easygenes

The Hailo-8L hat for the Pi is only ~$80 and has more than 3x the compute power of the Coral USB.

* https://www.pishop.us/product/raspberry-pi-ai-hat-13-tops/

Even the bigger Hailo-8 hat with >6x the compute of the Coral is only $135.

* https://www.pishop.us/product/raspberry-pi-ai-hat-26-tops/

I'd also keep an eye on the Rubik Pi 3 as it looks specifically designed to compete with the Pi 5 + AI kit and should provide a faster, cheaper, more efficient option. They're only just starting to ship and no support in Frigate yet, so just something to consider if you're not in a hurry to build a system.

* https://www.rubikpi.ai/

* https://liliputing.com/rubik-pi-is-a-compact-dev-board-with-...

* https://www.cnx-software.com/2025/01/09/qualcomm-qcs6490-rub...

I’m not sure what’s the most optimal for cost/performance, but a Pi 5 8G + Hailo 8 looks like it will be a good option.

* https://www.reddit.com/r/frigate_nvr/s/ncxP1YQDfB

* https://github.com/blakeblackshear/frigate/blob/e773d63c16d9...

Very cool! Nice convenient OKLCH palette.

Is it just me or do purples 5 through 8 not seem as differentiated as all the other colors across the palette?

Was this an intentional choice?

Note: I have no color sensitivity deficiencies and a reasonable understanding of color theory.

Yes, but this employs additional encoders to implement a more sophisticated cascading control loop: https://github.com/adamb314/ServoProject/blob/main/Doc/Theor...

You'd be surprised what you can do with servos modified with encoders, setup with an industrial grade cascading control loop. First demo video below demonstrates threading a mechanical pencil lead in and out through the writing tip. Second video demonstrates a 187gram weight at the end of a 470mm long rod. Third video shows that the modification is now a quick fitting of some 3D printed parts to the servo motor axle and servo housing.

  https://www.youtube.com/watch?v=ECLrLupFW10
  https://www.youtube.com/watch?v=_4mrb2T706s
  https://www.youtube.com/watch?v=Ctb4s6fqnqo
  https://github.com/adamb314/ServoProject

Cool! Is there a straightforward way to import your Pinboard?

This is neat, but what I really want to see is someone running it on 8x 3090/4090/5090 and what is the most practical configuration for that.

Yeah, ME/CFS in the extreme.

Mostly it just means that we’ve got a new tool we can leverage which is a positive feedback for improving models, and will ultimately increase both train and inference demand and shorten timescales for deeper capabilities.

OrcaSlicer is a fork of Bambu's slicer. It defeats the whole impetus of the project to not support Bambu Lab printers.

Look up Dubai Philips LEDs. The problem with most consumer LEDs is that they are overdriven, so their life is short. In Dubai, the Sheikh basically mandated that the bulbs need to be underdriven, so the lifetime is about 25x that of incandescents. They only deliver about 1/4 the power to the individual “filament” that most similar Philips lamps would at the same light output.

https://hackaday.com/2021/01/17/leds-from-dubai-the-royal-li...

It’s possible to install warm colored LEDs with very little blue light output though. You get all those benefits without giving up the more-suitable-for-night sodium light spectral benefits.

Sidebery is my go to for that.

https://github.com/mbnuqw/sidebery

Jobs had two iPhone teams working in secret against each other, and was setting up things on the side. He likely approached Otellini before either team was far along.

Yeah, Otellini famously turned down Steve Jobs' request to make the chip for the first iPhone, thinking the market wouldn't be big enough. When he got pie in his face, he tried to correct course. By the time he needed to retire, the board wanted to give up on mobile, thinking they would never catch up, and double down on data center.

Mostly I preferred the aesthetic, the menu and sessions system, and how I could configure it easily to handle my tab containers strategy.

I converted from TST to Sidebery years ago and haven't looked back, but generally agree with the sentiment.

Can you list the varieties that you know of to be problems? That at least gives a clue to which are less problematic.

Is the last line said ironically? Japanese brands long have and continue to absolutely dominate long term reliability ratings for vehicles, and the first mass market hybrid and full EV vehicles came from Japan.

If you’re talking about innovation and mass EV manufacturing, sure the US and China are leading, but the European Volkswagen and BMW Groups are still competitive. Japan is admittedly a laggard in the EV market, but largely because EVs are still a luxury good and Japanese brands are primarily mainstream.

It is also a training issue. The model has to be trained to reinforce longer outputs, which has a quadratic train-time cost and requires suitable long-context response training data.

Cool project!

Just FYI your chosen name collides with Nous Research, which has been a prominent player in open weights AI the past year.

I had a 5950x with similar issues to other posters here. It failed outright after just over two years.

Funny, just came across this 90 minute interview with the lead author from a few weeks ago on this topic yesterday: https://youtu.be/I0yil_3dvjs

My preference for this on the web is Wolfram Alpha: https://www.wolframalpha.com/input?i=57+tonnes+%2F+%286.1+g%...

(note I've slightly changed the prompt to match the "drums" OP used instead of barrels. This avoids the issue other child mentions of ambiguous units, as it clearly states the volume of drum it's using and gives options for others (e.g. the international 200L instead of the US 208L/55 GAL).

Another option which I wouldn't inherently trust to get the calculation just right, but is good for explaining the process if you can fact-check it is ChatGPT-4o: https://chatgpt.com/share/ead07fff-8d55-4147-a742-6e19b26f1e...

Generally I think this would be a better test if the phrase were “dropped off” Sue at university. As phrased, it’s plausible to assume that Joe stayed at university with Sue, and drove her back and had tea after.

Though a truly smart model should seek to disambiguate situations like this.

There's wide speculation that what will be branded as either GPT-4.5 or GPT-5 has finished pretraining now and is undergoing internal testing for a fairly near-term release.

Github doesn't pitch it as well as the videos he's uploaded on Youtube. They add dual encoders to cheap commodity hobby servos to allow for excellent backlash compensation. Result is movement accurate and consistent enough to thread a mechanical pencil lead in and out of the tip of a pencil. Project has been active for 5 years, with significant improvements in the last year.

https://www.youtube.com/watch?v=SioCwvR_PYY

https://www.youtube.com/watch?v=_4mrb2T706s

https://www.youtube.com/watch?v=Ctb4s6fqnqo

Anyone know how the accuracy of this compares to the similar-cost adamb314/ServoProject arm? [1] It utilizes a servo mod adding dual encoders to compensate for backlash and achieves accuracy of +/- 0.05mm (enough to thread a mechanical pencil lead in and out of the tip of a pencil). [2] He's been working on the project for 5 years, with significant improvements still in the last year. [3][4]

  [1]: https://github.com/adamb314/ServoProject
  [2]: https://www.youtube.com/watch?v=SioCwvR_PYY
  [3]: https://www.youtube.com/watch?v=_4mrb2T706s
  [4]: https://www.youtube.com/watch?v=Ctb4s6fqnqo

To see literally what's inside electronics, the book "Open Circuits" is enlightening. Many pictures of electronics carefully sliced open with explanations of the internals: https://www.opencircuitsbook.com/