HNHacker News
TopNewBestAskShowJobs

jstummbillig

6,608 karma · joined February 14, 2015

submissionscomments
jstummbillig··on Astra can read and write weights to improve Qwen
That seems both entirely obvious in hindsight and potentially super important.

I'll chalk up never even having thought of this being an option as yet another instance my human hubris, and a warning sign.

jstummbillig··on Claude Opus 5.5 Can Draw Permit-Level Architectural Drawings
I doubt that's going to happen.

1) Permitting is slow and, at some point, if you fuck it up big enough, the inspectors will probably take further steps, but otherwise not care much. You simply won't get a permit.

2) Who will care a lot are the builders and creditors. Building is expensive and if there's a couple of months of set back because of your fantasy drawings, that's not going to be in your favor.

Which is not so say AI planning is not going to happen. It will. But at that point it's probably mostly going to work.

jstummbillig··on Show HN: Subvocal – Talk to your computer without making noise
Category: A product, that's obviously super useful, as long as it will work.
jstummbillig··on Patrick Stewart taught us what good men do with power
More than anything, what struck me was the excellence of the team -- a striking combination of morals, skills, sense of responsibility and purpose -- that existed in TNG around Capt Picard.

Whatever was done to get that going was an act of absolute geniuses. (Granted, it helps when the hard parts can be explained through fantasy.)

jstummbillig··on GPT-5.6 Sol can beat Factorio on peaceful mode
Will be interesting to find out what they can bring over simply having better SOTA LLM coded opponents (I am sure, we will).
jstummbillig··on GPT-5.6 Sol can beat Factorio on peaceful mode
There has to be a AAAA studio building the new, greatest MMO, that also utilizes LLMs as part of the game loop, surely?

(Not sure how prohibitive it would be to do anything interesting with local LLMs of any kind in a big game, given how many platforms/setups you have to make the stuff run smoothly on, and the sheer amount of RAM they would bind)

jstummbillig··on GPT-5.6 Sol can beat Factorio on peaceful mode
How do you mean? (I have played 1 hour of Factorio)
jstummbillig··on 'Elite North Korean hackers' blamed for $6.3B Binance hack
Man, if only there existed some central societal body with the authority to essentially freeze accounts and unwind a transaction like this. What a security feature that would be.
jstummbillig··on AI isn't replacing SREs, it's deskilling them
It's fairly hard to understand to me, but also entirely unsurprising that swapping one occupation for another would be interpreted as a net loss at individual skill level.

You guys are still doing things, correct? Is my experience so outside the norm, or is stuff still challenging?

jstummbillig··on Ask HN: Are you tired too in this AI era?
Have you asked your system? Like, literally, "this new way of working is mentally overstraining me. What can we do at high level, that still is compatible with me in a competitive workspace?"
jstummbillig··on Claude Code regularly sends each background agent’s transcript to its model
But they don't. If that was how making money with LLMs worked, everyone would just generate models that generate tokens that do nothing. But of course that is not how it works and they don't.

It's just really bad to do something with a lot of waste that someone else can do with less waste.

jstummbillig··on Claude Code regularly sends each background agent’s transcript to its model
What market is going to care about the number of tokens somebody uses? That's like a market that cares about the number of if statements in a piece of code. It's nonsense.
jstummbillig··on Gemini 4 Carbon
Nobody is forcing Google to announce a model before they are cleared to release it.
jstummbillig··on Anthropic S-1
It seems obvious that the AI stocks are an interesting buy, but I wonder how other people feel about picking between Anthropic/OpenAI/Google here. Looks all fairly random, given the wild swings and interoperability of the technology so far.
jstummbillig··on SpeedCrunch 1.1
> The point at which new changes are more likely to degrade an application that's already excellent for its established use cases than they are to improve it further is the point at which we should consider the software "finished".

What I think: Being finished is not a property of the software. It's a function of skill of the maker to spot true opportunities to improve things, the skill of the maker to understand their own limitations to implement stuff (time, money, skill), and the current state of the software as it relates to the world.

It requires ongoing judgement and taste. If any of the inputs change, as they do, the answer changes.

