Hi, this might not be obvious to you so I wanted to explain it:
Testing on a single device, or two devices, is not enough. I'm on a laptop with 96 GB ram and a Ryzen AI 9 HX PRO 370, and it's running at 3 FPS.
The crucial part is browser performance and network latency. You need to test across multiple browsers, but also evaluate frame times, CPU load, GPU load, etc. so you can make a reasonable estimate whether it's fast or not.
Then, when you realize that its not good enough, e.g. on firefox, you use firefox's profiler to see which parts are slow. LLMs would guess, probably, but you have to measure.
I cant just take a photo of a person or even car on the street and just share it online unless its clear the car or person is not the central point of the photo standing alone.
But you raise an interesting point: if this would be against the law in Germany, why is Facebook waiting for these glasses to be banned? It never should have made these available in Germany, and should be using technical means to ensure that they cannot function inside of Germany's borders.
So it's just more proof of the hypocrisy. Meta doesn't care about the privacy of the plebs.
It's perfectly fine to take pictures of the Brandenburger Tor for example and post them online. There will likely be people in that picture but as they are not the main aspect of the image, it will be fine.
It's something different if you zoom in on somebody in that scene and he becomes the main point of the image. Then his personal rights overwhelm the right to document public spaces.
In Sweden for instance it's a bit in-between (unless the law has changed). You're allowed to film in public if you're using a non-stationary camera, but stationary cameras require a permit.
That's an important caveat. We kind of have something similar, in that if you install CCTV, and if it captures an area where people would expect privacy (ie, looking through a window, or into someone's back garden ... which has its own caveats around being a public place or not) you have a right to have that area removed from video coverage.
That would be the right thing. You want to film something, you don't use people passing by as a free extra and don't publish the details where and with whom they were going online. Is is just some American thing that they do not care about feelings of other people.
If you have any respect, don't film other people or at least blur them.
Yeah, Germany is different here. I am not 100% sure, but in Germany it is allowed if you are not focusing on any particular individual person. It is difficult to argue in absolute terms, which is why most consider public photos forbidden completely (unless for private/personal, very narrow use). Nowadays, in Germany, it has helped me to consider everything forbidden, unless it is explicitly allowed somewhere.
That's pretty understandable, and I've no particular issue with it (although I'm now aware I broke quite a few photo- and video- laws when I was last in Berlin ;) )
If a power source is profitable on its own, it can be financed by private investors. If it is unprofitable, it needs to be financed at least in part by the state, so it would come at the cost of the taxpayer.
There is a meaningful difference between a generic "load all flag values" call that is identical for all users and some actual telemetry which sends application context to the back-end.
I find that subagents usually burn more tokens and take longer, and produce about the same quality. A real killer use-case is using a VERY cheap subagent to do a lot of work, or reviews. Don't be fooled into thinking that a "scout" subagent will gather enough info for a "coder" agent to just start working.
OMP also does a lot of rewinds - where the orchestrator comes up with findings on its own and rolls back to a previous turn with a summary update. I don't know if it always chooses correctly between rewinds and sub-agents, but it seems to do a pretty good job. Luna is SO MUCH cheaper that even if it spends 30K overhead in context its still much less costly, and when they work in parallel its faster too.
I urge everyone to compare this announcement with Anthropic's announcements. From the post above:
> On FrontierCode, which evaluates whether coding agents produce changes ready to merge into real codebases, GPT‑6 Sol improves substantially over GPT‑5.6 Sol, and is able to match Claude Fable 5.1 xhigh at much lower cost.
I continue to appreciate OpenAI's attempt at some honesty here, showing that they are capable enough and have skilled engineers to a point where they can recognize that slop is hated for good reason, and that there is a real issue. Compare this to anthropic, where e.g. in the Opus 5.5 announcement[1] one of the first points on the page is
> One tester completed a 680,000-line code migration in less than a day—work that would have taken an engineering team weeks. It’s good at finding and fixing inefficiencies in software: when we asked it to cut load times across every page of a web app, Opus 5.5 succeeded 39 of 40 times, while Opus 5 made smaller improvements that also altered the app’s behavior. A different tester had several Claude models build a game from a single prompt; Opus 5.5 scored higher than any other model on the strength of its graphics and polish.
This is the kind of shit that is the very reason why I stick to OpenAI and deepseek. OpenAI is simply more honest and reasonable about their models' capabilities, while delivering models that still have solid value.
Notice how the OpenAI announcement doesn't make use of anecdotes.
Testing on a single device, or two devices, is not enough. I'm on a laptop with 96 GB ram and a Ryzen AI 9 HX PRO 370, and it's running at 3 FPS.
The crucial part is browser performance and network latency. You need to test across multiple browsers, but also evaluate frame times, CPU load, GPU load, etc. so you can make a reasonable estimate whether it's fast or not.
Then, when you realize that its not good enough, e.g. on firefox, you use firefox's profiler to see which parts are slow. LLMs would guess, probably, but you have to measure.
Welcome to software engineering.
reply