lawik

My in-laws Kia did this for me. It got really shitty when it got darker and presumably had to use an IR camera. And I am tall so the angle might have been bad. It flagged me every minute. Even when I intentionally focused right ahead.

Tracking gaze is not immune to assorted failure modes.

The cadence these topics were written in was so Apple keynote video that I had Tim Cook's voice presenting it in my head. I hope that's not just me.

More in the topic. Good that Windows Update will suck less. Did the Discover-something-or-other-imply less start-memu ads, I couldn't tell..

I believe the OSS part of this oban library covers a lot of the celery feature set and is free as well. You can run multiple nodes and knock yourself out without needing more infra than the existing postgres.

Oban has been a free and OSS project in Elixir for ages. There are some more advanced bits that are paid and that sustains the people that make it.

If you like Celery. Great. This is a different take. I didn't enjoy celery last I needed it but never got super familiar.

Not necessarily easy to do performantly, scalably and well. Oban has been doing this in Elixir at real scale for a decent while now.

Why do you need a tool? Well. You say celery. I am inclined to say Oban. And I'd say both are tools.

I wonder if the limitation of the application processor and Linux starting is mostly a default for the standard OS or an actual limitation. Typically with a hybrid SoC like this part of the point is that you can use the micro-controller as the power efficient thing that decides when the bigger application processor should boot or not. I'd be curious to see if that's possible with this one.

Shipping only Debian to start is fine by me. It has to start somewhere. And they seem quite responsive to making it work with other things. James Harton is plugging away at getting it working with Nerves (https://nerves-project.org) and he has it running with Buildroot already. Current repo: https://github.com/jimsynz/buildroot

Most recently they pushed their special sauce for the bootloader and how to produce the relevant mystery binaries. https://forum.arduino.cc/t/buildroot-support-for-uno-q/14108...

I share the sentiment that I don't trust that there won't be issues with Qualcomm over time. That company does have some pretty relevant chips though so I'm hopeful this means that we see them become more accessible on SBCs and embedded boards. I feel like they've been popping up more and more.

If they value this investment in Arduino they should now have a small wing of the company that pushes for things to be more open and even if they only consider that a marketing vector, if things are opened up for that purpose, quite possibly a win. But Arduino might also be absorbed into the amorphous megablob and this is the last we see. I hope not.

I don't think this board is that weird. It is just coming from the Arduino side and moving into Raspberry Pi territory. Personally I want to run Nerves on the application processor and get some practice with Zephyr on the MCU. Seems to already be supported: https://docs.zephyrproject.org/latest/boards/arduino/uno_q/d...

Also why no mention of the LED matrix. This is something RPi devices fail at. Providing som default way of neat output. First time plugging this in it starts doing fun stuff.

It seems to be RJ11 and they might be spec:ed slightly differently (thicker leads?) to power motors. I get the impression it is mainly to control what can be used in competitions.

This was what I found from skimming around: https://www.robotevents.com/V5RC/2018-2019/QA/35

I doubt it is Ethernet at all, so it wouldn't be Power-over-Ethernet. Just some useful connectors and wires making for an appropriate cable. Also seems like you can make your own perfectly fine. Or they might melt. I suppose try it.

Funny!

Now that Kurt doesn't have commit access, who do I ask to get internal Fly Slack bot fizz off of my behind.

I was in a devrel channel for a short while and ever since it has asked me to write updates in a channel I don't have access to. Frequently.

I think Intel wants to stay alive so they are looking for a lifeline and the current administration wants to bring chip production into the US.

Intel's death would be very embarassing to that whole effort. So Intel has incentives (survival) and the administration has incentives (jobs in the US). The method is "whatever can be claimed as a win".

No US party seems particularly capable or keen to hold an ideological line but especially not the GOP from what I've seen. Not saying other countries have particularly impressive parties either. I'm less than thrilled with ours over here.

I've done this. The NIF worked as in that it ran and was a correct enough NIF. It did not work in terms of solving what I needed it to do. Iteration was a bit painful because it was tangled with a nasty library that needed to be cross-compiled. So when I made a change it seg faulted and I bailed.

I essentially ran out of patience and tried another approach. It involved an LLM running C code so I could check the library output compared to my implementation to make sure it was byte-for-byte.

The C will never ship. I don't have practice writing C so I am very inefficient at it. I read it okay. LLMs are pretty decent help for this type of scrap code.

I wonder if they'll be satisfied there or add a chunk of others now that they've started. Parakeet is supposed to be good?

Should they add Voice Activity Detection? Are these separate filters or just making the whisper filter more fancy?

