HNHacker News
TopNewBestAskShowJobs

joshka

2,489 karma · joined June 6, 2011

[ my public key: https://keybase.io/joshka; my proof: https://keybase.io/joshka/sigs/ElI6j13SI-_2rTrkLnoKyV-uDepH4JvAQAzp2OWV_W0 ]
submissionscomments
joshka··on Anatomy of a Frontier Lab Agent Intrusion: A Timeline of the July 2026 Incident
The exploit gym setup explicitly allowed access to package registries and v8 sources. Putting a cache on that doesn't seem like a bad idea generally, except when there's a 0-day in the cache :D But yeah, for this sort of thing I'd be locking down very specific egress things and putting alerts on it that are entirely outside of the red network.

> I would argue that it's negligence and that's aside from the fact that if a human did this there would actually be repercussions.

I’m not sure “negligence” follows just from the controls turning out to be insufficient. Research involves mistakes, especially around novel failure modes. The question is whether the precautions were unreasonable given what they knew at the time, rather than whether hindsight suggests stronger controls would have helped. Doing it twice though would be negligent.

Caveat: I’ve worked with some of the people involved, so I’m probably biased toward a charitable reading.

joshka··on Now is the time to give LLMs access to the ACM digital library
Thanks for the reply here.

I guess I have some perspectives on a bunch of this. I'm for open sharing of academic work for all (but I'm not an academic, so my perspective is a consumer), so inferring your perspective here I think we agree on that. I maintain many open source (MIT/Apache2 licensed) libraries, and I've also worked in big tech (Amazon, OpenAI). I believe both in the idea of collective commons but also in the ideas that there should be the ability of people to sell software. There's tension in that social contract similarly, and it gets more complex when you look at copyleft.

I guess I'd be disappointed if this was just allowing big labs access and not more broadly allowing access to the ACM library for smaller open source models. Very much in agreement with your last points there.

joshka··on Toolcraft
This is freaking good.
joshka··on Now is the time to give LLMs access to the ACM digital library
How do you square away the idea that you do science for the increase in knowledge of human kind, but then say that a particular use of that knowledge is verboten?

I get the copyright aspect of this and I'm not arguing that here. I'm more asking about the moral / ethical idea of choosing who can benefit from your science.

Obviously there are the moral / ethical arguments about AI in general here to weigh against - those have been hashed out significantly elsewhere, and I'm not interested in debating them. What I'm asking about here is the impact on science by sharing it with tooling that distributes it in ways not generally considered when originally written.

A quick check of your post history suggests the frame that you work in strongly is privacy related research (observation - may be wrong). I'm curious how that impacts what you wrote here generally.

(Just to be perfectly clear, I'm not arguing your points here, trying to understand them better)

joshka··on Show HN: I left VSCode to build an IDE to handle many projects/agents workflow
Consider optimizing your demo for laptop user screen sizes and checking contrast of all elements. A lot of things on the demo are difficult to read fast due to low contrast, which makes animations using them less useful to show off what is actually happening.
joshka··on Anthropic publishes a practical key-recovery attack on HAWK-256
discussion / article: https://news.ycombinator.com/item?id=49087091
joshka··on Fast Remediation Is the New Trust Model (JFrog and OpenAI Zero-Day Findings)
https://huggingface.co/blog/agent-intrusion-technical-timeli... has that side of the write up a bit more

discussion: https://news.ycombinator.com/item?id=49089500

joshka··on All major LLMs are lib-left. Even Grok, half the time
Makes you wonder whether that reality has a left wing bias quote might have some legs after all (though it's probable more likely selection bias for the corpora)
joshka··on OpenAI’s accidental attack against Hugging Face is science fiction that happened
The other suspicion I had was jfrog artifactory (you mentioned OpenAI use it in a blog post in Jan). I accidentally conflated that with Nexus 14 july CVE. I think that's probably more likely.

I suspect it may be possible to throw a local model (same as hf did :D) at jfrog and find the exact mechanism in a small amount of hours.

joshka··on OpenAI’s accidental attack against Hugging Face is science fiction that happened
I wonder if it was this bug fixed july 14 in sonatype:

https://support.sonatype.com/hc/en-us/articles/5316501964136...

pure speculation here - no non public info

joshka··on OpenAI’s accidental attack against Hugging Face is science fiction that happened
There's no reason to think that a tool that can find an 0-day in a repo cache can't work out how to make that host send a post request rather than a get request once it has its keys and is able to get it to make arbitrary web calls.
joshka··on Codeberg: ToU extension to prohibit LLM-extrusions
https://news.ycombinator.com/item?id=49003386
joshka··on Late.sh – a command-line Clubhouse for computer people
yes, and... ?
joshka··on Show HN: Leaves – A text-UI disk usage treemap visualizer
Ratatui maintainer here. Looks great!

