Very impressive! Interesting how all other benchmarks it seems to surpass Opus 4.7 except SWE-Bench Pro (Public). You would think that doing so well at Cyber, it would naturally possess more abilities there. Wonder what makes up the actual difference there
Reminds me of when hedge funds started laying increasingly shorter fiber-optic cable lines to achieve the lowest possible latency for high-frequency trading.
TPU8t is for training. But even still, once you’ve trained, you need to run the model too. And these kinds of models already have a huge latency hit so there’s not much hurting running it away from the trading switches.
A lot of people are saying it’s disconnected, but even if it was, if a string of your country’s top rocket experts started disappearing, you wouldn’t just sit idly by
What's sad is, 5-10 years ago, no adversary would think simply off-ing American scientists was effective strategy, America was a new scientist generation machine.
Now thanks to Research funding falling off a cliff and massive immigration restrictions, this is no longer true.
Amy Eskridge - who publicly stated she was not suicidal before "committing suicide" reported to her friends that she received burns to her arms and hands through her window in an attack that sounded similar to this microwave/havana syndrome stuff. She was very vocal about the fact that she was being harassed over her work before she died.
I mean, sort of? Let's put aside whatever she claims to have been working on. Then, consider, if there is a group of people more likely to be attacked by odd advanced weapons? Probably people whose work puts them into contact with with or near research into odd and exotic things. If someone was murdered in their NYC apartment and a schizophrenic neighbor claimed it was a wild tiger then sure, you'd take that with some salt. If you then found out the deceased worked at the zoo? Well...
> Amy Eskridge - who publicly stated she was not suicidal before "committing suicide"
I really hate the discourse around this stuff. Like, yes, disguising murder as suicide is a thing and obviously three-letters agencies do it.
But someone saying publicly they're not suicidal gives you close to zero information. People with suicidal ideation almost never advertise it publicly because, one, there is a heavy amount of social stigma attached to it, and two, publicly declaring you're suicidal is a good way to get involuntarily committed to a mental health institution.
I see a ton of jokes on social media that go "remember, X is not suicidal". How the fuck would you know? This discourse is so disrespectful to people struggling with suicidal thoughts.
Subjectively, it seems like it's even prudent to consider that someone who is involved in a discussion about whether or not they're suicidal is probably likelier than average to commit suicide. Fair chance that "I'm not suicidal" should really even be understood to mean, "I'm not suicidal right now".
They publicly said they were receiving threats. And that if something happened to them, don't believe it's suicide. That's a bit different than just, you know, saying it at random, or because someone asked you how you're doing.
It's respectful to trust someone who says they're not suicidal. Saying "they could've been suicidal anyway" is disrespectful to people who aren't suicidal and are telling the truth.
If someone is struggling with suicidal thoughts and is publicly lying about it, they shall not have my respect anyway: I'm ok with being disrespectful to them.
Source? It is known (and studied) that even at low power levels that do not significantly raise body temperature, short RF pulses can cause rapid, microscopic thermal expansion in the brain. This creates mechanical stress waves that can lead to TBIs
That's the cool thing! No one really knows, so any remotely scientific theory is indistinguishable from truth. Maybe even the more outlandish the better, so as to scare people more effectively.
"...the importance of the energy being pulsed in order to have biological effects on humans. When you produce pulses like this, you can actually stimulate electrically active tissue like brain tissue and the heart, for that matter, mimicking what the brain normally does, but now you're driving it with your pulses from the outside."
Ah, so a portable pulsed microwave device. Current phased array technology is certain to be able to produce a narrow and powerful beam. Current energy densities of batteries allow for a significant amount of portable power. Expecting to see "Show HN" on this soon...
It’s been going on since the Obama admin. Could be longer. Purportedly a unit was smuggled out of some former Soviet republic and we now have a copy of the actual device. When tested on animals, the device produced injuries in alignment with those experienced by US foreign service personnel.
It’s been a great source of fodder for conspiracy theorists though.
> CRS calculated that President Trump's budget proposal for FY2026 included approximately $181.4 billion for R&D, $10.7 billion (-6%) below the FY2025 estimated level of $192.2 billion. The requested $181.4 billion, which included advance and supplemental appropriations, was to support federal investments in the conduct of R&D as well as R&D-related physical assets (such as the construction of R&D facilities or equipment).
181 billion! How will we prosper as a nation if we go back to 2021 spending levels??
Lets say an American scientist in a strategic area was offered a boatload of money (or some other piece of mice) from China or similar. Legally probably he can move, though export control probably applies to the brain content too. How sure the said scientist would be that he isn't going to have a car accident? Gerald Bull would have a word on it. So, "disappear" may start to look like an attractive alternative. A related example - Russia has put a bunch of top hypersonic missile related scientists into prison for supposedly working with China (and may be they worked, though official charges have so far been obviously fabricated - like for publishing in a journal of an research article on a non-secret project with that article making all the typical rounds for months through peer-review, etc) as well as making a law giving FSB full control over any scientific interaction between domestic and foreign scientists and institutions.
I suppose the top AI talent may become subjects of a similar game.
It doesn't have to be China or Russia. As others have mentioned, the current political climate in the US is... "weird". At least, as an outsider, I just don't know how else to describe it. It's like watching/listening to gibberish.
