dhon_

This says Claude was able to see the TV over wifi, remove apps and install new ones. It doesn't bode well for the security of these devices

Any policy that asks students not to use AI and trusts that they will comply is giving a massive advantage to dishonest students and punishing honest ones.

This article sounds like the professor was reprimanded by higher powers and tried to dig themselves out of a hole for having a badly structured grading system.

Students should be encouraged to use AI in their learning, because they will anyway. The understanding should be that they will be assessed in supervised environments so they need to learn the work either way. This gives the students incentive to learn without fear of using the best tools available to do so, and provides evidence to the powers that they actually did so.

Yeah my setup is pretty similar, I use a shared repo for my work laptop and home desktop. I test out new setups at home, and typically run nixpkgs-unstable there for a few weeks before migrating the work machine.

I use nixvim for generating my neovim config (and I have a flake output for running that on MacOS where I don't use nix a lot) although I've been pairing back my plugins lately as I find I'm spending less time in the editor.

Claude code/Git/NixOS is such a nice system to work with.

You can build your entire system from a worktree and merge if it works or boot the old generation if it doesn't. This is great for trialing new kernels, NixOS releases, or alternate Window managers, and rolling back is trivial.

One downside is that there is only one stable NixOS at a time, and no LTS. However in practice I haven't found this to be an issue across multiple desktops, servers and SBCs.

As a parent with kids at three different schools, help keeping track of their schedules would be amazing.

Currently updates from the schools are spread over multiple sources: paper notes, email, sms, "news" and "calendar" in school app (which are not kept in sync), and Facebook. Different schools use different apps.

Additionally the channels are often noisy, with messages about groups my kids don't belong to, and "helpful tips" like how to pack healthy lunchboxes. These notifications come in during the day while I'm working, get dismissed due to the low SNR and then the useful ones often slip through unnoticed.

Definitely an area that is ripe for disruption, but as others here have commented, I don't feel confident about Google being the provider.

Is copperhead related to this Copperbrain project?

https://github.com/deadbringer17/Copperbrain

They have similar names, seem to solve similar problems and both have first commit within a few days in July. One is Python and the other Typescript though and they don't seem to be forks.

I worry about the accumulating debt of maintaining all this "free" code if (once?) the subscription honeymoon ends.

Meanwhile I'm shipping tons of new features, upgrading outdated systems and delivering real value. But I do worry.

I'm waiting for the models to start responding with "Oh hi Simon!"

Language still matters because writing robust code in C requires discipline on many fronts, and LLMs will often do the minimum to get code running (as will most humans) unless prompted further. Rust protects you from some of these issues and gives you greater confidence that changes will not introduce bugs.

You can also push deployments to remote systems which is great for servers and headless devices like raspberry pis. I use colmena and deploy-rs on different projects because they have some nice features but you can also do it with bare nix.

I have a repo for kiosk appliances that can build an SD card image for a specific device in a fleet (so no on device configuration is required) but it can also push out updates or changes at runtime. NixOS isn't perfect for embedded devices yet but works well for cases like this in my experience.

What did-ger-eyes-doo?

I have older versions of litellm installed on my system - it appears to be a dependency for aider-chat (at least on NixOS)

Cue a flood of crass jokes as the bots attempt to prove their humanity

In this case though, the new driver has the same license as the project it was based on and explicitly credits the original project

  ISC License
  
  Copyright (c) 2010-2022 Broadcom Corporation
  Copyright (c) brcmfmac-freebsd contributors
  
  Based on the Linux brcmfmac driver.

How do you do it in nix?

Protothreads are amazing, but really expose you to a lot of subtle bugs. I would not recommend them for any new projects if something like async rust or an RTOS are options.

Here's the chat session - you can expand the thought process and see that it tried a few things (hands misaligned with the keyboard for example) before testing the Dvorak keyboard layout idea.

https://chatgpt.com/share/68e5e68e-00c4-8011-b806-c936ac657a...

I also found it interesting that despite me suggesting it might be a password generator or API key, ChatGPT doesn't appear to have given that much consideration.

I was concerned there might be sensitive info leaked in the browserbase video at 0:58 as it shows a string of characters in the browser history:

    nricy.jd t.fxrape oruy,ap. majro

3 groups of 8 characters, space separated followed by 5 for a total of 32 characters. Seemed like text from a password generator or maybe an API key? Maybe accidentally pasted into the URL bar at one point and preserved in browser history?

I asked ChatGPT about it and it revealed

    Not a password or key — it’s a garbled search query typed with the wrong keyboard layout.
    
    If you map the text from Dvorak → QWERTY,
    nricy.jd t.fxrape oruy,ap. majro → “logitech keyboard software macos”.

I switched to a candy-bar style dumb phone for a month and did something similar. My list was pretty much the same as the one in the article with a few small changes.

The most jarring was probably maps - other things like email, messaging etc could be delayed until I could reach a computer but not knowing how to get somewhere right now was problematic and required planning in advance.

I usually kept my smart phone in my car and did a sim swap on the occasion that I really needed it.

LLMs struggle more with embedded software due to the relative lack of examples in the training data compared to javascript etc. They also struggle more with visual reasoning tasks like the character example you provided.

For your first task - give it smaller steps along the way that you can validate. Provide context where possible (like docs for st7789, examples of other zephyr projects). Use Opus instead of Sonnet for tasks that are on the edge of it's capabilities like this. It will still make mistakes, be prepared for iterating on the design and providing feedback.

For your font example, you always need to validate the output of the LLM, but did it still save you time? If so then I consider that a win. If not, give it the task that it's good at (ie generating the surrounding code and definitions) and then fill in the font data yourself.

This is the best explanation I've found of the mechanics of pitch bending https://youtu.be/Fp-GxEaChr0?si=-E9uDTQx51gtnd9C

Essentially, match the size of the resonance chamber in your mouth to the note you're trying to bend to. This is different for every note you bend. You can find the right size by making "hissing" noises while breathing in (without harmonica) and matching the pitch.

Yes you're right. I'd assumed that tokio was used internally in the UI but from a cursory read it doesn't seem to be the case.

Could you elaborate on why using block_on wouldn't be an acceptable solution for users that want a blocking API?

Linus tech tips on YouTube did a video about a windows bug where sleeping while charging would allow the laptop to wake up to check for updates etc but often caused this issue of turning on in a bag

I've seen Gemini Flash 2 mention "in the OCR text" when responding to VQA tasks which makes me question of they have a traditional OCR process mixed in the pipeline.

I wish remote build/deploy for Raspberry Pi was in a better state - it seems like a perfect fit for NixOS.

I've got x86 servers running NixOS that are deployed using Colmena, but it seems to fall apart when I add cross compilation into the mix.

Cars radios are battery powered so will continue to work during power outages, and large enough that adding an AM radio + antenna is not really an issue - unlike a mobile phone. Seems like a good way to ensure most of the population has access to emergency broadcasts.

Edit clarified car radios are battery powered

Humans are complex. It depends greatly on the situation.

Try tracking the table tennis bat

What would be an acceptable error rate for your use case?

There are situations where AI is good enough, other cases where you need more accuracy, and others still where you should be reading the reference directly.

AI is improving quickly though, and context windows will allow for summaries to be tailored to each end user.