londons_explore

Not only are they shared kernel... They're shared process, shared address space, shared memory pool and allocator... In fact, there is very little isolated about them at all.

I bet there are a million ways to cause side channels allowing learning about other code or data on the same machine, and just one V8 bug (of which there have historically been thousands) let's you take over or modify code in another isolate.

For nearly all tasks, I want the fastest and smartest AI model.

It is vanishingly rare I ask an older model to do any task. Newer bigger and smarter models will just do the task better.

Therefore, I believe we are nowhere near 'good enough'.

I never drive my steam engine to work these days. It isn't good enough.

common misconception. In fact, it would probably be cheaper for most people since they wouldn't have to own a generator (or pay for use of someone elses).

Instead, people (especially the poorest) would need to put effort into tracking their own usage - and understanding that an airconditioner uses far more electricity than a phone charger.

I just don't understand why government didn't step in and raise the price of electricity 5x.

It is better to have electricity 24x7 but only 50 watts to run a fridge, phone charger and one lightbulb, than to have electricity 3 hours a day but be running 1000 watts of airconditioners.

I guess perhaps energy theft was too hard to police?

Would be pretty funny for an ISP to MITM all streams and put a "internet provided by Verizon" watermark in the corner of all video...

I made ~10 games with opus 5.5 (all multiplayer web games over web sockets).

About half the time it made a playable game in a single short prompt. The other half of the time a few follow-up prompts were needed for refinement (eg. Things like "the blaster weapon is way too powerful, divide it's hit points by 10" or "we need a way to reconnect a player whose network dropped mid round" or "the GPS doesn't work on iOS")

Yeah, but the server is most likely running the whole thing in a docker container with no permissions to do anything. Big companies security teams tend to require that when opening non-trivial third party code up to the internet.

The client on the other hand I would guess is running it's code as root, or at least something with full GPU access.

Kinda sad that it's 2026 and this data still goes over the internet unencrypted...

RTMP and all the video and audio protocols behind it aren't trivial either - I bet there are hundreds of exploits waiting to be found that any three letter agency sitting on the internet can use to take over your PS5 and all credentials stored within too...

Crash damaged cars have loads of good components though - typically allowing nearly all non-crash repairs to be done by just swapping parts about, unless the manufacturer has got software locks on swapping parts.

And the range extender was underpowered. Drive up a hill on a freeway at 70 mph and the battery would deplete and the car would slow down to 25 mph which was the fastest the range extender could make it go up a gentle incline.

It's dangerous to do 25 mph on a freeway... There are some roads I'm too scared to drive my i3 on now.

If you pulled over to let it recharge, it wouldn't do so, since the range extender refuses to run at full throttle when the car isn't moving fast (perhaps no cooling fan??)

It's fairly easy to run a program without access to the internet. OpenAI are employing a smart bunch of people. Why is this seemingly easy task outsmarting them??

Phones are charging ~8 hours a day whilst you sleep though. And few people mind the phone seeding stuff on your home WiFi whilst charging. Many users could also spare 32gbytes for seed data too, especially if the app was smart and automatically reduced the store size before the phone started to notify you of low storage space.

If a typical user uses preertube 30 mins a day, that still gives a 16 to 1 ratio between seeding and leeching, and a 1024 to 1 data storage to data viewed ratio, meaning there is a good opportunity for a big back catalogue.

I want these things to implement peer to peer caching, so that when 10 people are watching the same video, only one downloads from youtubes own servers.

Then it's just a matter of letting people add their own videos directly to the service and it can become fully independent of YouTube.

This thread reminds me that there is a huge benefit to using a super common OS+hardware.

In 30 years time it will all be outdated... But emulators will exist and a business process in 2050 that still depended on something I built today could still run with minimal effort or risk.

Doesn't look that unhinged to me.

An in-app notification saying 'your access ends on X date and here is who you should contact if that isn't what you want' is exactly what I'd expect.

And the hutter prize disallows GPU's. If you allow use of a powerful GPU, you can do quite a bit better.

I suspect GP's question is answered by the fact the system is not in thermal equilibrium. The view the antenna has is approx half of the sky whose temperature is far lower than room temp. Even indoors, the roof of your house is sufficiently transparent in RF that you are still seeing the sky behind.

If the same experiment was done in metal box with everything of constant temperature, I don't think it would be possible to transmit any bits.

But training LLM's is also a task one can do whenever you have a spare GPU-minutes.

I wonder why they don't have some kind of scheduler which makes sure there are never any idle minutes. One would imagine they at least would have autoscaling on their production serving workload and use the freed compute capacity for model training for example.

TL;DR: Mathematician community broadly unhappy with AI because it might destroy the community by doing all the mathematics for them.

Google internally has lots of possible redundancy levels for data.

They don't sell any of the lower and less reliable levels to the public, I suspect simply because the reputational damage from losing user data is so bad, and the news will take no notice of the fact the user got a discount for less reliable storage.

A camera has literally tens of millions of thermal noise sensors...

At a minimum I'd like cloudflare to mark such traffic in some way, and browsers to put a little cross through the "Https" indicator to show the traffic is passing unencrypted across the internet.

Great. Next can cloudflare stop providing Https to the user (giving the impression of security) when the connection back to the origin isn't strict HTTPS and is often plaintext?

This seems like the kind of thing that the USA will explicitly grant an exception to.

It is clearly bad if the whole of Iran gets their own CA infrastructure which the NSA can't as easily spy on.

surely someone in apple would have run some common neural net architectures through the rtl simulator before tapeout and checked that the performance meets expectations??

> it has received ~8,000 requests from two of your scanning hosts

If it were 8000 requests per second, this might be worthy of some investigation.

But 8000 ntp requests alone consume far less than 1 us cent of compute + bandwidth. This isn't worth lifting a finger over.

And how much did this investigation cost, when you account for all the hours spent by investigators, lawers, judges, etc?

My guess is it was a huge net loss for the taxpayer.

The more effective move for the nation would have been for this guys boss to have confronted him and said "computer parts keep going missing on your shift. You have 2 weeks to make sure all the computers are back to their exact original config or we're gonna investigate and you could end up in prison".

Not soon there won't be.

There is no reason you can't have phased array antennas of tens of elements on the handset and thousands on the masts, to give a massive bandwidth increase.

It will take a lot of generations to get to the limit physics sets.

In quite a few countries, a false positive can lead to someone getting put in prison, entirely without any human viewing said image.

It's all a matter of 'the image hash was on the database, therefore OP is guilty. 10 years incarceration'.

I wouldn't want to be enabling such a system.

It also has a fairly high false positive rate, and due to the nature of the content paying someone to verify false positives is near impossible both legally and morally.