Did it though? EFF says that the number of impressions their content received is why they left:

Those [2018] tweets garnered somewhere between 50 and 100 million impressions per month.

Then

Last year, our 1,500 posts earned roughly 13 million impressions for the entire year. To put it bluntly, an X post today receives less than 3% of the views a single tweet delivered seven years ago.

But, I wonder what the real numbers actually are. Do we think Elon is honestly reporting real numbers to people? And, of the “impressions” that are real, how many of those are actually from bots rather than actual human users?

IMO, one of the biggest tricks Elon has managed to pull with Twitter is to convince celebrities and brands that it’s still a thriving site full of other people, leading to them sticking around because supposedly no other site gets as much traffic.

This is good for bitcoin.

IMO if a door has an electronic lock and normally opens with a button (which is dumb), the backup system has to be something you can use if you’re on fire and have a concussion after a crash.

Apparently Audis have an electric lock but still use a regular latch instead of a button. In an emergency you can open the door by just pulling harder on that door release latch. In a Fisker you can do it by pulling on the handle twice. In a Ford Mustang if you yank hard on the door handle it acts as a release. These are all things I can see someone doing in an emergency without thinking.

Also:

5. How is it that the company that makes this shit has a price-to-earnings ratio of 326 and a market cap of $1.3 trillion. Meanwhile Toyota makes good dependable cars and has a P/E ratio of 11.41 and a market cap of $266 billion, even though Toyota sells 10x as many cars.

Musk is much more of an Edison than a Tesla. If he’d been honest he would have named his cars Edison. Then, the cool rebadge could have been Tesla. But, even he was smart enough to realize what an asshole Edison was, even if he didn’t recognize the Edison in himself.

It seems like there’s a market for a company that will buy Teslas ultra cheap, modifies them heavily, then rebadges them like Alpine does for Renault, AMG does / did for Mercedes, Abarth for Fiat, etc.

These days those are all subsidiaries of the main brand, and even before that they had a cooperative relationship with the main brand. But, I can imagine a setup where the main brand doesn’t support or approve of what the modifier company does.

Wonderful. DRAM and NAND flash companies are a cartel and are able to make their prices absurd by cutting supply when demand is high.

they’re reducing production in places.

Really? Where have you seen that reported?

So, maybe this will result in the companies that make SSDs ramping up production. Maybe once the bubble bursts there will be a huge glut of SSDs and so they’ll be cheap.

C’mon bubble, do your popping thing.

I hope datacenters are ordering pallets of 8 TB SSDs and are instead getting pallets with bricks on them.

They don’t support sending messages over a serial / USB / network connection to say “battery is almost dead, shut down cleanly while you can” right? As far as I can tell, that’s the one key feature that a UPS has that a portable battery doesn’t.

I took a look the other day and was amazed at how little UPSes have improved in the last few decades vs. everything else battery-related.

At this point, I’d expect a consumer-grade UPS to have something like a Raspberry PI attached, and run a web server. I’d expect it to not just have a serial port for signaling, but to be able to run custom BASH scripts to send messages out to any attached device warning it about being on battery and keeping it up to date on the battery status.

Some form of “AI” will always be around. It has been around nearly as long as we’ve had computers. We’ve even had AI chatbots since 1966.

But, the dot com bubble is a bad comparison. If you look at a graph of Internet users over time you can barely even see the dot com crash. The Internet was a massively useful phenomenon and more and more people kept using it. The dot com crash was basically an overestimation of how quickly people were going to adopt it combined with a massive drop in the value of Internet-based ads.

What’s much more likely with AI is another AI Winter where a few things stick around, but mostly AI goes back on the back burner for a few more decades.

I don’t think it’s an “obsession with cars” or that people are “ignoring” electrified rail.

The problem is that there are things that are in your direct control, like buying a car and using the roads which exist. Then, there are things outside your direct control, like trying to get your government to install electrified rail. Even if you have a really responsive government that isn’t captured by special interests, getting rail built and up and running can take a decade. And, if you need to get from A to B, you can’t wait for a decade. Even if you’re really pro-rail during that decade you still need to travel, so you’re likely to be forced into getting a car. Once you have a car, then rail might become less of a priority because you are now a car user. Maybe eventually you’ll still want to use the rail system, but for now you have a car, so your priorities are still going to include car priorities.

