phh

"You can just make it type words, what's the risk in that?"

I don't know if it's a useful answer to people saying this kind of stuff, but here are some examples of other attacks arbitrary USB pwn allows.

A USB device can appear as a network adapter and most OS will happily route all your traffic there, so your speaker can know which porn you're looking at!

It can also appear as a DisplayLink dongle, so it can see what's on the screen (it does require those specific drivers installed, and uh yeah, no way in hell it's technically possible on that MCU).

It can also turn it into a mouse jiggler to prevent lock screen (yes it's technically the same thing as your first point, just HID, but different angle).

It can also appear as a USB-storage: You don't trust the cloud, so you're writing those super secret documents to give to your boss on the USB drive you just plugged in? Surprise, you actually sent it to the attacker.

You can do an IR remote without a RTOS, but as soon as you do BLE you realistically need a RTOS. You have timers for keep-alives, connection states, competing interrupts, CPU-"intensive" tasks that can be preempted (for crypto)

I'm sorry to announce this to you, but the last PowerBook was released 21 years ago

I agree with you that it is the very obvious conclusion, but it isn't obvious to everyone. But it could still be relevant to you if you find yourself discussing with someone saying "why do we even spend money making an API, the AI can just control my computer?"

I'm on the side of "clever, fun, but feels useless". But to defend the project, all sensors require a powered central system. It's pretty common for Zigbee to have one repeater per room [1], which is just what is needed for this system.

[1] Because any AC-powered Zigbee device is a repeater, so just a bulb or a plug is enough

Laying out the math (assuming earth is an homogeneous sphere) just in case it's not clear:

F_gravitational = G m1 m2 /r^2

g = G Mass_earth / r_earth^2

Mass_earth = r_earth^2 * g/G

Density_earth = r_earth^2 * g/G / V_earth

Density_earth = 3*g / (4*Pi*G*r_earth)

Prior to Cavendish we already new g and r_earth, just missing G.

TFA says "After 1,000 fast charges, the battery should retain more than 90 percent of its original state of charge, the company said."

I can't really judge whether 1000 charges is a reasonable target for a car, though i think that 1000 fast charges is reasonable. It should probably be able to push to 5000 slow charges and 500 fast charges, which should fit a lot of use-cases.

Well, you stopped hearing about it, not because it stopped being used, but because its support became mainstream for a very long time. Google strongly recommends using zram on all devices, even on devices with a lot of RAM since like Android 10?

Unironically. Every time I asked a LLM to make something faster, they always tried blind code optimisations, rather than measure.

Absoultely.

I'll add to your comment that it isn't a bug of MCP itself. MCP doesn't specify what the LLM sees. It's a bug of the MCP client.

In my toy chatbot, I implement MCP as pseudo-python for the LLM, dropping typing info, and giving the tool infos as abruptly as possible, just a line - function_name(mandatory arg1 name, mandatory arg2 name): Description

(I don't recommend doing that, it's largely obsolete, my point is simply that you feed the LLM whatever you want, MCP doesn't mandate anything. tbh it doesn't even mandate that it feeds into a LLM, hence the MCP CLIs)

An ESP32-C3 Super Mini can be found for below 3$ (cheapest I had was 1.58€). Since the original clock is 3.88$, it can't be that much cheaper.

Since GCC is lacking such an essential optimization, you should consider have one of your junior interviewee contribute this basic optimization mainline.

You can make frequency inertia with solar (even without batteries if you accept running with a constant reserve so with reduced efficiency). Spain showed that there is a learning curve, that's for sure, but their issue was a "simple" oscillation problem that can be fixed by adjusting frequency-follow rate and grid-disconnect rules. It wasn't like a peak of energy consumption or loss of energy production that only a rotating mass could compensate.

80% is catastrophic though. In a classroom of 30 all honest pupils, 6 will get a 0 mark because the software says its AI?

Is there a survey of SoTA of what can be achieved with CSI sensing you would recommend?

What is available on the low level? Are researchers using SDR, or there are common wifi chips that properly report CSI? Do most people feed in CSI of literally every packet, or is it sampled?

I'm curious why they are using actual modems rather than just doing it with VoWifi that merely requires a SIM card reader (pretty much just an UART)

Except I'm guessing they are not selling their equity, they are making debt backed by their equity?

Well you can't really do -600C sand (or anything), so the benefits of sand VS water largely diminished. "just" freezing water already gives you around 300C equivalent of sand (if my napkin is correct).

Also the point of this plant is to exploit the counter-correlation of cheap electricity and cold. Usually there is a bigger correlation between cheap electricity and heat.

Well in FP4

I dream of getting mcp with interoperable micropayments before ads.

> Tail recursion IME is a bigger foot gun

This is true for some languages, but not all.

Useless anecdote that tail-recursive can be a foot gun even in Scala.

I did a (screening) job interview at Datadog, they asked for "give the spare change back on that amount of money" exercise (simple variant), "in whichever language you want". I did my implementation in tail-recursive Scala (with the annotation). I ended up trying to explain that tail-recursivity doesn't explode in memory for the rest of the call (and failed)

It depends on espeak-ng which is GPLv3

You can technically run it on a 8086 assuming you can get access to a big enough storage.

More reasonably, you should be able to run the 20B at non-stupidly-slow speed with a 64bit CPU, 8GB RAM, 20GB SSD.

Well if nVidia wasn't late, it would be runnable on nVidia project Digits.

- Gesture navigation steals user input when swiping on the left/right edges of the screen,

Well I've seen even 1B+ dl apps failing to handle that (on a Google Pixel), so at this point I'm putting the blame on Google. I've switched back to three button navigation. Though even some trivial OS gestures like screen unlock fail reliably on my Pixel 6a. (As in, I do the gesture, it fails to register the gesture, i try to make the gesture "with more conviction" through the whole screen and it still fails, and after few minutes it ends up okay somehow)

Kyutai's unmute has great latency, but requires a fast small-ish, non-thinking, non-tooled LLM. What I'm currently working on is merging both worlds. Take the small LLM for instant response, which will basically just be able to repeat what you said, to show it understood. And have a big LLM do stuff in the background, and feeding back infos to the small LLM to explain intermediary steps.

So, my brain defaulted to "people are smart, so it makes sense", so it understood it as "it's the toy you keep within reach". But if you look at Pollen Robotics product, you see they have a "Reachy", which can indeed move, and has arms to interact with its environment. So yeah, it's a weird name. It reaches your heart through the feeling it communicates to you with its antennas?

Looking at the BOM (6 cheapest servos, one usb camera, a usb hub, a microctronller ,two mics, 30cm-high low-precision plastic) the price looks fairly realistic to me. I could imagine it at half the price on aliexpress. The manufacturing or sourcing doesn't seem complicated. So overall it looks like a very realistic endeavor.

The only negative point I see: Pollen Robotics doesn't seem used to do mass market/cheap products. But as I said, it seems to be a pretty simple production (I mean, they are probably running everywhere like crazy because nothing is ready and everything is broken, but they should be able to accomplish this)

Could we rather get a link to the original blog post? https://huggingface.co/blog/reachy-mini

As someone who had to tweak many times LMK and Android's way to compute which process/app has which priorities, and how is a service restarted, I want to scream very loudly. But I guess that works.