HNHacker News
TopNewBestAskShowJobs

jstummbillig

6,609 karma · joined February 14, 2015

submissionscomments
jstummbillig··on OpenAI discloses four more rogue incidents
> If OpenAI does it, it's gone rogue and it is nobody's fault.

Where did OpenAI, or anyone, claim that?

jstummbillig··on How the Claude Code team does code review
I am not sure what it is you are saying. What I am saying: I would have supposed that Claude writes 99.x% of Claude's Code code, and that ~no code at Anthropic is written by humans anymore.
jstummbillig··on How the Claude Code team does code review
That's a shockingly low number.
jstummbillig··on Show HN: Gemini 4 Argon plays StarCraft
How do they avoid the LLMs simply working with pre-training data? Which apparently they do, since the human written bots win. But it's a mystery, is what it is.
jstummbillig··on Waymo in Toronto
Not much snow in any of these, or occluded vision. Which drivers seem to struggle with, a lot.
jstummbillig··on Gemini 4 Argon
> That said, the em-dash is back w/ this model.

noooooooo

jstummbillig··on Gemini 4 Argon
Clicked on "Performance". Scrolled to "Safety".
jstummbillig··on The gap in shared understanding of LLM capability is widening
As far as I can tell, we seem to keep moving past states and redefining them while we do.
jstummbillig··on Muse Wonder
Timing is not a detail.
jstummbillig··on Codex CLI installs a persistent, self-updating background daemon by default
Being able to turn auto-updates off is okay, but that being the default would certainly not be okay in Codex, giving the reach, what it can do and how immature it is.
jstummbillig··on Why aren't AI companies being charged for hacking?
> "Costly" how, and for whom?

For multiple parties, probably mostly OpenAI (which is, of course, as it should be). I would assume the people involved in investigating, cleaning up and coordinating the outfall do not do it for free or cheaply, and neither is the infrastructure required to run this operation.

If it was not mostly OpenAI, or if OpenAI failed to cooperate, we would probably see a lot more lawsuits already.

> "Free marketing" seems like a more relevant characterization.

Yes, nothing screams free marketing quite like an AI company losing control of their AI. The public has been ecstatic in response.

jstummbillig··on Why aren't AI companies being charged for hacking?
The consequences, just so far and just from what I can see as a bystander, have been huge and costly. This has literally run the news cycle for weeks.

If you want to define consequences as exactly "not being charged", maybe that is correct in this case (so far). But a fairly wonky definition of "consequences" would be required to support this broad claim.

jstummbillig··on Why aren't AI companies being charged for hacking?
> Consequences so far: none.

What?

jstummbillig··on Passkeys are a trap
I have dedicated about 0 seconds to thinking about any of this: I store mine in Bitwarden, because it offered to do so. Is that a bad idea?
jstummbillig··on Terence Tao: Math 2.0 [pdf]
We struggle, and will for some time, to understand how transformed our lives will be.

The questions we asked were originally not about math. They were about a thing that we invented maths for to do or explain, a question that existed, because it touched us in some way that was already real to us. There is nothing that would not allow this to happen in the future. All this requires is attention and connection to the world around us. The maths required to answer our questions can be done and developed by something else.

To me, all you need to believe for this to be true is to agree that understanding maths is also not a stated requirement of reality to get a thing, if something else understands the maths (or something that does the same job). This is demonstrated by billions of people who do not understand maths and get things that, currently, require other people to understand the maths.

But the last part is entirely optional as it pertains to reality. That's just the best we can currently do (and in some important sense it is holding us back as a species, and in some other sense doing the opposite).

If that goal is understanding math, you certainly will be able to understand maths, more than ever before.

If that goal is something that required you to understand maths first in the past, you won't have to do that anymore.

jstummbillig··on IRCv3
Everything decentralized will keep being used. Stuff is sticky.

More illuminating: To what (relative) extend? And what's the trajectory?

Looking at Google Trends: https://trends.google.com/explore?q=%2Fm%2F03t2s%2C%2Fg%2F11...

jstummbillig··on Computers Cannot Make Decisions
I have no clue why people are so confused about any of this: The owner makes the decision, or pays a CEO/politician to make the decision. Owner is somebody vested with that right, for not particular reason. Could be ownership through capital. Could be voting rights in a country. Could be the son of a king.

