eis

I explicitly mentioned he did not post any comments.

> He never even uploaded a single video nor post any comments.

He wasn't really using it in that way. It was a pure video watching account.

Still, the way they handle it is extremely bad. It takes a long time to get the account re-instated and they just take down the whole thing in the meantime. What they should do is make the account read-only and if it was taken over by a malicious actor then just rollback to a previous state asap. For accounts with millions of followers they should have emergency contacts and a process that resolves this within less than 24h. These Youtube channels generate serious revenue that easily would justify the costs.

That's just insane. Imagine Mastercard or Visa shutting down card payments for Amazon.com randomly without anyone with half a sense doing a proper review. No circuit breaker in place to stop automated takedowns of huge accounts.

Google and especially Youtube seem to be run in a reckless, careless and incompetent way in this regard. Similar to Youtube seemingly being uninterested in solving the huge comment scam problem. I don't buy that the company that successfully filters out billions of spam messages in Gmail can't detect these really obvious spam/scam comments on Youtube. It's just baffling.

This happens way too frequently nowerdays. A friend of mine had his Youtube account banned. He never even uploaded a single video nor post any comments. How could it even possibly have created a reason to be banned? Of course support requests went into the void and he just gave up. Luckily his gmail was not affected.

A company I was involved with had a business account with Wise (Transferwise). One day out of the blue they froze our account and said only withdrawals from now on. The "reason" they provided was Violation of Acceptable Use Policy. They wouldn't even say which clause or give any details whatsoever. When asked for review we got an reply that after extensive review they can't reinstate the account, the decision is final. During the whole process it was painfully clear that they didn't read our messages or even try to understand the situation. After a few weeks and many emails with supporting evidence that we really can't see how we broke any of the terms the account was reinstated - a surprise to us. It seems like eventuall someone actually did take a look at the issue. But we never heard why it was frozen in the first place or what made them reinstate it. We probably just got really lucky. During that time we couldn't do payroll. We couldn't collect customer payments. It could have killed us did we not quickly scramble to get an alternative setup.

The common theme is that a company makes ill-advised banning decisions that have a big impact on a single user and then appeals are not treated seriously. This can have devastating effects. Usually this happens after the company who is running the platform grows to a certain size where they don't need to care about a low percentage of users anymore and rather save on support staff costs. And to add insult to injury they lie in your face about having conducted "extensive" reviews and having "carefully" decided all the while it being clear that they spent no energy or time on the matter. What's the solution here? Some mandatory third party appeal service which you can call upon? Many victims of wrongful banning would be glad to carry the costs of a review by a real human to get their accounts back.

Yes that tachyon visualization is pretty good. Here a video of a hypersonic jet flyby: https://www.youtube.com/watch?v=PQydRIxoAU0

You cannot hear any sound as it approaches. When it is at the cameras location, you hear a boom as a lot of soundwaves hit you at nearly the same time. After that you hear a mix of the soundwaves that the jet produces while traveling away and the other part is the soundwaves arriving from earlier during the approach. Unfortunately we can't really distinguish between those with our ears though a machine could. With light you would see indeed the object suddenly appearing out of nowhere and then splitting into two. One traveling into the direction of time and one seemingly going backwards.

What makes things a bit weird are for example time dilation and length contraction. Our equations for those would result in imaginary numbers (e.g. with "i") and at least to me it's a bit unclear as to what that would mean.

It is also unclear to me if a particle traveling faster than light would still see light traveling at the speed of light from its frame of reference.

FTL comes with causality issues/paradoxes, the easiest to comprehend video on that topic that I've seen is https://www.youtube.com/watch?v=an0M-wcHw5A

From the perspective of a being that can sense only sound the hypersonic traveling jet would be equally confusing - it could assume that there must be things happening in another dimension, one that it cannot sense/observe. And so one could theorize that an object being able to travel faster than light could exist in a dimension outside our 4 spacetime ones - one that we cannot observe directly.

To an observer who is only able to sense sound, in his reality nothing should be able to move faster than the speed of sound because otherwise he'd observe event B happening before event A. Of course in our reality moving faster than the speed of sound is possible. And in terms of sound we hear the object actually moving back in time. That is we hear it first when it's at our location and then we hear the sound it made further and further back in time. From the jets point of view though he was not moving backwards in sound-time.

