uSoldering

Why feed it photos of a photo when you could just feed it the photo that was printed out?

It’s hard to imagine anyone wanting this to be real more than I do, but this is nowhere close to being ready to do actual work. Photographing real PCBs is hard, there is no fiducial maths, no actual probing is being done. It's just photos of a photo being piped into an agent. If it actually did what it claims, no reasonable person would exclude it from the demo video.

https://gainsec.com/2026/04/16/autoprober/

For the bottom 99.9% of wealthy people, it is not a good thing.

Eric Schmidt is, in his own words, an arms dealer now and is driving the R&D of autonomous A.I. weapons.

I guess I missed it and was hoping for a deeper dive.

I think it's odd to make a whole video about a DIY rear-projection television without using the phrase 'rear-projection television'. It's novel to bounce it off the ceiling, and it repurposes e-waste so it's a great project.

As someone with over a decade in industrial manufacturing, there is only ever the in-group of the masters, and the out-group of the serfs. Actually making things is merely the means to the end of neo-feudalism.

I am voicing my opinion in which I believe in strongly, nothing more.

https://www.youtube.com/watch?v=TlO2gcs1YvM

You are missing the forest for the trees. This is an aerospace vehicle capable of flying in the air indefinitely. How many drones can you fit on a vehicle? How many vehicles can you fly at once? 60 miles is already reachable by the cheapest and most deadly weapon class of short-range drones. And now all they need to do is fall down. At freefall that is 140 seconds until it reaches the ground.

I like how the article glazes over the fact that this is drone swarm delivery tech. You are naive to think otherwise.

Apparently, the rabbit hole goes deeper. The wayback machine puts the "AI Generated" description on the cheese website at August 7th, 2020. The AI didn't hallucinate anything because it didn't generate anything, the entire premise is simply fake.

https://web.archive.org/web/20200807133049/https://www.wisco...

The (edited) cheese ad: https://www.youtube.com/watch?v=I18TD4GON8g

What probably should be the target link: https://www.theverge.com/news/608188/google-fake-gemini-ai-o...

If this worked well you would imagine a demo consisting of more than a single through-hole component.

What is the overlap of people who are reading a blogpost about Cloudflare standards and people who need a metaphor to understand what compression is? You have 7 paragraphs of highly technical information then just in case, you need to explain how compression works? Just tell your reader you think they're a moron and save yourself the keystrokes.

Driven at 1 amp, ~2 inches from skin, ~5 minutes, twice daily. 660nm seemed to work better. Note that 850nm light is not visible and exceptionally dangerous. You will burn your eyes by looking at anything the invisible light is bouncing off of.

Anecdotally, I designed and built a simple 'LED on a stick' to emit 660nm or 850nm red light at levels that cause eye damage and found it cuts wound healing time roughly in half.

Wait until you hear the news about eggs.

If you photograph and stitch a panorama of a CD at a resolution with the data visible and distribute it under your copyright, is it piracy?

Truly remarkable that technology can disrupt any industry— even one that has been around for millennia, like pimping.

There is a direct parallel from companies using open source code to build monetization platforms while paying open source developers nothing; to companies using user generated content to build LLMs. At least with LLMs it is a self-poisoning well. The internet is dead, long live the internet.

For this PCB, the smallest targets are about 0.2mm in diameter. In terms of precision robotics, this is very manageable. A robotic soldering iron, or even a laser soldering system, with a wire feeder and cutter could be used to make something like a rudimentary die-bonder, that simply solders the wires to the correct pads.

My background in process engineering made me lean towards a figuring out very manual process that could be automated, instead of figuring out a highly automated process.

I know very little other than it appears the two most popular sources are Zillion x Work, and XinZhiZao, both of which appear to be based out of China.

Thank you so much for the bandwidth. I would like to do this for boardscans going forward, but I don't have the hosting infrastructure. I know OSD can do overlays, it would be awesome to have the functionality of OpenBoardView as a webapp.

I didn't take a depopulated panorama because I did all the photography without an automated stage, which is what I'm currently working on. There are some boardscans that are depopulated, with the various layers you can go through here: https://balika011.hu/switch/

I am okay at programming, but slow. I think it's definitely possible, but processing of computer vision is still magic to me.

Yep!

It's just unique spot to hold each part. There are 8 trays with 100 pockets each, so if you wanted to know a specific component's electrical properties, I could look up which tray and pocket it's in and measure it. Or if I get around to measuring all of them, I can push that data into the boardview itself.

At 20kHz it only takes 3 minutes to run the extraction program. I run it multiple times and at slower frequencies, but the output is stable.

No kids and my job is running/programming SMT production lines, so when the process is stable I get to supervise the machines and read technical documents as training.

A bounty hunter-like crowdfunding system would probably be ideal. I could probably hack together some forum software with each thread being a different crowdfunding campaign. Thanks for the suggestion.

Making a machine to automagically remove a tiny bit of material and image the result over and over would be easy for me. The image processing to take the stack of 3D sequential images and automagically process them into a netlist is well beyond my programming capabilities. If anyone thinks they could do this, contact me.

The original goal was to just turn an idea I thought was possible and figure out exactly how to execute it. The current goal is something like improve and iterate, while seeing what the market interest for something like this actually is.

I think most of the value is in the imaging technology, and could easily be offered as a mail-in service. I can also bulk manufacture the extractor PCBs and sell them at a small markup, while open sourcing the rest.