This all changes if you live somewhere where there’s already great rail service. In that case, you might already have rail available when you move somewhere and all you need to do is encourage your local government to keep funding rail and not subsidizing cars. At that point, the car driver demographic is small and easy to ignore.

The problem is in switching from one system to the other. You need a government that is going to weather the complaints from drivers for years while the rail infrastructure is being put in place until you get to a point where drivers can start selling their cars and switching to rail. That’s really hard to do though, because going from poor rail infrastructure to good rail infrastructure can take a decade, and politicians often have terms lasting only 4 years or so. That means that they have to take on the expense and pain of starting a rail project and then facing an election long before the system is up and running. It’s actually surprising how many politicians are willing to do that, given that it’s so hard on their political careers. It’s unsurprising that most of them don’t want to do it because it means getting re-elected is much more difficult than if they just stick with the status quo.

Meanwhile, the special interests like car companies, car dealerships, gas stations, etc. are all going to be lobbying against any rail projects. In North America it’s even harder because car companies are local, whereas the companies that make trains are mostly European. So, the car-related lobby can talk all about the impact on local jobs, whereas the rail lobby has to deal with the jobs mostly being in Europe. Even without that, it’s hard to change things because of the issue of diffuse costs and concentrated interests. Hundreds of thousands of commuters might benefit from a rail system, but it’s probably not their #1 priority, it’s something they care about, but at best it’s #4 or #5. Meanwhile for car companies, etc. it’s a top priority. While you might not want to go to every city council meeting where this is being discussed. It’s almost certain that the auto lobby will ensure their voice is heard because it’s at the top of their list.

In the end, it’s a lot more complex than just people being obsessed with cars, or ignoring light rail.

It will slightly improve the chances. But, is that enough?

Imagine you had an intern working with you on a project. They didn’t know anything about SQL injection, cross site scripting, etc. You probably wouldn’t give them a task where that was a concern. If you did, you’d watch them like a hawk. Because they’re an intern, the amount of code they’d produce would probably be pretty low, and it would be pretty low-quality overall, so it would be easy to spot mistakes that would lead to these kinds of vulnerabilities.

An LLM has the understanding of the problem space that an intern does, but produces vast amounts of code extremely quickly. That code is designed to “blend in”, i.e. it’s specifically trained to look like good code, whether it is or not. Because of “vibe coding”, people trust it to do all kinds of things, including implement bits where there’s a danger of XSS or SQL injection. And the way Claude Code ensures it doesn’t generate those vulnerabilities is… someone says “hey, don’t do that, ok?”

Having that statement in there is better than not having it. But, it’s just a reminder that these things aren’t appropriate for writing production code. They don’t actually understand what XSS or SQL injection are, and they can’t learn. They don’t know why it’s important. They don’t have a technique for checking if their code actually has those vulnerabilities, other than passing it to themselves recursively and asking that other version of themselves to generate some text that might flag if those vulnerabilities were spotted. But, AIs are famously sycophantic so even recursively using itself, it will generate text to “please” itself and probably write something like “your code is great and I can’t spot any vulnerabilities at all! Congratulations! [Emoji] [Emoji] [Emoji]”

The system prompt discovered in the leak explicitly warns the model: “You are operating UNDERCOVER… Your commit messages… MUST NOT contain ANY Anthropic-internal information. Do not blow your cover.”

This is so incredibly stupid.

You’ve tried security.

You’ve tried security through obscurity.

Now try security through giving instructions to an LLM via a system prompt to not blow its cover.

You’d never live long enough to grow your $1b into enough to worry someone like Elon Musk, Bill Gates, Jeff Bezos, etc.

The problem is that everything I’d want to do would be opposed by other billionaires worth hundreds of billions.

this is the beginning of the end for the AI bubble

The end of the AI bubble has been beginning for years. The end of the beginning of the end of the bubble might take a few more years.

Those aren’t pictures from Brisbane.

In places where biking is well established as a way to get around, people think they don’t need safety gear. But, when a city doesn’t have a lot of bike infrastructure, cars are going to be constantly driving way too close to bikers.

This is an image that’s actually from Brisbane:

Or you could have gone with this one:

From this story: https://www.abc.net.au/news/2020-11-11/many-busy-brisbane-cycling-routes-make-no-allowance-for-riders/12862402

Thanks, that’s helpful.