For us, the same would apply to something moving faster than the speed of light. We would observe it moving backwards in time but that does not mean it actually does from its point of view.

Point being we have to be very careful about distinguishing between when we observe something happening and when it actually happens. Observing something going backwards in time is fine while actually traveling backwards in time probably is not.

This is the second time that Monty left a company that he had no control over even though it was his software "baby" that the company was selling. At least the first time he made some good money. But I'm not sure if that is the case this time.

Third time's the charm? Monty you should retain control over the companies that you want to sell your product under. If you don't control over 50% then it's not "your company" anymore.

  md5: d56d71ecadf2137be09d8b1d35c6c042
  sha1: 8de2fdb04edce612738eb51e14ecc426381f8ed8
  sha256: b1674191a88ec5cdd733e4240a81803105dc412d6c6708d53ab94fc248f4f553
  size: 184292 bytes

These are from MacOS 10.15.7 The file matches the officially released whitepaper pdf.

How could it be that versions of MacOS released in 2019 have the paper due to a court case from 2021? Even ignoring the temporal issue I really can't see any plausible connection.

You said Electron would kill the performance gains of Perspective. You also said it would make a workstation fans spin somehow.

The performance gains in rendering of Perspective are due to using canvas + Webassembly. This would be exactly the same in Tauri or Webview2 because it would run in the same browser rendering engines which execute the Canvas renderer and Webassembly like in a browser or in Electron.

> The key benefits observed from the transition from Electron to WebView2 include reduced memory usage and a lowered disk footprint as resources are shared with Edge.

That's exactly pretty much where the benefits end. The memory usage reduction is even debatable if you are not running Edge as your browser. See WebView2 as an installed Electron core that's pre-installed on Windows. Doesn't change rendering performance.

> In the case of Sciter (quoted in my parent comment), Sciter is a bit different: The frontend is HTML, CSS and JS, but the app's local backend is whatever you want it to be: C#, C++, Python, Assembly, etc. That's why it's popular among performance-critical software such as malware scanners, where you need native performance but as well as a you know customers that like a nice UI.

Yea that's why I didn't list Sciter in my reply. With Sciter you don't get the system webview - you get a custom minimalistic web html/css rendering engine that supports a limited subset of the HTML5/CSS3 spec. There's no WebGL and there's no Webassembly. So you'd have to bring your own libraries for that. Easy enough with Webassembly and they could even skip that part and go native but with Canvas and trying to run on all major platforms... it's just a lot of added work for minor gains.

I'm not against libraries like Tauri. Perspective could totally use that to make a desktop app. It would not have any advantage when it comes to rendering performance though.

I don't see why rendering this in Electron (which is essentially just a bundled Chromium browser) would perform differently than in a standalone browser. Neither do I see how projects like Tauri or WebView2 which just embed the system webview would give better rendering performance. Microsofts webview is even using the same rendering engine as Electron - Blink. What you'd get is significantly smaller binary size but most people don't care much about that.

Online demo that runs in the browser: https://segment-anything.com/demo

Initial impression: it seems to be pretty good at identifying objects but when you look at the cutouts without the glowing outline you'll quickly realize the edge is rather rough and thinner parts sometimes get omitted completely. This might be though because those borders they draw are too thick and omitted from the resulting cutout.

Example: https://i.imgur.com/S57c5Cj.png

In the kids drawing both extracted person had no arms.

With the classic double slit experiment both space coordinates affect each other. With this time axis double slit experiment one has to ask if t1 affects t2 or if t2 also affects t1. The latter would raise a lot of questions. I don't understand the paper well enough to know which of these cases they are observing but I assume only the t1->t2 case? I'd appreciate if someone with more domain knowledge could enlighten.

Edit: after more thought I guess it would make sense that two slits after each other in time show a change in wavelength if they are brought close enough together.

Essentially the time separated slits act like a space separated double slit experiment wherin the slits are oriented on an axis parallel to the direction of the light path instead of at a perpendicular one. The wave like nature does not just expand in one dimension in space but all three and so there must be interference in the direction of travel. Do I have the right idea with this?

