired me to read the article.

I've done this despite also feeling silly about it, but then was surprised to get value from the response. Like something I didn't think about or would have otherwise forgotten to do.

So now I do it regularly.


> We need new words!

The words we have are fine.

We just need to assign liability by ownership/initiation: if your "agent" destroys something, even though you didn't tell it to (because it had "agency"), you should be liable for the damages.


>> We need new words!

Can make distinctions and can choose actions - applies to both humans and AI. I'd replace 'agency' with 'distinction & choice' language.


I believe their complaint is between "agents" and (them not having) "agency" — by definition, agent is something which has agency.

If you want to use "distinction & choice", you'd need a new word for an "agent" too.

I actually like the appropriatelly directional "harness".


Because a PE is a fact not an interpretation.

It's interesting how people who stare only at P/E make consistently bad investment decisions. Perhaps P/E is not such an important indicator overall?

I agree. But that still doesn't make it a lie or illegal.

As i write this, an article about Claude is 6th from the top on the linked site.


this website reduces the priority of content, rather than filter it entirely, so some articles with high score will still appear, but lower down than they otherwise would.

This is configurable though, so you can make it whatever you want.

see below for a complete demo (view on desktop): https://sprinklz.io/demo/hackerNews

as well as another example I've exported from above with stronger penalty: https://sprinklz.io/public/5vsn97zcgxex

you can also visit the main homepage at https://sprinklz.io if you want to build/save/export arbitrary feeds


A curved wall?


Well look, the wall is already losing every point by volleying your serve so I feel like I'm on pretty strong ground here already. But give me an exact specification for your curved wall and I'll tell you how I'd beat it.


Yes. And right after we kill golf lets pave over Central Park. And we can tear down sports arenas and concert venues after that. It's a huge waste when you think about how few people actually go there and what benefit they actually have. Sports overall are a waste of time. Everyone should just walk in the woods instead. Except soccer. You're right that we need thousands more soccer fields. Its way more efficient than any other kind of exercise. And there is so much demand for soccer in the US, you can't even find an empty field. In contrast to golf courses which are mostly empty like you noticed.


go ahead and reductio ad absurdum.

you aren't countering my core point that golf courses is the possibly the worst use of space and basically anything else would serve the same or better utility for way more people.

heck keep it golf but switch to driving ranges and its still like 10x more capacity for users than golf courses.

counter arguments that point out other things that are still 10x better uses of space than golf are not very compelling. a sports arena that serves 2.5m people/year is a terrible argument against a golf course that maxes out at 117k. (using sofi in LA numbers as its a similiar footprint as well 150acres)

im also not saying abolish golf, just only play it in places where the space/water use dont matter. like rural scotland. don't put golf courses in/near cities where we should use our space better.


> you aren't countering my core point that golf courses is the possibly the worst use of space and basically anything else would serve the same or better utility for way more people

My counter point was that everything we do doesn't have to be motivated by efficiency gains. Sports, golf especially, isn't trying to cram the most people possible into a small space.

By your own logic, we should replace soccer fields with kicking a ball into a net.


> Guess we’ll need even more human software engineers to fix it.

This is what we've been doing for years before LLMs. Hiring a ton of devs to spit out code, which became another's burden. It just took longer and cost more.


Major critical software has always needed thousands of engineers to maintain million line codebases. LLMs can't reason across edge case failure modes, and better yet, LLMs generate compartmentalized modules which exponentially explodes the number of failure modes as LLM based codebases grow. Read how everyone here laments about Sol or Opus or GPT6 generates 10x the required number of lines


> LLMs generate compartmentalized modules which exponentially explodes the number of failure modes

It sounds like you're arguing against modularizing which is an unusual stance. Especially for a million line codebase.



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: