Hacker News Reader

daemonologiston F-15E jet shot down over Iran

If you scroll to the bottom of that page, they discuss possible evidence of damage to the radar from satellite imagery.

iNaturalist is cool, but it'd be a lot cooler if they released their models.

daemonologiston Google releases Gemma 4 open models

The implication is that there is (should be) a major speed difference - naively you'd expect the MoE to be 10x faster and cheaper, which can be pretty relevant on real world tasks.

daemonologiston Google releases Gemma 4 open models

Gemma will give you the most, Gemini will give you the best. The former is much smaller and therefore cheaper to run, but less capable.

Although I'm not sure whether Gemma will be available even in aistudio - they took the last one down after people got it to say/do questionable stuff. It's very much intended for self-hosting.

daemonologiston Escaping the Ogallala Trap

It will probably lead to more cars traveling at any time, but potentially far fewer cars parked.

However: turn most of the street parking into bus and/or bike lanes, the parking garages into apartments, seems like an absolute win. (Except for Chicago, which is presumably going to have more problems with its privatized street parking.)

daemonologiston What Is Copilot Exactly?

I'm pretty sure they're referring to their coworker as "he," not an LLM.

daemonologiston VR Is Not Dead

VR's problem, in my opinion, is that I can get immersed (fully, exactly as the author describes it) in a 2D game just fine - the lack of stereo vision or head-tracking or motion controls is no more an impediment to my immersion than the limited binocular overlap or peripheral vision or lag in a VR headset. And 2D is a heck of a lot cheaper and more convenient (and less nauseating).

That's not to say VR can never be successful, but I think it needs to offer something more compelling than just "immersion." Exercise or AR might be viable routes.

daemonologiston Intel Announces Arc Pro B70 and Arc Pro B65 GPUs

If you stick with your OS/package manager-distributed version, installation isn't painful anymore (provided that version approximately overlaps with your generation of GPU). It's okay for inference, and okay for training if you don't stray too far beyond plain torch. If you want to run code from a paper or other more esoteric stuff you're still going to have a bad time.

I don't have an Intel dGPU, but I suspect the situation there is even worse. I mean you go to the torch homepage: https://pytorch.org/get-started/locally/ and Intel isn't even mentioned. (It's here though: https://docs.pytorch.org/docs/stable/notes/get_start_xpu.htm...)

daemonologiston Make macOS consistently bad unironically

Yes! After many years of using only linux or windows machines, I was assigned an iMac at an internship and noticed the friction with fullscreening things. I decided not to fight it and spent the next year happily working in little windows and making frequent use of the "mission control" gesture.

However, after the internship I went right back to fullscreen/window tiling in linux, so I can't say I really preferred it. Even now as a Gnome user with a big monitor and magic trackpad on my desk - which gives me ~equal access to either approach - I fullscreen everything.

daemonologiston Intel Announces Arc Pro B70 and Arc Pro B65 GPUs

7800 XT has 624 GB/s as well, and can be found for $400 used. 16 GB of course.

daemonologiston Ball Pit

Had to break out Chromium for this one - Firefox+Linux does not like webgpu (my whole DE started flickering).

daemonologiston ARC-AGI-3

Or the classic from Dijkstra (https://www.cs.utexas.edu/~EWD/transcriptions/EWD08xx/EWD867...):

> even Alan M. Turing allowed himself to be drawn into the discussion of the question whether computers can think. The question is just as relevant and just as meaningful as the question whether submarines can swim.

(I am of the opinion that the thinking question is in fact a bit more relevant than the swimming one, but I understand where these are coming from.)

daemonologiston ARC-AGI-3

That one was interesting - I found it a lot of work to plan in advance but trivial to complete because at every point there was only one sensible course of action. After a couple of rounds I didn't bother planning and just lined things up as I went.

daemonologiston FCC updates covered list to include foreign-made consumer routers

Hey, let's not undersell America's high-tech manufacturing capability. We could easily produce morse code keys and copper wire, for a price of course.

daemonologiston Ask HN: How much are you spending on AI coding at work?

