pgwhalen

So if the matters to you, run the LTS release, right? I'm not sure I follow the concern.

Distinguished version: write the outline of the slide deck for the talk you plan to give at conferences about it, without having shipped anything or even written code yet.

These takes Arne necessarily incompatible. It can be a great tool for helping you do this kind of big picture design, but still need you as a guide and taste-maker to get to a good end result.

It’s worrisome, but it’s also worth pointing out that no one else has succeeded at it, or even gotten close. It’s easy to forget that there are not good well-integrated person AI solutions yet, it’s just chat bots. I think it’s just harder than a lot of people think it is.

If it helps, I could have written this comment just about word for word, and you can check my account and see that I’m clearly human.

I would probably add that I’m nervous about AI search results and how it affects the future of the internet and content creation in general, but from the perspective of a user, I’m pleased with the direction.

Oh I'm sorry, I didn't see you were different than the commenter I was responding to.

Can you link to it? Your comment history does not show such a comment.

Can you explain why? I'm getting better at detecting some kinds of AI writing, but I constantly seem comments like this on HN for things I'm much less suspicious of, and I want to understand why people make them.

Did you read the article? They did not believe that the token the agent had access to had the ability to delete production data using it.

Who is building an artificial wall? Maybe I skimmed the post too fast, but it doesn't seem like this information is being presented as "you have to know/do this before you start agentic engineering", just "this is some stuff to know."

As not a shill myself, who comments on a variety of stuff regularly, I could totally see why the mention of a product someone loves is the thing that pulls them out of lurking.

It’s just how discussion on the internet works, for basically anything at all worth discussing. It’s exhausting, but I can hardly blame HN specifically.

You could break it down further if you like, yes.

Everything is complicated by money. I wish we were better about shielding politics from money. So much about society in general is about money, it ain’t easy.

Lobbying is literally half of what representative democracy is. First, you elect representatives to office. Then, you try to get them to do what you want. The latter is lobbying.

Of course, when money becomes a significant portion of how the second one happens, things can get complicated.

Isn't this post an announcement of time-based billing? Just in a kind of indirect way (not billing, rather than billing).

Also, my (extremely naive) understanding is that at the cutting edge, hardware is diverging for training vs inference. That might not be true for Anthropic though.

While it may not have “agency” it definitely doesn’t necessarily do what you tell it. I’d put it as “it may do what you let it.”

I really am quite in awe of Claude Code recently, so definitely not a naysayer, but this is a really important point. It’s so easy to create code, but am I shipping that much to prod than I used to? A bit.

Obviously this highly depends on your company and your setup and risk tolerance and what not.

I’m a heavy user for about four months now, and it’s definitely getting better for me. How would you say it’s getting worse?

What is that spend compared against though? They already spend hundreds of billions of on various things in a year, but what is the marginal spend?

When you present these numbers alongside each other, you imply that they will go from making ~$20b/quarter to losing ~$30b/quarter, which is not plausible to me.

They literally don’t have the cash to do it.

I don’t understand this? Amazon is a profitable company, on the scale of tens of billions of dollars per quarter. They very literally do have the cash.

Am I missing some subtlety in their financial reporting?

Is this how comping actually works? I’ve never worked in a restaurant, but I assumed there was some system for it (if sometimes ill-defined) and not just employees stealing.

Thanks for this link. I’ve failed to find specifics on this for a while but this is pretty good, particularly the example about which lane to choose when cones are set up.

This won’t happen with general software though, that ship has sailed, the space is too competitive.

I think he would too, but they’re obviously not going to do that.

The top of the readme links to the accompanying paper, which explains very clearly why you would care: https://badrish.net/papers/bftree-vldb2024.pdf

I admit I’ll agree that that extra hop was a little confusing to me though. I guess people just like GitHub and don’t like PDFs.

This is the second time I’ve seen Goedecke criticized as cynical and honestly it quite baffles me, I see it almost completely the opposite. His writing acknowledges the common cynical views of working at large companies but then works to rationalize them, in a pragmatic way.

Like https://www.securityweek.com/hackers-target-popular-nx-build... ?

I only had time to skim this, but it doesn't seem like prompt injection to me, just good old fashioned malware in a node package.

Your other two examples do seem to open the door for prompt injection, I was just asking about documented cases of it succeeding.

Thanks for sharing the example!

Going back to typing all of the code yourself (my interpretation of "writing by hand") because you don't have the agent-managerial skills to tell the coding agents how to clean up the mess they made feels short-sighted to me.

I increasingly feel a sort of "guilt" when going back and forth between agent-coding and writing it myself. When the agent didn't structure the code the way I wanted, or it just needs overall cleanup, my frustration will get the best of me and I will spend too much time writing code manually or refactoring using traditional tools (IntelliJ). It's clear to me that with current tooling some of this type of work is still necessary, but I'm trying to check myself about whether a certain task really requires my manual intervention, or whether the agent could manage it faster.

Knowing how to manage this back and forth reinforces a view I've seen you espouse: we have to practice and really understand agentic coding tools to get good at working with them, and it's a complete error to just complain and wait until they get "good enough" - they're already really good right now if you know how to manage them.

I was hoping for a real world example.