Jeff Geerling

oof, yeah I tried to fix up one of the plushie toys one time... not fun!

YouTube video by Level 2 Jeff This kids toy is engineered better than my dishwasher

My kids' $45 learning table toy is more repairable and will probably last a lot longer than my new $800 dishwasher.

It was easy to take apart, engineered for repairability, and... I fixed the issue it was having with a spritz of deoxit :)

www.youtube.com/watch?v=ACG5...

You keep making it better! Meanwhile I still haven't set up my V2. Each time I've done to do so I've forgotten something. Finally have all the parts I think lol

Ah pulling the ol OpenAI playbook maybe

GitHub - exo-explore/exo: Run your own AI cluster at home with everyday devices 📱💻 🖥️⌚ Run your own AI cluster at home with everyday devices 📱💻 🖥️⌚ - exo-explore/exo

Is Exo Labs going to keep developing Exo (github.com/exo-explore/...)? It seems like there hasn't been any activity for 5 months, and there are a bunch of easy bugs to squash.

Continuing my AI cluster testing on distributed-llama (has a lot of limitations) and llama.cpp RPC.

Heltec V3 is cheap, muzi.works has a nice R1 that I prefer for mobile use.

Note that Docker can run rootless now as well; both are fine, same here in terms of just having muscle memory and tested workflows. In production, it's neither!

Debian 13 "Trixie" is just around the corner — I just published my Ansible test image for Docker/Podman, you can get it here: hub.docker.com/r/geerlinggu...

YouTube video by Jeff Geerling Spying on the Mesh with a Raspberry Pi and GNU Radio

Hacking the Mesh with SDR: I took a Raspberry Pi laptop and HackRF to SFO last week to see if I could decode messages at Open Sauce.

While it wasn't complete success, I learned a lot about GNU Radio, Meshtastic, and even tracking airplanes with ADS-B! www.youtube.com/watch?v=1_lb...

No it would need a custom kernel compile, plus any patches for the AMD GPU that are required

Ha!

Seismic waves move fast!

Yep! www.jeffgeerling.com/blog/2022/wa...

Ha, as I was typing that, my clothes dryer switched off, and yes, looks like it's a motor hum/vibration from that. Our AC motor is currently running still, which I'm guessing is the other dimmer line. The dryer is directly above the shake location, while the HVAC is about 20' away

That could be my clothes dryer, which is electric and runs for an hour or so at a time. We had a lot of laundry today so that's my best guess.

Wow... did my Raspberry Shake catch the massive 8.8 #earthquake all the way in St. Louis, like 8,000 km away from the epicenter?

RP2350 A4, RP2354, and a new Hacking Challenge - Raspberry Pi New A4 stepping of our RP2350 microcontroller now available, with security and other improvements. Plus: another RP2350 Hacking Challenge!

New RP2350 stepping just dropped, E9 fixed.

www.raspberrypi.com/news/rp2350-...

There's a trade off with any process of course. Photonic runs a gauntlet of tests, so shortening the individual test time is probably part of that decision.

Those tapes are holding data, just not much of it :)

LTO is pretty pricey, some archivists use it, for sure. A lot of places store their offline backups on tape, and rotate tapes out to an offsite secure location.

I just use Amazon Glacier Deep Archive since it's more convenient

Do you mean like:

Didn't seem to be any on this analog board! Nor mentioned in the guide I found on AppleFritter or included in Console5's kit.

The leaks on the analog board were easy to clean. The caps on the motherboard's were a little harder to see before I cleaned the board. Then I noticed a couple iffy traces.

Hey look, I found the save icon! 💾

Indeed! Learned a lot more about this computer in the process of bringing it back to life. Also about 8 of the caps started leaking, but luckily nothing catastrophic.

I thought it might be an inside joke or easter egg... but more likely it's abbreviating "SINGAPORE" www.sgpbusiness.com/company/Appl...

My Mac Classic's motherboard has an analog board called 'Apple Computer S'Pore' - never heard a reference to that before...

Definitely a fun rabbit hole; though I wouldn't agree with some of the commenters that GPIO access is inherently insecure (or secure, or anything really... it's quite neutral as far as interfaces go—it's a kernel level thing).

My solution is to always write through a multiple of the available RAM (so like 32 GB if you have 8GB of RAM, or 5 TB if you have 1 TB of RAM) so even if the first portion is hitting cache, you'll bust it at some point.

Yeah, the larger numbers didn't make sense for the UHS-I / SDR104 interface on the Pi, but do make sense for IO being cached somewhere.

Benchmarking ZFS is a royal pain because you might be benchmarking a combination of RAM cache, NVMe storage, PCIe bandwidth, SATA, etc. in one volume. 1/2

Ten Thousand

Ah.

And lol, I guess I'm one of the 10,000 today! (xkcd.com/1053/).