Claude Code and Codex reverse engineered my 1994 Jeep's ECU over a $3 Arduino [video]
Discussion 157 comments
Even if AI goes away tomorrow, the tools will still exist.
Is this just "vibes"? Maybe, but I now have a full, verified spec, against which to build software for myself, also with a little bit of "vibes".
Is this a good or bad thing?
that is entirely dependent how accurate the spec is
Hard to judge though, most people are just winging it though and shoving it out the door as if its impressive.
But yes, I think its probably bad because you probably didnt learn all that much and became more reliant on Anthropic. Unless you dont care about learning, then who cares.
Same with my bespoke IRC client, RSS feed hydrators and a dozen other small utilities I've vibe coded. All stuff that I've been planning for a long time, but haven't found out the time to dedicate to them.
Now I'm saving actual money every year because I could drop subscriptions applications.
Not every application needs a weekly release with new groundbreaking features. They're just one and done. Maybe an update if you hit a sharp edge while using it.
I suspect you're right BUT still maintain a bit of optimism. We miiiight end up in a future where most hardware comes with an OK-ish app/interface and some sort of LLM-Friendly technical documentation so if you don't like $aspect of the device, you're free to roll your own after indemnifying the manufacturer via something like an eFuse.
or maybe I really am naive :(
I think most people on here fail to realize the demand to customize and build things hasnt gone up. Just because llms exist now doesnt mean normies all of a sudden want to start getting into programming, they dont care.
The only number thats gone up is people who were slightly interested in programming/CS before but were too undisciplined to learn things themselves. Llms are great for those lazy people.
I still get the joy of translating ideas and orchestrating and architecting solutions to problems I have in real life, but just at a quicker pace.
If some of those tools and solutions can help other people, that's great. If not, that's okay too. I hope everyone sort of has that mindset who's building today.
Me!
haha I dont even know github I use it as google
Yes I have full time IT job xD
(to stay within physical machinery, like the post.)
actually I have no idea what would llm do to make their machines better
Lack of imagination on here can sometimes be disheartening
people were able to revive, repair, optimize a lot of day to day devices, very nice view. that said cars and heavy appliances.. should be done in a big public warehouse safety lab
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.
That's true but we have the benefit of unlimited time, having physical control, the analog hole.
> As far as I can tell no manufacturer disables the car yet when detecting modifications, or take drivers to court for doing it, yet.
As far as I know cars are a special case in much of the world (US, EU, others?). Right to repair laws laws usually give you the right to use non-OEM parts.
Aren't DMCA anti-circumvention provisions the bigger concern in the US? I think the EU has an exemption to allow reverse engineering for the purposes of interoperability and repair.
Nothing prevents you from unplugging the cell modem. The car will still work.
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.
Making something impossible to do is more thorough than making it illegal to do
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.
Let's be clear that it's just incentives. If the EU economy was based on creating companies and new products there would not be as many incentives for consumer "rights" there.
Not for car owners/consumers in the EU. Auto makers and dealers get away with a lot of shady shit fucking over their customers in the EU all the time. For example we in the EU don't have the same trigger happiness of NHTSA mass recalls the Americans have, but matters of such auto issues/defects is left to individual member states or even at individual case by case to decide what action is needed, and responses are a lot weaker and more limp-dick than in the US, almost never leading to mass recalls, even for the same issues[1], mainly because the US is more litigious making car manufacturers to be on the edge and proactive to issues.
Like sibling said, there are incentives to regulate stuff that has no impact on EU businesses, shit like USB-C or replaceable batteries, but they won't touch the auto industry to mandate easier repairs and less lock-downs.
Brand & Models | Defect | US Action | EU Action
------------------+--------------------+----------------+------------------
BMW (2011-2018) | Coolant pump plug | NHTSA recall | Excluded from
N20/N26 engines | short / fire risk | (720k+ cars) | recall
------------------+--------------------+----------------+------------------
VW / Audi (08-13) | Timing chain | 10yr/100k mi | No recall; case-
2.0T EA888 | tensioner failure | warranty + $ | by-case goodwill
------------------+--------------------+----------------+------------------
Ford Focus/Fiesta | DPS6 PowerShift | 10yr/150k TCM | No recall; few
(2011-2016) | transmission fail | + buybacks | buybacks/payouts
------------------+--------------------+----------------+------------------
Hyundai / Kia | Theta II / Nu GDI | 3M+ recalled + | No mass recall
(2011-2019) | bearing/fire risk | lifetime wrty | (diff specs/plants)
------------------+--------------------+----------------+------------------
Tesla 3/Y/S/X | Autopilot nag & | 2M+ NHTSA OTA | No EU recall
(2021-2024) | UI font compliance | recalls | (UNECE rules)Pure battery EVs are considered to have no emissions-related components, which is why Tesla gets away with that they do.
Don’t hold your breath.
The smarter the models get, the bigger the loss for the big 2 when an open source model gets released that doesn't talk down to you. The value of open source increases and closed source falls exponentially.