Yes, MacOS never had similar limitations as on iOS. I'm using Firefox which uses its own renderer as a daily on my Mac. There are signs that iOS will (be forced to?) drop that restriction as well in the future. Google started working on a Blink based Chrome for iOS.

The results are pretty impressive. I wonder if the general idea can be applied with a bit more data than the roughly 21 bytes in this version. I know it's not a format that lends itself to be configurable. I'd be fine with placeholders that are say around 100-200 bytes. Many times that seems enough to actually let the brain roughly know what the image will contain.

The news hit yesterday and got them +10% before close but it seems pre-market they are down currently -23%. Doesn't look like there is much reason to rebound to anywhere their previous levels in the short term.

A Macbook Pro M1 would have the base M1 CPU while he was referring to the M1 Pro CPU in something like a Macbook Pro w/ M1 Pro. It's confusing naming by Apple.

Ignoring the question of who is actually doing something wrong and the fact that most customers on lower tiers wouldn't even be able to get most of the presented information, it seems absurd to me that a customer on an Enterprise plan that costs thousands of dollars a month can't get a simple "this is why those requests trigger the firewall" answer from CloudFlare. The diff between the request that passed and the one that triggered the rule is simple enough and the WAF is not some opaque AI I hope.

Why can't the triggering request be replayed through the WAF with output that shows the scores for each bit?

That does not get you hourly time slots.

You either want something like

    floor(timestamp / 3600)

or

    timestamp - timestamp % 3600

The energy crisis in Europe is very real. But the problem is the fossil fuels that have been cut off because of Russia. If the supply from Russia were not cut or if the EU had increased production from renewables faster in the past decade then this situation wouldn't have arisen. There is ample wind, solar and hydro energy potential left. We've accessed a small fraction of it.

Now Europe has been shaken and will hopefully learn the lesson and become completely independent in terms of energy in the next decade or two.

In contrast to fossil fuels, Europe can ramp up electricity generation fairly easily as there is plenty of untapped potential left. That would reduce prices though renewables are already cheaper than fossil fuels. But even if renewables are not cheaper than oil and gas, the gained sovereignty is worth it alone. That much the war in Ukraine has shown very clearly.

In Germany charges, taxes etc on electricity are more than 50% already.

Of course there is. Hydro storage is probably the biggest one that has been in use for decades. You pump water up a mountain during over supply and then let it run down through turbines during times where there is no wind/solar etc.

There are plenty other methods to store energy but most involve gravity. Batteries can be used for small scale rapid reactions to changes in demand or production. Actually fully switching to EVs as a side effect can create a distributed system of energy storage.

What major changes to the grid are required to make EVs work and why do they have to be "smart"? Electricity consumption will increase but it will be quite uniform if everyone is switching to EVs. A few upgrades for added capacity will be needed but I don't see any major issues with that. And while all new cars have to be emission-free by 2035 that does not mean all cars in use will be. It'll take probably till 2040 to 2045 for the vast majority to be EVs and that's plenty of time to do the necessary upgrades.

While coal was a major part of the electricity generation in Germany, it has been declining rapidly in recent years and is being phased out completely before 2038 which is roughly in line with the switch to EVs.

In the meantime renewables, especially wind have been growing rapidly and have accounted for more than 50% of all electricity generation in 2020. In 2021 Wind alone was nearly a quarter of the whole electricity production.

It seems to me that the rollout process is flawed.

A change should be rolled out to a small fraction of nodes and then monitored for an extended period. 5% of requests returning errors would easily be spotable. Only when a new release has run stable on that portion for some time should it proceed to a bigger subset. You probably want to do this in several steps at the size of CF.

We also learn that it took customer reports to get the investigations rolling but 5xx errors are easily monitored so it points at internal monitoring being lacking even though it's hard for me to believe that they don't have an eye on this already.

It's not the first time that a deploy has brought CloudFlare (partially) down. From the timeline we see there's several hours between incident investigations starting and the rollout being stopped. The rollback should be the first thing considered even before looking at what the actual issue is.

Ideally you have someone sitting next to a red rollback button during a rollout whose only job is it to have an eye on all automatic and customer error reports. :)

An impressive speed run. I wonder if it counts as tool assisted.