When things get super trivial, you can probably consider them done at some point, but I hesitate. Let's for example just consider "cd" (the shell function). A typo like "cd /home/user/projcts" could offer a smart correction. If you went to "/home/user/work/clients/acme/backend/api" previously, through better usage of nav history you could maybe just get there by using "api".

Not claiming these off the cuff examples are great, and that they might not survive additional consideration, but it's fairly easy to see that they are arguable (and implementations like these already exist in some tools)

SpeedCrunch is much more complex and very far from inconceivably improvable, in my opinion.

jstummbillig··on We should be more ambitious with the software we build
Idk what that number does, other than stoke this kind of (very respectfully put) response. Seems fairly unnecessary/random.
jstummbillig··on SpeedCrunch 1.1
> If that's true, how do you explain the observable pattern of many consumer applications degrading in UI functionality and coherence after a certain point?

Well, that's easy: Bad choices. The better something is, the easier it is to change it in a way, that makes it worse.

I agree with GP, everything can be improved, but that doesn't mean that you will succeed, or that you should try, for various, separate reasons.

jstummbillig··on Ask HN: If the models are so good, why doesn't OpenAI just start 100 companies?
I don't think they ever claimed that companies are easy, just because models are good?

1000 super smart humans don't just automagically make a good company, or make creating one easy. Idk, seems to just be confusing different things.

But: I am sure more of that will be happening in the future. Anthropic and OpenAI are building out their product portfolio, going into research (medical) more broadly.

jstummbillig··on Claude Duet
Can anyone explain these to me? What does it do in practice that ChatGPT or Claude by itself does not. I look at it but it really doesn't click.

Paint me a real world picture.

jstummbillig··on Claude Duet
Product remains hard. I feel the AI companies are certainly not good but actually surprisingly good about it.
jstummbillig··on AfD Leak: What 421,179 Hacked Emails Show about the German Far Right
Pretending to care is worse than not pretending to care.

It's hard to work against pretense, because pretending is lying, and shortcutting reality is hard to compete with, if people start to believe in the lie.

For example: I can promise to gift everyone a unicorn for upvoting this comment (thanks in advance!). Instead of working on a concept of how to make that happen, I can instead just pretend to work on making that happen and mostly work on selling that to you.

It's not necessarily easy, but much easier.

jstummbillig··on SimCity, but with Permits
"free, because you are loved"

And because LLMs are making love cheaper.

jstummbillig··on DeepInfer: High-performance inference framework for DeepSeek models
Not sure what you are promoting here. The "open" model currently seems to work, if you a) burn outsized amounts of money that you have in the bank, in an attempt to get good eventually (Meta) or b) distill other models.

Both ask for a fairly narrow definition of what constitutes a good idea. There is plenty indication that both Meta and all of the Chinese labs will go closed, as soon as that path makes more economical sense for them. They are certainly already steering in that direction.

jstummbillig··on Facebook now charging to post more than 2 links a month
Those free and open models though
jstummbillig··on MiniMax M3.1 Flash
> The perverse incentives around benchmarks

I am not sure. The people who actually care enough to look at benchmarks will run evals, and find out quickly. If you post silly benchmark scores too often, you will simply be ignored. You are paying with your reputation.

jstummbillig··on Mercury 3: Diffusion at scale
They probably did not get it to work well?
jstummbillig··on There's never been a better time to go into math (by: 3blue1brown) [video]
You are certainly welcome to present all the things that humanity has lost that you consider meaningful, and I will be here to evaluate them.
jstummbillig··on There's never been a better time to go into math (by: 3blue1brown) [video]
I have learned that people find progress debatable, so I will just point to broad advances in life expectancy, child mortality, extreme poverty, literacy, educational attainment, real incomes, undernourishment, and access to clean water, sanitation, electricity, and healthcare and leave the rest as an exercise to the reader.

Equality is not a prerequisite for everyone to be better off (and please do note that this is not me saying I prefer inequality. That would be an entirely separate sentence.)

jstummbillig··on There's never been a better time to go into math (by: 3blue1brown) [video]
You can chose to do whatever you like, but that does not mean others now or in the future are obliged to pay you to do what you like if it's no longer economically viable or socially willed.

We can and should have it as something that people do in the future because we have the resources for everyone to do so.

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?

← PreviousPage 2 of 34Next →