The average person is likely to feel overwhelmed by this, which I think is what is likely going to shape the new economy gig of "AI engineers". Those that aren't overwhelmed by it and that may not only enjoy doing this but also potentially figure out a way to scale operations that allow doing these kinds of DYI reverse engineer hack jobs.
At least until the robots take over
To protect "corporate interests" :( That article is about gaming, but this will only be the start of it of course.
So, I switched to a Chinese model to do the work. Anthropic and the US AI companies are trying to fight a tsunami with a teacup.
I think the whole market for aftermarket keys could be completely eradicated with good software UX (and low pricing).
It's dominated by Chinese manufacturers of specialty tools (Autel, Xhorse, ...), which inevitably mandate bundled accounts, subscriptions, mandatory cloud features, etc. Even the older cars, you have either cheap Chinese tools, where you never know wheather it'll really work, or the manufacturer's tool which needs a dealer code, which of course the dealer won't have after 20 years.
I wrote about my 20 year old car-key-adding adventure here: https://jonas.bike/low-cost-diy-key-/-transponder-for-my-200... - it requires desoldering an eeprom, two pieces of hardware from aliexpress which come with pirated copies of software, and lots of hope that it'll work...
An open source community curated collection of scripts, which works with a cheap OBD dongle (which is all the chinese devices are: OBD hardware + per-car software), would be amazing.
The corsa also needed a "security code" that wasn't with the log book like it was supposed to be and would have required a trip to the dealer to get butI read about ripping the eeprom off the back of the dashboard to get the code but decided to cut my losses.
However they can't match and forward everything (not enough RAM) and miss loads of packets, which makes them useless for things like reflashing firmware or interfacing with OEM tooling.
But of course, the major US labs are going to get pressure from these vendors whose business models depend on obfuscation, and you know they're going to bend the knee and nerf the LLMs.
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.
I told it 'Let's find the limitations of this Linux system' then it made a benchmarking set of tools for disk i/o, RAM, etc
When using Gemini on my phone it also told me it'll be 'human reviewed.' when asking it about car prices.
Also when saying 'let's use best security practices'
o_O
Even my old reliable Muse Spark has gotten a little testy lately. I'm back to GLM 5.3.
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.
yolo engineering for ya.
I've got a "get you home box" for EFI systems with separate ignition, like Bosch LE-Jetronic that I built a few of as clones of an insanely expensive device some 20 years ago. It really just consisted of a couple of 555 timers and a bloody great MOSFET, with one 555 wired as a Schmitt trigger picking up ignition pulses and the other delivering a 1ms pulse to the injector coils every time the spark fired. This dribbled enough fuel into the engine roughly proportionately to engine RPM to let you drive it slowly to somewhere you could work on it safely, or even just onto a trailer.
And of course, Speeduino, Megasquirt, and a bunch of others are a thing.
very cool project anyway.
How tf are you getting it to RE?
But don't use it, obd read functions are fine but other functions for other diagnostic/actuator etc stuff is not.
I am taking this slow as I have other priorities, but as models advanced the stuff I Was able to do with limited time have increased, so with new version i still have locally, it may be possible to do more interesting stuff.
An example I recently did with my car was disabling auto start-top. not possible with consumer tools. Another simple one is maintenance resets - some consumer tools can do this. ABS bleeding - often you can't do it with consumer grade. etc.
This guy clearly put $5K+ of time and video production effort into this cool hack.
What he did was technically interesting, but probably not a great solution to the actual problem.
It definitely feels like a classic case of the operator reinforcing Claude to keep pursuing a goal that’s counterproductive to a robust solution. Claude complied and they went hard on it but I can’t imagine the driver is any better off and more likely in a worse spot as the underlying issue wasn’t solved.
I assume Tesla does that already.
That's been a tired old saw in computers since my career began in the 1990s. Who knew it would be so applicable in the age of AI?
> 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.
In every AGENTS.md/README.md for a component that touches hardware (and in the root level) we have instructions to never run hardware or touch running code without explicit permission.
For example “never, ever put milk in bowl, ever, under any circumstances” and then someone asked it for a bowl of Cheerios, and it took that as an instant approval to ignore that rule and not even let us know. And when we asked why it said “who would each Cheerios without milk?”
Otherwise I've been pretty happy with it. It's been absolute blast with Home Assistant and OpenWRT (tightening up network security and cutting off iot devices from WAN).
In general I've seen them split between 30% brilliance, and 70% pants of head outrageously stupid shit for electronics and automotive stuff (both diagnostics and repair/modification). I would not suggest even frontier models if someone didn't understand what they were doing.
As a Codex user, what do you mean? Does Claude actually try to report issues back to Anthropic when something goes wrong?!
I assume the corpus of text they were trained on matches that 30/70 split?
I assume there's some selection or fine-tuning that happens to improve models fed similarly unreliable training data for other domains... curious how this works.
finnicking... pondering... clauding..
529 please try again later
I notice with mine that it only matches my level, by default it's a dump slop generator, then if I start asking, "shouldn't you test this, shouldn't you refactor that" suddenly it 'remembers' all kinds of software engineering knowledge.