Fascinating, but if he had low pressure on the rail that would point to a clog or possibly a pump issue, right? Wasn’t really clear whether the lean condition was due to the map or a mechanical issue.
I have a dirtbike and experienced some stuttering and stumbling under load after getting hot. Used Claude to help troubleshoot, as the plug was white and it was real lean. Replaced the pump (a 300$ module) and saved a ton of money and annoyance having a shop deal with it. 2000mi later zero issues. Absolutely game-changing. I use a dumb aftermarket fuel controller to add fuel under certain conditions, but would love to try and tune it myself some day.
> Claude suggested that it would run the fuel pump on its own. Um and I said, "Hey, if you can do that, why not? I don't think you can do that." And it started priming the fuel rail and it kept priming it and kept priming it. The fuel pump was on for about 10 seconds and I started to freak out uh and requested it to stop because that was the first time that Claude had actually done something inside of the computer to have a change.
Although we've seen this play out with handing cloud keys to Claude and it just going off on its own, seeing it poke around and make physical, real-world changes in its clumsy, cavalier way is both awesome and terrifying.
Would love to see this being used to add keys to cars. A lot of cars now have tablets and screens that could be used to do this, but for some reason (money) I have to take my VW to the dealer...
I think the whole market for aftermarket keys could be completely eradicated with good software UX (and low pricing).
Not these days. It won't even look at my projects for security flaws any longer. Has a little pop-up that says "This output cannot be viewed." or some such.
You fall back on having to use Chinese models. In theory the Chinese models could also be trained to sneak in new security flaws while Claude and Codex would avoid looking for them.
Claude and codex could also be trained to sneak security flaws. They could even target specific people. Not to mention that all their data goes through NSA headquarters so you don’t even have the option to run it locally in any shape or form.
LLMs end up changing the consumer market not the way big corporations would like to. It is impressive to see that we'll be able to give ownership back to the customer, or enable others to do so.
Very nice demonstration.
Wait until those big corporations start using LLMs to learn how to lock things down even further, or implement checks that disable your car when it detects alterations to the software “for safety reasons”, or go after drivers who do these things for loss of revenue, etc. A ‘94 Jeep will be fair game but a new car might not be.
>big corporations start using LLMs to learn how to lock things down even further
Why would they? They just need to fund lobbyist to make it illegal for any AI to reverse engineer anything copyrighted or with IP.
Sure, you can use open source AI to do it but not everyone will/can do it.
So, they don't need to ban everyone, they just need to make sure the final code isn't available freely on GitHub. That would minimize the damage, it's not a perfect solution but more than enough to protect the company.
They've been doing that for a decade. I think overall the balance will shift in our favor but Big AI labs will do everything they can to try and stop us.
Claude is quick to refuse to help with reverse engineering and I suspect it's only going to get worse over time.
Yes, they have been on that road for years. Tesla showed everyone the way with things like always-on connectivity or not being able to change a headlight with another original one without Tesla’s blessing. What I meant is that whatever I can do with an LLM they can do more. And if they can invoke “safety” that'll have the law on their side for some time.
As far as I can tell no manufacturer disables the car yet when detecting modifications, or take drivers to court for doing it, yet.
They are now required to do just that, if they want to sell to the EU market. When you hear the term "cybersecurity," the real subject is the protection of revenue streams from attack by end users.
Lockstep processors, hardware security modules with immutable unreadable keys, and PKI keys that the mfg holds the private for that can only be accessed on their online services.
It’s locked down now and this started in 2018 or so. It’s just worse now.
To automotive, there is really still no right to repair. They get around it by selling “3 day passes” to their service, but with their hardware and at their prices.
They’re all moving the Tesla system of more centralized brain systems which will be even more protected.
Sad to hear. I am not a car owner, but lived in the EU for most of my life. The consumer market owes quite some consumer rights to the EU, so I guess all we can do is hope that they'll eventually crack down on these measurements again.
can someone explain the root cause of the smog test failure to me ? it sounds like the car was using a California smog test-specific fuel map at the speeds of the test. Wouldn't that make it pass the test, not fail it ?
Yeah doesn’t really make sense, there’s very likely a mechanical issue the LLM “fixed” with a richer fuel mix. But also it’s a 94 XJ the reason the CA map was there was probably because they got abhorrent gas mileage without it.
You can adjust the mixing ratio of fuel and air in a way so that the exhaust has reduced emissions. This will reduce performance of the car but allows the exhaust to pass the emission tests.
Definitely one of the Claude video skill stacks that are going around. Dude should make a whole separate video about the animations, he'd get as many views again just for that.
It also didn’t solve the root cause of the issue. ECUs don’t need to be overwritten to get stock performance and emissions. Something is wrong with hardware/mechanical that needs to be solved.
What he did was technically interesting, but probably not a great solution to the actual problem.
LLMs have been pretty good at breaking bad crypto in cars. In my Hyundai, they were able to figure out the right incantation to get the car to run unsigned code off a USB stick. Although I guess we're in trouble once carmakers use LLMs to lock things down even more
I have a dirtbike and experienced some stuttering and stumbling under load after getting hot. Used Claude to help troubleshoot, as the plug was white and it was real lean. Replaced the pump (a 300$ module) and saved a ton of money and annoyance having a shop deal with it. 2000mi later zero issues. Absolutely game-changing. I use a dumb aftermarket fuel controller to add fuel under certain conditions, but would love to try and tune it myself some day.
> Claude suggested that it would run the fuel pump on its own. Um and I said, "Hey, if you can do that, why not? I don't think you can do that." And it started priming the fuel rail and it kept priming it and kept priming it. The fuel pump was on for about 10 seconds and I started to freak out uh and requested it to stop because that was the first time that Claude had actually done something inside of the computer to have a change.
Although we've seen this play out with handing cloud keys to Claude and it just going off on its own, seeing it poke around and make physical, real-world changes in its clumsy, cavalier way is both awesome and terrifying.
I think the whole market for aftermarket keys could be completely eradicated with good software UX (and low pricing).
Have to make the Chinese one do it.
You fall back on having to use Chinese models. In theory the Chinese models could also be trained to sneak in new security flaws while Claude and Codex would avoid looking for them.
Why would they? They just need to fund lobbyist to make it illegal for any AI to reverse engineer anything copyrighted or with IP.
Sure, you can use open source AI to do it but not everyone will/can do it.
So, they don't need to ban everyone, they just need to make sure the final code isn't available freely on GitHub. That would minimize the damage, it's not a perfect solution but more than enough to protect the company.
Claude is quick to refuse to help with reverse engineering and I suspect it's only going to get worse over time.
As far as I can tell no manufacturer disables the car yet when detecting modifications, or take drivers to court for doing it, yet.
Lockstep processors, hardware security modules with immutable unreadable keys, and PKI keys that the mfg holds the private for that can only be accessed on their online services.
It’s locked down now and this started in 2018 or so. It’s just worse now.
To automotive, there is really still no right to repair. They get around it by selling “3 day passes” to their service, but with their hardware and at their prices.
They’re all moving the Tesla system of more centralized brain systems which will be even more protected.
very cool project anyway.
What he did was technically interesting, but probably not a great solution to the actual problem.