I've had on my someday bucket list doing up a web based version of daisy disk as a docker image I can slap on my various servers, but this is probably just as good an approach. I like it.

joshka··on Fixed three bugs that made Qwen3.5-122B a daily driver on Mac Studio
I caught the title "The real work..." labelizing things in a sort of weird phrasing like this is the one I've seen a lot.
joshka··on AI Can't Recreate the Thrust Game (But It Can Help You Understand It)
This is often people invoking chesterton's fence, where the quickest way to have something corrected on the internet is just to say something wrong.
joshka··on OpenAI Forked Git on GitHub
This is related to them hiring Taylor Blau to work on SCM stuff recently:

https://x.com/gdb/status/2074652518135804321

https://x.com/ttaylorr_b/status/2074523670278742457

joshka··on Write code like a human will maintain it
I wonder how effective it would be to just put the blog's title as a line in your AGENTS.md...

This is kinda the "this could have been an email" of AI complaints.

As an AI acolyte, I strongly believe that AIs should be being trained for this sort of mode by default, but they tend to be optimized (or at least tested) to solve the single shot benchmarks rather than multi-step stuff that you really need in a long term maintainable project.

There is a https://www.scbench.ai/ SlopCodeBench that checks this and a variety of others. These don't seem to often come up when frontier labs are advertising their new models.

joshka··on Build your own vulnerability harness
or just let pi create that approach - i.e. slap the above input into a clanker's input field and see what comes out.
joshka··on Build your own vulnerability harness
If I was doing this at scale, I'd be looking at two features in particular. Cached input tokens - leveraging the LLM's token cache to get a 10:1 cost advantage, and batch tokens, 2:1 cost saving.

The benefits of doing your own harness here is that you get to explicitly program around those specific things to optimize cost. And they both heavily benefit you in the way that these sorts of jobs work - at least for the hunt side of things.

TLDR - context management is pretty important for cost management when you're doing something repetitive in bulk.

joshka··on Postgres rewritten in Rust, now passing 100% of the Postgres regression tests
If you examine what you're saying here and slippery slope it a little, examining past effort for misses and correcting them at scale is not worth doing? There are many ways that the second part of your perspective are incorrect (burden / lack of point). I bet that a coding agent could in an automated fasion find at least reasonable additions that you'd say add value and reduce potential for error long term that you'd find valuable. They've probably already done so (I don't know postgres dev at all, so just supposition here. They will 100% do so in the future.
joshka··on My thoughts on the Bun Rust rewrite
I have a working port of ghostty to rust ... not even kidding
joshka··on Postgres rewritten in Rust, now passing 100% of the Postgres regression tests
It would be reasonably easy to audit and automate this...
joshka··on Postgres rewritten in Rust, now passing 100% of the Postgres regression tests
A lot of the signal (github, forums, mailing lists, discord, etc.) can be turned into signal. Right now it's easy enough to collect. In future it will be easy enough to cluster and generate preferences, experience, etc.

Every bug report, code change as a result, PR / commit message, PR comment that steers preferences, etc. is solid signal to generate future tests.

joshka··on FrankenMarkdown
The license is the most interesting thing here...

https://github.com/Dicklesworthstone/franken_markdown/blob/m...

joshka··on Regression to the Mean: on LLMs and the quiet death of the new
Technically incorrect (the best type of incorrect).

A conditional probability model (LLM) is not necessarily a mean-seeker.

Possibly mode-seeking, rather than mean.

joshka··on Better Models: Worse Tools
I've been trying to push for this perspective about the error messages of jj vcs. There's some push back from people that don't perceive that making tools work well with LLMs is also making tools work well with humans. (Obviously there's more nuance to the arguments than this one sided perspective).
joshka··on The cost YAGNI was never about
But here's the rub - if you want your clanker to do those steps, it's usually a simple matter of adding them to your AGENTS.md and then it always does them.

I'm a big fan of the characterization step step being added. And it can be reasonable to add this before or even after the fact as a commit prior to your actual commit (assuming you're familiar with using tools where that's easy to do - e.g. jj or git with rebase). And the agents can do this - they just don't tend to without you saying to do so.

A lot of engineering practice comes from choosing which elements are reasonable to use given the context of what you're building. Providing that is your job. When you do that poorly, you get poor results. But garbage in garbage out has always been a thing. Any advanced automation amplifies ambient assumptions

joshka··on The cost YAGNI was never about
Citation needed.

But in seriousness, this intuitively feels like something (as phrased) that would be easily influenced by loud noise and quantity rather than hard facts. The "piloted poorly" part is applicable to any tool use. AI is no different there other than its adoption rate.

joshka··on I built a 100% local network privacy appliance to stop smart home spying
I read every bit of the site and still have no idea what this is / does.
← PreviousPage 4 of 34Next →