$40/month at work and $10/month at home, and it's more than I can use.

I cannot imagine productively spending $250k/year on LLM coding - you'd need some kind of massive tree of agents reviewing each other's work and I think even then you would struggle to keep them on-task and sanity-checked. However, I don't make $500k a year so what do I know...

daemonologiston Department of State advises Americans worldwide to exercise increased caution

"It's," perhaps?

daemonologiston Flash-MoE: Running a 397B Parameter Model on a Laptop

Most definitely - the popular engines have extensive support for doing this and controlling exactly which weights end up where (llama.cpp: https://github.com/ggml-org/llama.cpp/blob/master/tools/cli/... , vllm: https://docs.vllm.ai/en/stable/configuration/engine_args/#of... , sglang (haven't tried this): https://docs.sglang.io/advanced_features/server_arguments.ht...).

Even with a MoE model, which has to move a relatively small portion of the weights around, you do end up quite bandwidth constrained though.

daemonologiston The Impact of AI on Game Dev Jobs. Open to Work Crisis

Of interest to me regarding Roblox is that it's been around for a long time - public release in 2006, almost 20 years ago! It was around when I was a kid, and I vaguely remember being aware of it and thinking it was dumb (certainly, I didn't know anyone who played it).

I wonder what happened to make it so popular in the present day. Was it incremental improvement, or a better fit with modern devices/culture; maybe simply being free to play?

daemonologiston A Japanese glossary of chopsticks faux pas (2022)

Did you also play Thrice today? (This was one of the daily questions.)

daemonologiston OpenRocket

Just reading the name I wouldn't have been surprised if it had nothing to do with rockets whatsoever - I was half expecting it to be some kind of "agentic platform to accelerate your product development" etc.

I think I need to go for a walk.

daemonologiston North Korean's 100k fake IT workers net $500M a year for Kim

I had the same thought - I guess there's additional overhead in paying the in-country proxy and probably also a lot of churn (being found out and fired, and then taking a long time to find another position).

daemonologiston Starlink Mini as a failover

Five or six! My kingdom for one fiber provider.

daemonologiston What is agentic engineering?

You are in violation of the HN guidelines. Please review the link at the bottom of the page ( https://news.ycombinator.com/newsguidelines.html ).

daemonologiston Tons of new LLM bot accounts here

I would propose July 2024 as the cutoff; early on it was unusual to just set an LLM loose to run amok on a forum. I'm sure state actors and some corporations were experimenting with it (e.g., Ultralytics on their own GitHub), but it was usually very obvious (or very subtle) and the volume of the noise has only picked up recently.

Date picked based on this Trends page: https://trends.google.com/explore?q=agentic&date=all&geo=Wor...

Of course I'm biased, having an account created after November 2022.

daemonologiston Claude March 2026 usage promotion

Would be cool to have a $5-10/month plan that only works off-peak, for people who want to do the occasional side project after work. Right now it's hard to justify anything but Copilot (because it's cheaper, offers the same models, and I'm nowhere near the usage limits).

daemonologiston John Carmack about open source and anti-AI activists

The point is not that it's not "real" open source, the point is that he has less interaction with the big part of the open source ecosystem which is feeling the brunt of the downsides of AI, namely, giant useless bug reports and PRs.

(I do agree that it's still OSS even if you never maintain it or anything.)

daemonologiston Document poisoning in RAG systems: How attackers corrupt AI's sources

Holy moly what's with all the AI comments in this thread?

daemonologiston Ask HN: What are you using to mitigate prompt injection?

    Don't post generated comments or AI-edited comments. HN is for conversation between humans.

daemonologiston No, it doesn't cost Anthropic $5k per Claude Code user

Even if it's larger, OpenRouter has DeepSeek v3.2 (685B/37B active) at $0.26/0.40 and Kimi K2.5 (1T/32B active) at $0.45/2.25 (mentioned in the post).

daemonologiston Neural Boids

I agree that this would be a more interesting approach; I think you might need more incentives to create a flock though (aerodynamic benefits, predator protection, etc.)