It does not matter if the CEO is an AI or a human. In either case, of course they can make decisions -- and be held responsible. Not in any emotional sense (that we seem to want to bake into the concept for added confusion), just that if you are not satisfied with the result, you can replace the human or AI.

I am not saying that this is a fun vision of anything, but why are we making it more complicated than it is? The parts are all there and explained.

jstummbillig··on Anthropic AI model submits false tip on unsolved Philly murder, police say
> Sam Altman said yesterday "we believe that the world should accept some bad things happening for the benefits of this technology"

Technology requiring this is neither novel nor extraordinary (exhaust engines, factory farming, medicine). The only extraordinary part is somebody building it addressing it directly and inviting the outrage.

jstummbillig··on Yes, and
Thanks for your thoughts! Can you explain why in your model being able to read code will stay any more important for tomorrow's coders than it is for today's managers? Is it resting on the assumption that tomorrows models will be worse at that than tomorrows coders?
jstummbillig··on OpenAI annualised revenues $20B less than previously signalled
What I was saying: What FT presents in the way they do should not be enough to move the needle and convince anyone that this happened, regardless of what might have happened and that it could have happened.

People can just write stuff. That does not make it wrong but that also does not make it right. If your claim rests on some claim that some anonymous actor got some information, that's just not super convincing and neither is pointing at it as some sort of truth.

jstummbillig··on OpenAI annualised revenues $20B less than previously signalled
How does that contradict what parent said? Yes, obviously they are pointing at something for their numbers, but something as vague as "based on information that was provided to investors" might still just be entirely nonsense and is certainly not enough to establish confidence as to the validity of the claim.
jstummbillig··on Step 5 Preview, a 1M-context MoE from StepFun, shows up on OpenRouter
Why is that interesting?
jstummbillig··on “Math 2.0” will need to value mathematical progress more holistically
Alas, you will notice that article is 1.5 years old. Update on this issue (albeit soon from a year ago too): https://waymo.com/blog/2025/10/creating-an-all-weather-drive...

More telling: Waymo just rolled out in Denver (1 month ago or so), apparently fairly confident they got this handled given the upcoming winter.

Progress on the obvious stuff keeps happening (which kind of brings me my to the first comment here).

But also: It does not have to do "snow" to be useful! A lot of cars/people don't drive when it snows heavily, and that's something we have always been okay with (at a societal level, YMMV of course). If was only useful 95% of the year that's still great. A lot of technology works like that.

jstummbillig··on Claude Haiku 5.5
Probably explained by implementation around a particular model, no? I suspect if you put similar effort in, you could get Haiku 5.5 to do what you want.
jstummbillig··on “Math 2.0” will need to value mathematical progress more holistically
> And here we are, a decade later.

With Waymo and Tesla increasingly doing what they said they would do, and a small number of early adopters happily paying money for their services, that do work.

So what's the critique? That the timelines are not correct? Sure. And how about the timeline of the people who said "research level math, never in my lifetime" and the people inside the ai companies who are apparently increasingly spooked by how quick the progress is? How about the various levels of code/programming jobs that AI was supposedly never going to be able to do, but, in reality, now just does?

We are engaging in some very one-sided discrediting, and I am not sure, why.

jstummbillig··on “Math 2.0” will need to value mathematical progress more holistically
It's kind of astonishing that after all we have seen in the last years people still find the position that AI will not be able to do an obviously valuable thing likely and it requiring an explanation (instead of the other way around).
jstummbillig··on Claude Haiku 5.5
Because it's increasingly useful and the thing you are substituting (human time) is much more expensive.
jstummbillig··on Claude Haiku 5.5
How about if we get away from written text as the input, to something more fundamental, that then also is able to produce text (among other things)?

Given that humans learn to talk while having encountered a measly number of word instances, and, given enough time, we should always be able to improve on the lottery that is biology, it does seems fairly likely.

jstummbillig··on Claude Haiku 5.5
Maybe they think it's of interest.
jstummbillig··on Claude Haiku 5.5
Opus 5.5 does :^)
← PreviousPage 3 of 34Next →