There's https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Ff-droid.org.%2F It has NewPipe, which is a YouTube viewer/downloader. I just side load that without f-droid since I already have the Android developer tools and am comfortable with that.

Any state of the art video call uses ICE & STUN (or just uses WebRTC which includes these) to punch through most NATs.

The original reason is that it helps during crash investigations to establish speeds, seat belt use, how the safety systems operated, etc. Good for both the engineers and the lawyers.

At this point, a breakup of Amazon seems like a great idea. Split them into a logistics operation that handles the warehouses and last mile delivery and sells that handling service to anybody at an equitable price. Split AWS off into its own company. Amazon Prime gets split into it's Netflix competitor and a logistics membership. Amazon Basics and all the other self-dealing Amazon does on its marketplace gets its own company, without access to the data from the marketplace. That leaves the marketplace itself, which should have capped listing fees or accept liability for the products, but no more fat rake with no guarantees.

I think they are trying to go beyond just detecting humans, and are now looking to detect click farms. But sophisticated privacy-conscious users click fast and aren't signed in or have workflows that require lots of captchas, and it looks the same.

Lines of C code or kB of compiled C code doesn't have any relation to the size/complexity/development cost of an ASIC.

Basically, you should constrain the problem by figuring out what the cost of the cheapest general purpose CPU and associated IO hardware would cost, and then figure what it would cost as a SoC by summing.

The ASIC, to be marketable, needs to either exceed that solution in performance or cost.

You should also look at the cost of ASICs that do similar work, possibly combining several if that's required. That gives you another estimate.

Writing your algorithm in verilog and programming an FPGA is a 3rd method.

The problem with the bell is that a substantial portion of pedestrians, upon hearing the bell, act as if they've never heard such a thing before and are unsure if they should keep left or keep right.

What was the method used to steal these employee's stock? Is there a news article about it?

> What is a realistic time frame?

It's a societal change that needs to occur. Current SDC can handle good weather, well-maintained roads, etc, but struggles with ambiguous situations (unprotected left, pedestrian walking on the shoulder). If we, as a society, decided that SDCs are more like trains, and you had best stay out of the way, and anyone getting injured by one is to blame, then we could have SDCs very soon.

There is precedent for this in jaywalking. Look up the history of that term and how the auto manufacturers basically created a crime out of thin air to shift liability away from their creations.

Could this just be clever "parallel construction" to cover insider trading? Buy some satellite data or phone location data and use it to prove what you already know.

Airlines could pressure the FAA to change how PIC hours are calculated to make it a bit easier, and basically promote some of the current stock of low hours pilots immediately.

Lots of people are using mouse and keyboard through 3rd party adapters on Xbox to gain competitive advantage (aiming in an FPS is much more precise with a mouse).

Step #1 here is to let people use it legitimately against other M+KB users, and build a data set of what M+KB input looks like versus controller input

Step #2 is maybe using the ML model from step #1 to ban people using M+KB adapters to cheat.