So I can imagine American allies recruiting scientists en-mass, to protect themselves from America. The US has currently demonstrated a desire to take over allies completely (Canada, Greenland), and I'm sure few know who may be next. Some scientists may have simply wished to move abroad, and also, have quite valuable skills which are restricted in some way, hence them "disappearing".
not necessarily from America. The goal #1 of the US dominated NATO for example was to prevent Germany from getting nuclear weapons in exchange for protection by US. Now with US de-facto withdrawing, Germany would have to quickly get nukes (as well as missiles to carry them) - i don't see other option for Germany here giving the environment in Europe and MidEast. So they would also need such scientists. South Korea, Japan, Australia seem to be in the similar situation too. (and everybody understands that a nuclear weapons program can't be a long multi-year endeavor - somebody will try to stop you - and so it must be very fast once started, and thus you have to have ready-to-use skills and knowledge)
Keeping the FRG from getting nukes wasn't part of NATO strategy. The succinct reason for NATO was to keep the Soviets from marching to the Atlantic. The more pragmatic was expressed as "Keep the Russians out, the Americans in, and the Germans down."
No shit? Why would they have to? Is someone ready to nuke them if it turns out they’re no longer under the U.S. nuclear umbrella, or are they some special snowflakes who should have them while Iran (and most other countries) shouldn’t?
No. The nukes prevent the aggression even by a conventionally armed aggressor. Nukes ins't to win a war, it is to prevent one. Lets say Germany has successfully repelled Russian tank-and-soldiers invasion - it would still be a devastating thing for Germany which the nukes would help prevent from starting at all.
>are they some special snowflakes who should have them while Iran (and most other countries) shouldn’t?
Yes, i listed those several special snowflakes who were kept safe by the US nukes, and would need their own umbrella with US no longer providing the one. Iran's situation is obviously very different.
Yes, very different as in 'Our blessed homeland vs their barbarous wastes' meme.
We (and our allies) should have nukes because we want to prevent wars. But no one else should have them, since the situation is obviously very different (we wouldn’t want them to be able to prevent wars).
And I used to think that Little Rocketman was a crazy bastard, but it looks like I was wrong.
>Yes, very different as in 'Our blessed homeland vs their barbarous wastes' meme.
exactly. Iran's policy declaration of destroying whole countries (US and Israel in this case) and conducting of actual proxy-wars in order to achieve those goals make them barbarians from whom the civilization must be defended.
>we wouldn’t want them to be able to prevent wars
they don't even try. They want nukes to be able to conduct wars.
>And I used to think that Little Rocketman was a crazy bastard, but it looks like I was wrong.
absolutely. For all their tremendous faults, NK uses their nukes for deterrence as they want to genocide their own people in the comfort of personal safety. Whereis ayatollahs are hellbent on waging wars and destruction in order to spread their Islamic Revolution.
I’ve heard that lower and middle education aren’t exactly the US’s strong suits, but still? The US organized a coup in Iran over 70 years ago and has never really stopped meddling in Iran’s internal affairs. The US runs proxy wars around the world on a daily basis, and when we’re talking about barbarians, they’re certainly near the top - almost a GOAT.
Or the scientists and engineers themselves are wanting out of the US and were offered secret offers to "dissapear" and live elsewhere under a new identity
it does not pass the smell test, because what's the purpose of communicating about this FBI ongoing investigation ? at best it won't harm the investigation. it's probably propaganda
Backface culling has been common since the late 1990s when we started using face normals to determine lighting rather than per-vertex lighting. Pretty much every 3D game engine since about 2004 has included and enabled it by default. How is it that you made a game that doesn't use it?
For the curious readers, backface culling (at least in the way fixed-function OpenGL does it, and probably newer APIs still do) is not based on face normals, it's based on winding order of triangles, so it works even if normals are not used.
Also face normals (flat shading) are generally considered older tech than per-vertex lighting (Gouraud shading). Newer stuff since 2008-ish is generally per-pixel using normal maps for detail.
If I remember my graphics accelerator history correctly per-vertex lighting using Gouraud shading was the method SGI made standard with OpenGL in 1992, before 3DFX and ATI came in and made per-pixel via Blinn-Phong just as computationally efficient in about 2000 or 2001. Used before Gouraud and then alongside per-vertex and per-pixel was per-face flat shading, which kept being used up until the Xbox 360/PlayStation 3/DirectX 10 era. Face normals for per-pixel started being used in 1999 by SEGA, but stopped being common around the same time as normals flat shading got abandoned.
If I'm wrong please feel free to correct any of this, it's been about eight years since I last learned all of the different methods.
Ahhh. So you used a wrapper or a library? Interesting then. I had assumed that almost every rendering method enables frustrum, occlusion, and backface culling by default if only to clear the number of objects needed to be tracked in memory. One thing I noticed in your game is that it's based on the absolute mouse position, which with a 16:9 window makes it difficult to turn in certain situations because your horizontal movement space is much larger than the vertical movement space and that adversely affects turning speed. Changing so that is based just on horizontal mouse movement or adding keyboard controls might be better.
Whenever I read your articles, I get distracted by the space invaders and just play that instead. Maybe this is a problem with me being a bit ADHD, but I feel like I am not the only one
I just played for a couple of minutes and didn't even finish the article. Also suffering from ADHD but in my defense space invaders is pretty good with mouse controls.
Wow, if the benchmarks checkout with the vibes, this could almost be like a Deepseek moment with Chinese AI now being neck and neck with SOTA US lab made models
You need the infrastructure to train and run it regardless though. Kimi is great but I'm not getting the same performance from it running it on my MacBook or a 3090 as it running on a H100 or a Grace Hopper supercomputer. Pretend you did have said moat. Why wouldn't you also books infrastructure to run it on?
I've got a 12T model on my machine, built it myself. It's called Mytho. Too dangerous to even release a fact sheet about it. It can hack into the mainframe, enhance ultra-compressed images, grow your hair back, and make people fall in love with you.
According to the benchmarks, you are wrong. It is on track and slightly above some sota. Just the benchmarks speaking there, they can be/are gamed by all big model labs including domestic.
reply