pentamassiv

LLMs are surprisingly good at reading Morse code. Some time ago I pranked my flatmate. I installed a wifi switch in the wall and then wrote a script to convert text to Morse code and send the code with a slight delay through his lamp. I also built a website to do some legend-building to make him believe a ghost is haunting our apartment. He sent a video of the flickering lamp to chatGPT and it correctly deciphered it. We had a good laugh when I revealed the prank to him.

Sadly it looks like it will be a dead monarch butterfly

https://www.emilydamstra.com/please-enough-dead-butterflies/

Hi, I'm part of the team that built Sai. If you are curious about the tech, feel free to ask.

We are still limiting signups to keep our costs under control, but you can use the code "SIMULAR" to register. The first 1,000 users get access to Sai for $20/month (regular $200/month).

Fefe is a different person

How did the wildlife adapt to that? There must be some cool species there

I find the Swift tooling very lacking. There's no way to lint dead code, there no way to auto format the files exactly as Xcode would do it and tell the linter those rules so that it doesn't lint your auto formatted code. Xcode project files are impossible to edit except with Xcode and Xcode often has issues and I need to manually empty the build folder. These are just some of the issues I remember

Fun read!

Now that some bikes have electronic shifting, you can attack the bike itself. I wrote two blog post about how to downgrade the Shimano Di2 shifters and do a replay attack to remotely shift it. You can find them here:

https://grell.dev/blog/di2_downgrade https://grell.dev/blog/di2_attack

It doesn't look like it does since "fund I" >> "fund II"

I think an important factor is also how long you had the same number. I switch numbers every couple of years because I get a new contract and moving the number to the new company is such a hassle. Over time you share your number with more and more companies and people who sell it or get breached.

My parents had their number for ~30 years. I never get spam calls or texts. They get one once a week or so (this is in Germany, we get a lot fewer of these calls).

In theory you cannot even say for all programs and all inputs if the program will finish the calculation [0]. In practice you often can break it down but the number of combinations of input is what makes it impossible to test everything. Most developers try to keep individual functions as small as possible to understand them easier. You can use math to do formal verification, but that gets difficult with real programs too.

[0] https://en.wikipedia.org/wiki/Halting_problem

CRLite sounds like an elegant solution. Are there reasons why Chrome is not implementing it or do they just have other priorities?

I am not sure the highscores get updated. I don't see HTTP requests or a websocket to communicate with the server

UPDATE: Two people from Anthropic recommended me internally and their HR department already rejected the application. They recommended me for jobs where you need more experience with AI, so I agree that I wasn't a good fit for those positions. Thank you for your recommendations anyways. That was very kind.

A number of other people contacted me with offers so it looks like there will be a happy end to the story :-)

Correct, you cannot remove a version or the whole crate unless very specific criteria are fulfilled. You can "yank" versions. That prevents people from adding the version as a new dependency, but if you relied on it before it got yanked, your build will succeed.

I wouldn't delete old versions even if I could though. My goal is to publish a rock solid library that everyone can depend on and build awesome projects with

Wow, what a great compliment. Thank you :-)

Yes, I applied through the website. Unfortunately I don't know the people good enough to do that

I don't know. I have no comparison but it is common for crates to be released under MIT. I took over the maintainership from the original author so the license was already there. I rewrote pretty much everything so I guess I could try changing the license now but that's not something I wanna think about.

I do the work because I see it as payback for all the great open source software I use all the time.

The role I applied to was not really AI related

I don't think so. They use an outdated version straight from crates.io (at least in the publicly available version of Claude Desktop).

Right now it is just a hobby and there are still a number of bugs remaining. Since I don't have an income from it, I can't dedicate more time to it. Hiring me would allow me to work on fixing them full time and make the progress much faster

Hey, I'm the author of the blog post. Thank you for submitting this. If you have any questions feel free to ask and please let me know how the writing was. It's one of my first posts so I'd like to improve

You're welcome, it's nice to hear from people using it. I hope everything is going well with your new startup :-)

Sure, it would be hard to monetize and while it took countless hours to iron out many of its bugs, it is definitely not rocket science. I contribute to open source software expecting nothing in return because all software I use is also open source. It's my way of giving back and I love the knowledge that it is useful to people and hearing about their projects. So far I did not have any benefits from it but continue doing it anyways. It makes me happy to see more and more people using it.

When you apply, you have to confirm that you did not generate the application with AI. As soon as you send of the application, you get an automatic email confirming your application. They also say they don't reach out, if you are not a good match and that they only contact the people they want to hire. Maybe they changed their mind on that policy, because I received a rejection letter a few months later. It was very well written. The people I showed it to said it is one of the nicest rejection letters they have seen.

I am the maintainer of a library to simulate keyboard and mouse input. I didn't start the project but took over the maintenance and have since rewritten pretty much all of the code. I recently found out that Anthropic is shipping it in Claude Desktop for some unreleased feature which is probably like "Computer Use". I noticed they had an open position in exactly the team responsible for the implementation and applied. A few months later I received a rejection. The letter said that the team doesn't have the time to review any more candidates. The code is under MIT so everything is perfectly fine. It is great that a company like Anthropic is using my code, but it would have been nice to benefit from it. I wrote a slightly longer blog post about the topic here:

https://grell.dev/blog/ai_rejection

I think so too. It helps to form good habits. If you have a roommate for example, you can't leave your dirty dishes everywhere and need to clean it as soon as you are done. You can learn a lot from your roommates too. One of my ex flatmates was super crafty and I got a different perspective on things thanks to that.

I used to live in a flat with one flatmate who changed every half a year or so because they were usually interns. Never knew them before they moved in but 90% of the time we became friends. I liked that they changed after a while so I was never stuck with a bad roomate.

I don't recommend the other way around. If you have a good friend it's more likely to notice their annoying habits so there is little upside but in a worse case you can damage the friendship.

Wouldn't you be able to just set up buckets of water and then dump the water every few days before the mosquito larvae hatch? Is the point of the bacteria that you can set and forget about them?

Fairphone does!

https://www.fairphone.com/en/bootloader-unlocking-code-for-f...

Unless you are using a rooted Android, putting your own certificates on your phone is annoying. They need to be in the system certificate store which is, as far as I know, only possible with a Magisk module.

An easier way is to run an Android virtual device with an older Android version on your computer. You can then use some scripts to add the certificates and proxy the traffic to Burpsuite or mitmproxy. That way you also don't have to switch devices.

It would also be interesting to use APKLab or Jadx to look at the code of the app. Maybe you can find the key derivation algorithm. The app and the washing machine must somehow generate keys or have pre-shared secrets.

If I understand correctly, the app only works if both devices are in the same network? I like that

Only that there is no way back. Those physical magazines would contain the same AI generated content.