mungoman2

Imo the difference is enormous between social media and forums.

Infinite feeds are designed to game you for attention, whereas the forums of yore were there to facilitate discussions.

I'm sure some forums would also have liked to game you if they could, but they didn't have the scale to always have something juicy to serve up.

To me it's super uncomfortable to expose my kids to a product designed by large teams with the goal of making it addictive.

> Naturally, trains already have track-based location systems, but they are usually based on a train being within a “moving block”, so their accuracy is down to metres rather than centimetres. If you want to monitor track conditions, the more accurate the location of the suspected fault, the less time staff spend repairing it.

Agree in principle though, is this extreme precision really needed?

Yes, I believe the paradigm shift will be to not treat the code as particularly valuable, just like binaries today. Instead the value is in the input that can generate the code.

The money argument is IMHO not super strong, here as that Mac depreciates more per month than the subscription they want to avoid.

There may be other reasons to go local, but I would say that the proposed way is not cost effective.

There's also a fairly large risk that this HW may be sufficient now, but will be too small in not too long. So there is a large financial risk built into this approach.

The article proposes using smaller/less capable models locally. But this argument also applies to online tools! If we use less capable tools even the $20/mo subscriptions won't hit their limit.

The money argument doesn't make sense here as that Mac depreciates more per month than the subscription they want to avoid.

There may be other reasons to go local, but the proposed way is not cost effective.

Hm the first statistic sounds pretty low

136 years That's how long Europeans collectively spent waiting for delayed trains in 2025.

Is that just summing up each train's delay without taking number of passengers into account?

I hear the team running the site was laid off.

No, the tool rtl_433 repackages payload data in json for easier downstream consumption.

I agree! Seems like very interesting work.

As proposed in the article, calculating the area weights all points in the viewshed equally. I wonder if it makes sense to give more weight to far-away points? Or maybe less weight? Not sure! But would be interesting to explore and see if it gives anything.

Wow, with this in place the incentives are enormous for OpenAI to allow sponsors to pay for a slight nudge in the recommendation this way or that.

This will replace the current ad economy.

I don't really get comments about unreasonable labor in this comment and TFA. It's what, 16x16 = 256 SIM cards per such machine. At 10 seconds per SIM card, that's like an hour to fill one machine. In a week at 8 hours per day you can fill 40 such machines with SIM cards.

Interested as I have a basement full of stuff in questionable order.

Could you explain a bit more? If a box in the basement box is marked with 4 emoticons, how does this help you understand content, context, history of it?

Wouldn't a slower tick make it easier as you get more wall time to do the same challenge.

For story writing you generate into the framework of https://news.ycombinator.com/item?id=45134144

This is extremely cool! Is this the first type of tools that are genuinely enabled by AI?

Can we distill the story into something to feed z3 to prove there are no plot holes?

Yes, agreed. What we see is simply a clever way to differentiate the customers that can pay a premium from those that can't. The end goal is to extract the maximum amount of money.

This is very cool, if the reviews are useful it can really shorten the submit/review/rework cycle.

This is super neat, cool that you have an online demo as well.

I wonder if there are any ideas on how to make this a OpenSCAD-style editor instead of interactive? I like the text-based style for simple regular shapes, but they tend to end up too simple and regular. Maybe tools like filleting edges SDF-style is a game changer?

Well if you accept collectively shift all times by an hour, you're effectively arguing for the system we have.

Currently, 146 UN member states (75%) recognize the State of Palestine.

The blend operation happens on linearized values, so the frame buffer being gamma corrected is not a problem. 8-bit sRGB is a very good gamma corrected format.

Do I have a treat for you https://youtu.be/G8YBV7GoF_s?feature=shared

Yes it's very possible to do this. Without seeking support it's trivial, just instruct the encoder to encode with a low bitrate for a few seconds and then increase it.

To support seeking you could encode a low bitrate stream, and a high quality stream, and then a number of ramps between these. So when you seek you start with the low bitrate stream and then after a few time units go on the ramp to the high quality stream.

Very cool! As I understand the flow is transcription with whisper, followed by note generation with GPT4. I guess the prompts in the second step is the secret sauce?

That triangle doesn't mark the stop line though, it's further back. If cars stopped at the triangle they would block pedestrians crossing the street.

This is an annoying change, but the cars in the video weren't actually running any red lights. Doesn't help the case to exaggerate.

This is not a bad idea. It's not dissimilar from an autocomplete engine. If reset per word and using a static probability table it will work its way into muscle memory. In practice fingers will only need to leave the home row for some initial letters.

Super interesting how the example of "Rule 105 Reversible" inserts a mirror 3/4 down, but only on the right half.

I wonder how it continues.

Love it! Beautiful to avoid the redundancy of a special encoding when the same effect can be reached through other means. Simplifies the decoder.