Honestly, Claude Code running as an agent on your machine under your own user account is both really powerful as a troubleshooting tool, but also very scary from both a privacy and risk point of view. If you use it in your local environment, then definitely put it on Manual permissions and approve every action, and realize that everything it reads goes up to the cloud. I only run Claude Code in a virtual machine with its own very limited local user account. I sometimes use it that way for troubleshooting a Windows problem but I have to copy log files into the VM, and run the commands it proposes myself. But at least that way I've got some protections in place. Everything you copy or paste into the VM potentially goes to the cloud, so remember that.

It's a bit odd to me that ships just sit rusting for decades when they're mostly made of a 100% recyclable material, which we eventually just sink to the bottom of the ocean rather than scrapping. I get the economics of it because it's dangerous to do shipbreaking, but it's still weird.

Communist Combatant Cells is not particularly better.

We use wheels on Mars rovers that are pretty capable for traversing rough terrain. I think that when you combine wheels with the ability to "put the brakes on" and attach those to the end of articulated appendages you get a pretty robust solution. When going downhill, the robot can actually use regenerative braking to slow themselves down (and in that case actually recharge the battery a bit). An organism that had wheels and the ability to adjust the braking force would be pretty nimble.

I think you're correct. If it was possible for evolution to have created wheeled actuators, then the organisms with those appendages would have dominated long ago.

If you've never had a boss like this -- a lazy narcissist -- you're lucky. But if you have, the pattern is familiar and unmistakable:

* Doesn't understand the issues.

* Doesn't do his homework.

* Believes he's a genius whose "gut" overrules people who do understand and do do their homework.

* Plays his subordinates off against one another.

* Highly susceptible to information from random sources: thing he saw on TV, or heard on the golf course.

* Not particularly fussed about contradicting himself because he doesn't care about being right, but does care about being in charge.

* The more chaotic he can make things, the more in charge he feels.

* Has a cynical view of the people he works for but has a genuine gift for flattery when it suits his purposes.

Once you've seen this up close it's really obvious when you come across it again. You can't *not* listen to a lazy narcissist boss because he absolutely will try the stupid shit he says he will. But you can't count on him either, because he has no commitment to anything other than how he looks and feels *right now*.

Neither. Commerce didn't drop it, it got taken away and given to DoJ, but not because Commerce couldn't handle the invetigation of where the chips were going. The case goes to DoJ because we're talking about criminal charges of smuggling, corporate fraud, violation of national security statutes. Also because the Commerce investigation was tained by potential political interference. And possibly because it ran afoul of a counterintelligence operations siloed in the DoJ -- although by this point the cat is so out of the bag everyone must assume they've got a FISA wiretap warrant on them.

However, the *effect* may be to slow the progress of the investigation and delay any actions being taken.

The first reason for this is that criminal investigations are slower, the standards of proof higher, and potentially involve a criminal trial. Commerce could simply make an administrative law determination that put Megaspeed on some shit list and come to some kind of agreement with NVIDIA.

The second reason is that putting this into DoJ's hands is putting it into a department that has had staff turnover and loss of institutional knowledge, strife between career personnel and political appointees, and difficulty replacing lost employees. It's also led by lame-duck appointees, facing turnover in the next administration. So if they don't make progress in the next two years, allowing for transition delays, it could be years before anything gets accomplished. By then a lot of people will have left their jobs, memories will get hazy, evidence will go missing, and defensive positions constructed.

So while the investigation hasn't been *killed*; it has been delayed so long that it's effectively in suspended animation. It's not possible for Commerce to sanction anyone because the investigation is out of their hands.

That would depend on the exact wording, of course, and a lot of those contracts get written so that the agreement is only with the original owner, and I assume most of those vehicles would be on their second owner by now. Interesting though.

