Hacker Newsnew | past | comments | ask | show | jobs | submit | rx_tx's commentslogin

On the topic of tricking the automated phone usage detection cameras this youtuber had an entertaining video where he built a car phone holder by molding his hand and making a replica.

https://youtu.be/Ud8kFCmalgg


> Based on 50 Cent's name being coined

I see what they did there.


I suspect hiding the manufacturer/model was very much on purpose, they blurred the markings on the PCB and hid the domain name for the manufacturer's API calls (and in the console logs as well).


I agree, hopefully it helps not getting the article taken down because its a very good primer on getting any ESP based device locally working.


H.I. Sutton did a great video [1] about it that also explains how it was beneficial due to the way enemy submarines had to estimate speed and heading and could get fooled.

[1]: https://www.youtube.com/watch?v=Kw7vq_YD6JM


When you're trying to hit something moving 20kt, with something moving 30-35kt, from a few thousand yards, it doesn't take much error in estimating speed, heading, or distance, to make them miss. It's honestly more remarkable that they hit at all in those conditions, even with a "spread" (shooting several along slightly different headings hoping one or two will hit).


Anyone with a copy of Silent Hunter 3 https://store.steampowered.com/app/15210/Silent_Hunter_III/ can experience this for themselves. Trying to estimate another ship's heading from a vantage point near sea level is maddening, even with the aid of a little recognition book which shows ships of that class at various relative headings, and as TFA says it's especially hard to tell the difference between perpendicular and near-perpendicular relative headings.

Mind you, this also explains why dazzle was a promising idea in the first place. Calculating Angle On Bow was already the hard part of ship-against-ship targeting, why not try to make it even harder?


SH3 vets plotted relative positions to get heading much more reliably, which was something that was also insanely difficult for sub captains.

With some trig you can get relative headings by taking bearing measurements, waiting 30s, and then taking more, but speed must be known.

To get full heading and speed you can do the same thing multiple times but now we're in a least squares problem and it has plenty of singularities.

The most reliable method for me was to pass in front of the ship submerged the fire tail tubes or turn around if you have enough time.

Insanely difficult for fast moving targets.

I played so much SH games from the first days on, that I went and did my PhD on tracking using bearing measurements (albeit mostly applied and only a tiny theoretical contribution).


Well, at least in WW2 they had mechanical computers that could use the input width of the vessel and the class to estimate the range, heading, and firing angle to set the torpedoes up. There's a good series of youtube videos by a sim player that teaches how to use the TDC: https://www.youtube.com/watch?v=ANk6hZCcVRw - it's very in depth, and demonstrates how much effort it takes to get a firing solution, as you say.


The worst thing is that the TDC's calculation is very much GIGO if you get the Angle On Bow input wrong, and that's the hardest thing to get right.


Even with a spread, you're really just narrowing the odds, not guaranteeing a hit


Adding to the difficulty is the possibility of observers seeing the incoming torpedoes and a fast-moving, maneuverable ship dodging them. Fast and maneuverable as in .. the USS New Jersey, able to sidestep a spread of five torpedoes from a Japanese destroyer.

https://en.wikipedia.org/wiki/Japanese_destroyer_Maikaze#Con...


This article has good background, context, and explanations [1] They skipped CUDA and instead used PTX which is a lower level instruction set where they were able to implement more performant cross-chip comms to make up for the less-performant H800 chips.

[1]: https://stratechery.com/2025/deepseek-faq/


> Moreover, if you actually did the math on the previous question, you would realize that DeepSeek actually had an excess of computing; that’s because DeepSeek actually programmed 20 of the 132 processing units on each H800 specifically to manage cross-chip communications. This is actually impossible to do in CUDA.

You can do this just fine in CUDA, no PTX required. Of course all the major shops are using inline PTX at the very least to access the Tensor cores effectively.


So can people do the same in SPIR for OpenCL or amdgcn?

https://en.wikipedia.org/wiki/Standard_Portable_Intermediate...

https://www.khronos.org/spir/

Or even better in the unified language like SYCL?

https://cdrdv2-public.intel.com/786536/Heidelberg_IWOCL__SYC...


Gastropod is a great listen, and they did an absinthe episode, interesting history and details.

https://gastropod.com/absinthe-the-worlds-most-dangerous-dri...


the "World's longest song" is from John Cage, called As Slow As Possible and is played continuously on an organ in a church in Germany.

It will be completed in the year 2640.

https://en.wikipedia.org/wiki/As_Slow_as_Possible


It's been a big pain for Tesla as well, where their tiny 8GB emmc on the center screen would fail since they logged to it too much... 134,000 vehicles recalled eventually after they denied it was an issue.

https://www.tesla.com/support/8gb-emmc-recall-frequently-ask...


Jesus Christ are they amateurs? These are steel boxes on wheels and we're dealing with the same issues as shitty 200 dollars android tablets from 10 years ago.


That's because all who gets hired at these hyper-fast startups are fresh graduates who can do leetcode by heart.

The people who have been in the field for a decade or more can't be arsed putting up with all that and so you get stupid issues which were solved years ago but the devs were not aware of them.


So where do those people go to work instead? I want to work with them.


Defense companies.


That affects the infotainment computer only so driving is unaffected. You just wont be able to check your speed etc. But yeah, it's way too common of a mistake.


I don't know how their cars' UI is designed (and I hope I never have to) but if it's the only way to determine car speed or battery capacity, then this goes well beyond infotainment. No wonder they were forced to do a recall.


the specs on the one in the article mention it's a multitouch RESISTIVE trackpad, so I don't expect it to be great, at all.


Try using a capacitive touchpad in the pouring rain.


> Integrated tetra-array mic

That's definitely an uncommon way of saying quad microphone array.

It's an AI toughbook due to having a unspecified "1.4Ghz AI Boost NPU" which seems to be something Intel offers, and according to other websites, operates at 34 TOPS.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: