Hacker News Reader

Care to share your setup?

rzkon Ask HN: How to deal with shipping AI-written code you don't understand?

A lot of comments here saying “don’t do it.” I think that’s shortsighted. The speed gains from LLMs are massive, and the code they produce (Opus 5.5 at least, in my experience) is correct most of the time. Once a developer or a company realizes that, there’s no going back. There are two problems, however:

1. If you keep shipping AI-written code you don’t understand, you become entirely dependent on the AI to maintain it, because you lack the mental model you’d need to jump into it quickly.

2. You can afford to do this as a senior, because you can tell when the AI does something suboptimal. But how is a junior supposed to develop that intuition when all they start out with is AI-generated code?

rzkon Updated Google Maps shows destruction of the city of Rafah

> but often we don't acknowledge the horror of Hamas on Israel and Palestinians when talking about this conflict

Often we don't acknowledge the horror of Israel's multi-decade siege on Gaza¹ and occupation of Palestinians when talking about this conflict.

There, I fixed it for you.

[1] As early as 1967, U.S. Senator Albert Gore Sr. called Gaza “a vast concentration camp on the sand”.

rzkon Anonymity Trilemma: Strong Anonymity, Low Bandwidth Overhead, Low Latency–Pick 2 [pdf]

Slides: https://petsymposium.org/2019/files/hotpets/slides/coordinat...

rzkon Play GTA Vice City in the Browser

Thanks, edited.

rzkon Anthropic Sued Over Limits on Its $200-a-Month AI Plans

https://archive.ph/nRaAL

rzkon Coordination Headwind: How Organizations Are Like Slime Molds

Here is an article by the author that expands on the ideas presented in the slides:

On Schelling Points in Organizations — https://medium.com/@komorama/on-schelling-points-in-organiza...

rzkon Singapore offers parents $55,000 in battle to boost birth rates

https://archive.ph/uuTKn

rzkon Nvidia AVO achieves 100% in ARC-AGI-3

NVIDIA AVO's 100% is only on ARC-AGI-3 public demo set.

rzkon And then the men with guns tell you to do it anyway

And then the men with guns ask you to put your name in the "sender" field.

rzkon Claude Is Down

Claude is down, but their status page doesn't mention it yet: "All Systems Operational" – https://status.claude.com.

rzkon AI isn’t outthinking mathematicians, it’s out-remembering them

From the article:

The physicist Eugene Wigner, who knew both John von Neumann and Albert Einstein, wrote that no one he had encountered possessed a mind as “quick and acute” as von Neumann’s. Von Neumann could absorb vast amounts of information, follow extraordinarily complicated arguments and move between mathematical fields with astonishing speed. Yet Wigner still regarded Einstein’s understanding as deeper, more penetrating and more original. Von Neumann may have had the greater raw intellectual processing capacity, but Einstein was more likely to reconceptualize the problem itself.

Present-day AI appears more like a machine-amplified version of the first set of abilities than the second.

rzkon 96.2% on ARC-AGI-3 with Opus 5

Code: https://github.com/jerber/arc-code

rzkon Text AI watermarks will always be trivial to remove

rzkon Position: LLMs Can't Jump

This might be of interest:

Building AGI Using Language Models – https://bmk.sh/2020/08/17/Building-AGI-Using-Language-Models...

rzkon [dead]

It has already vanished from the front page.

rzkon [dead]

https://archive.is/IZ7ew

rzkon The Buffett Indicator Keeps Flashing Red. Is It Broken?

https://archive.ph/Bpy8j

rzkon "Please just use plain English, I don't understand what you're saying."

An interesting solution suggested in the replies (https://x.com/andrew_n_carr/status/2081534245370314816):

    The fix for this is to say: only report to me in ASD-STE100 Simplified Technical English.

More info about it here: Simplified Technical English - https://en.wikipedia.org/wiki/Simplified_Technical_English

rzkon "Please just use plain English, I don't understand what you're saying."

Damn, I've been telling the same thing to Claude, good to know I'm not alone. I've also noticed that Opus 5 is much worse in this regard than Fable.

It's as if the model is speaking Mentalese, or in some "bag of concepts" language.

rzkon GrapheneOS protections against data extraction from locked devices

I think this has been posted in response to this news story [1] to clarify that GrapheneOS has strong protection against data being extracted even without a duress PIN/password.

On a related note, a recent article [2] also describes how GrapheneOS helped a journalist protect his work and his confidential sources citing the 18-hour auto-reboot feature that returns the device to Before First Unlock (BFU) mode, where keys cannot be extracted.

[1] A US man is being prosecuted after allegedly using a GrapheneOS duress PIN to wipe his Pixel during a border search – https://www.theguardian.com/us-news/2026/jul/23/cop-city-pro...

[2] A Journalist had his mobile phone seized. Did using GrapheneOS protect his data? – https://www.computerweekly.com/feature/Journalist-Richard-Me...

rzkon Opus 5 is currently #1 on Artificial Analysis Intelligence Leaderboard

I had a long session about SQL with Fable and at some point, it started to falsely trigger censorship for any message I write in that conversation, even for the simple string: "random message."

rzkon DKSplit – Fast Word Segmentation for Python