What are your best local models, and what hardware do you run them on?

It says 12-15 hours of recording in the article.

I really don’t want to wear a battery in that form factor.

Sure a phone or watch can burst into flames, but at least you’ve got a chance of dropping it or taking it off.

I also don’t see the bother of talking to your wrist rather than your hand.

Does Starlink have a temporary or “pay as you go” option?

Not anymore. You need a $5/mo charge to keep your account hot.

They still give you 500kbps of speed, which is enough for checking emails, voice calls, navigation, music streaming, etc.

That's pretty reasonable.

Historically Starlink roam let you pause/suspend the service and restart it when you need-

In August they changed their plans so you’d need to cancel and re-subscribe.

I didn’t know they’d added an alternative kernel. The CGAL one used arbitrary precision which massively slowed it down.

Also, fillets are made using the Minkowski operation, which is super slow.

They added it in the dev branch. There hasn't been a stable release since 2021 and there has been a lot of ongoing development in the meantime, many people use the development release since it's significantly faster.

Where, out of interest?

If that’s Octopus Intelligent Go, then it will also give you the 7p rate outside the normal nighttime slot if the car is charging and their algo calculates they can do it.

The one you linked looks ugly. Horrible seams, odd and too-long red buttons, unaligned slots.

On the contrary, I think you’ve done a fantastic job with the enclosure. Far from looking “half melted”, you’ve made good use of the bed texture, and the design is top notch.

Always impressed with the PIO capabilities.

Reminds me of the old BeagleBone Black that had two small separate cores that worked in a similar way. Someone used it to create a 3D printer control system.

Are there any other chips out there like this?

Application / realtime split cores are very common, STM32MP* was designed around the application processor / realtime processor split, I don’t remember if that’s what was on that Beaglebone. Any “big” application processor these days will have a variety of smaller generic cores (Cortex-M style) around it, some which usually handle programmable I/O.

A lot of microcontrollers also have pretty sophisticated interrupt controllers and timing analyzers which can be used to accomplish similar tasks, although they’re usually “programmed” by chaining register effects so it’s nowhere near as elegant as PIO.

Specialized IO coprocessors which are programmed using “code” like PIO is are a little less common, Infineon Peripheral Control Processor springs to mind.

I just refreshed my memory - they were called Programmable Realtime Units on this chip: https://uk.rs-online.com/web/p/microprocessors/1219716

Can the “little” cores in big.little arches be run entirely independently then? That’s pretty cool if so.

> Can the “little” cores in big.little arches be run entirely independently then?

Well, that too :) What I’m referring to is more like Qualcomm “safety island” on Dragonwing, Xilinx RPU, or Allwinner AR100 (I think this is used in 3D printer projects using A64, actually), though - where most modern large “embedded” Linux SoCs have some real time island to talk to the outside world. Cell phone SoCs and stuff like Apple M also have realtime cores hiding in them running blobs, although they’re usually connected to more specific RF or A/V blocks rather than generic IO.

> Specialized IO coprocessors which are programmed using “code” like PIO

Which is exactly what GP was talking about: the TI PRU cores are this.

Yeah, I didn’t realize those were in a Beaglebone ever, I thought it was STM32MPx. So that’s like PIO or PCP more directly.