I'm a little surprised that someone buying a Tesla 10 years ago would believe the pie-in-the-sky promises about self-drive, or using it as a robo-taxi. I certainly thought self-drive was a long way away, but I have experience with computer vision. I just wouldn't buy a product with the promise that some software update sometime in the future would deliver the rest of the features. Either way it doesn't matter to me because I've never owned an electric car, and I want to see real people in my area (rural Canada) using them for a few winters driving 40 km each way to work, on unplowed snowy roads. But if I lived and worked in a city, I'd probably have bought one by now, and would still be very hesitant to use self-drive. But I know of a guy who has a Tesla and has it self-drive him from his driveway to his destination about 2 hours away pretty regularly. So I guess they finally delivered on the promise.

Money may be an aphrodisiac but come on, Jensen, look in the mirror. There's limits to anything.

I understand the "we're all supposed to take every opportunity to hate Elon" thing, but your premise doesn't hold up when checked against the facts. The whole genius of Tesla was that it was the first company that made electric cars that *looked* like regular cars, and that people wanted to actually buy, and he's proven to be quite good at business decisions so I have a hard time believing he didn't have a hand in that. The cybertruck was an anomaly. But if you look at the problem of electric pickups, they're a significantly harder problem to solve than an electric car because transporting a couple relatively light humans vs. transporting heavy cargo doesn't scale well with our battery technology. You can try super-sizing the battery, which just cranks up the cost and makes the whole thing even less efficient, or you can do what Tesla did and re-think the frame and body from the ground up to get rid of as much weight as possible. It just so happens that the material they picked to solve this problem can't be formed like regular door panels, so they decided to lean-in to this angular weird look and see if people would accept it. I certainly disliked the look of it from the start, but I've also been wrong about which products would succeed, and you don't know until you try. It turns out that people buy trucks on looks more than they buy them on specs, so it didn't pan out. But this hardly means the people at Tesla are constantly shooting down Elon's crazy ideas. This was a one-off.

Look, I mean... yeah, it doesn't work exactly the same way human intelligence works, but I use Claude Code as a coding assistant every day, and within the domain of writing code, if this doesn't qualify as AI... then I just don't know what the difference would be. No, it hasn't really solved code review, and it hasn't solved big software architecture yet, but as far as day-to-day coding of small to medium tasks, I would argue that it's already surpassed the intelligence of the average programmer.

If you're on the Ultra plan with Claude Code, it's $200 per *month* and honestly, running about 4 separate sessions constantly working through to-do items, I don't even hit 25% of my weekly usage anymore. I was hitting the limit with Fable 5, but Opus 5.5 is much more efficient token-wise. I was thinking of dropping to the $100/month plan. Though I have suspected that the Claude Code plans are being sold at a loss, and they're much, much cheaper than pay-as-you-go pricing where you pay by the token.

As soon as everyone switched to using Claude Code to write all their software, it became affordable to rewrite in Rust, and people are seeing significant speed-ups and a reduction in memory usage, which is amazing. Using Claude to migrate an existing app to another language is a pretty good task for it because it has the existing one as a target, and AI agents work really well if they have a goal to iterate towards. The fact that human programmers find Rust to be pretty ugly doesn't seem to bother anyone, once the agents do most of the interaction. For the line-of-business app that I work on, I'm leaving it in C# for now because that's what I know, and I still review all (or most of) the code it writes. But if I were vibe-coding a weekend project, then telling it to use Rust is kind of a no-brainer.

Yes, it really is getting good. But I've still been reviewing almost all of the code it writes (I'm using Opus 5.5 mostly). Just yesterday I caught it writing the same logical rule (something like "only show this button if such-and-such condition is true) a second time about 10 lines away from where it was written the first time. Sure, it fixed it when I pointed it out, but that seemed like low-hanging fruit that it should have picked up immediately. At the same time, it can investigate pretty fast, so it does a pretty deep dive when writing a design document, and will pickup on things I probably would have missed in the first pass. Plain coders who just write what they're told are now a thing of the past. This has happened in some industries before, but I don't think it's ever happened this fast. Opus 4.x that could actually write mergeable code was released last November, and by this summer we had Fable 5 and since August I've hardly written any code. I'm a solo dev working at a small company and I have a lot of autonomy, so this technology just tripled my output over the span of a month, even with me still reviewing every spec, design doc, and almost all the code it outputs.