I will consider it :)

The advice in every consultation is either mouth-spoken or finger-written by a 100% organic (but not currently certified) entity.

I really should see how many 1 million process BEAM VMs can fit...

I found out that Ampere is into edge and telco usage way after we got connected to do this work actually. I've been an Elixir dev and through that connected to Erlang for 7-ish years.

But I will certainly try to leverage my telco-connection to get to play with more of their kit if I can.

I was aware an Erlang being a unit though I'd forgotten what it measured. I Need to have my fun when giving titles to these things. Hope it fell within bearable tolerances.

According to Robert Virding at an unnamed bar in Berlin ~3 years ago they just wanted to be like Pascal in terms of picking a mathematician. But Ericsson Language certainly helped sell it internally, I'm sure.

True, worth tuning

This probably got second-chanced or something because I posted it over a week ago. Funny that this went up here since the follow-up is probably more in-depth.

https://underjord.io/booting-5000-erlangs-on-ampere-one.html

Hope people enjoy it, both the technically interesting bits and the general goofiness of it all. I personally have enjoyed making the number go up.

I hope my co-fouunder on NervesCloud does some write-up on the allocator variations he has been trying. Every day he pops in and asks me to throw 5000 devices at the servers and see what they do. mimalloc, jemalloc and some others. Those two seem to do a better job with fragmentation for the BEAM than whatever we had by default, presumably from glibc.

No. It is not necessary. I don't think I said it was.

You can absolutely simulate more IoT-devices. But in this case I don't have to simulate receiving OTA updates, rebooting, keeping state. It has some upside but it isn't the reason I've done it.

I think I mentioned that in both this and the previous piece that I started out doing the VMs for fun and that's still kind of the central point. "What can I do with 192 cores?" We'll have a good use of the artifacts of this work for testing the framework itself in ways where devices are useful.

As you say, if I wanted to stress-test the OTA service for real I could definitely just set up some libraries to make multiple connections instead of starting the whole machinery. You can make a lot more connections, throw a lot more work at the system. And depending on how fancy you want to get you can simulate a lot of device behavior.

Given that I happened to have a couple of thousand full-fidelity devices in my pocket we got some incidental testing of the service.

Well I applaud their ability to get that bag. But in terms of raising the floor for a lot of people, this is not that.

I don't feel strongly about these salaries beyond them being an indication of deep dysfunction in the system. This is not healthy, for a market or for a society. No-one should be paid these amounts but I don't care about these developers because they don't run the system.

I've benefited from devs being paid well. Not that well. But same thing in concept.

I love following their adventures. Glad to see that it continues.

This reminds me of when a student pub hosted a wedding party. As we got to the dance floor there were drinks on offer. I grabbed a Cuba Libre and essentially went "oh, that's the roughest on I've had, must be cheap cola on truly awful booze". Kept on sipping. Someone else complained. Then another one. We go back to investigate. Bartender has a taste. Grimacing.

Like bad and maybe salty?

Takes a minute before they track it down. They'd mixed up the salted ice cubes they used to keep champagne cold with the ones they use for drinks.

That one was rough. I had politely worked through my drink before we launched the investigation. Good party though and I did get a new drink which tasted fine.

I'm sure a tiny pinch can do wonders. Salted cubes, not recommended.

pinout.xyz is a treasure when working with Pis in general.

Right. This was incredibly useful.

But animated gradient outlines on text is the only use-case I care about.

I think this proving out the concept. A dev board costing. 150 doesn't matter for professional projects. It latters for tinkerers. What matters is unit price for desired qty.

And this has 4G/LTE (because it is a smartphone) so comparisons to base RPis are largely irrelevant.

And in industrial embedded Linux stuff there is essentially no correlation between price and performance. Most don't need performance and they aren't really cost-optimizing this bit of the production line very hard. It just needs to be certifiable, reliable and replacable.

I do hope they come down a lot in price and prove this out over many more phone variants.

A lot of the glory of Elixir and Erlang are in that you can write code that is very straightforward but performs well.

There may be cool cases for extra GenServers or gen_statems but a lot of uses of Elixir and Phoenix don't warrant making any fancy architectural choices. Partly because the runtime is already fancy for you.

No benchmarks. Hard to take this seriously.

AI voice will detract because voice is very important in video. Especially in essays. And AI voices are very common for low effort slop video so it would not be novel.

But it is not necessarily a dealbreaker. Good research, good writing, funny timing, great editing. All of these can elevate the whole. Or being exceptional at one or two can compensate for missing one.

CGP Gray does well without a face and started with terrible animation.

All names are unfortunate. There are too many things.