One of the best and most complete Hackaday posts I’ve ever read, lol.
Some stuff, it’s okay to let the robot do.
Finding and filling in a slick little BOM display component, for example: people don’t just come up with extra 30+ minutes to do that when they’re spending hours on a post.
We just did this https://news.ycombinator.com/item?id=49478426 this is starting to become the risc v of fpga resumes. It's certainly work and making anything that exists and works even somewhat is commendable but it's getting played out
Author here. One thing that didn't make it into the post: there are no hand-drawn traces on this board. I find mouse-driven PCB layout maddening, so the whole card is a Python script driving KiCad's API—antenna coil, lettering, everything. Retuning the antenna meant changing one number and re-running the script, which as a hardware novice was the difference between finishing this and giving up.
The chip is NXP's NTAG I2C Plus. I picked it for the energy-harvest pin (that's what lights the LED) and only realized later that it ships without NDEF formatting—which is why every consumer NFC app refused to touch it. Sending it raw write commands through Core NFC turned out to be much less scary than I expected: the part that matters is about 60 lines of Swift.
I still can't get over JLCPCB assembling ten of these for about $75.
I can understand because I find MCAD drafting to be a chore and like that I can describe a script to generate geometry now for most things.
Even still, it's difficult for me to fully get it. There is so much art in ECAD layout. There are many dimensions to weigh and high levels of thought and care can result in a higher quality, lower cost product. Many nuances are not properly captured by simplistic automations or chatbots and won't be for some time. Things that accelerate schematic capture, layout capture, library management, and simulation are all useful tools. They're most effective when used with care and thoughtfulness.
> I still can't get over JLCPCB assembling ten of these for about $75.
I remember years and years ago I made some business cards out of 304 stainless with a laser etched design and bottle opener on the side (before that whole idea became mass marketed). It cost me over a hundred bucks for ten, wild that you can get whole pcb assembly done for less now days.
That made the post? Everything in your comment is in the post, AI summaries aren't very popular.
> I designed the whole board as code, which I’d recommend to anyone who, like me, finds dragging traces around with a mouse maddening. KiCad—the free PCB tool—has a Python API, so the entire card is generated by a script.
AI:DR
> Once you’ve felt that, you start seeing NFC taps differently. You’re not reading the tag. You’re powering it.
Your loss.
One of the best and most complete Hackaday posts I’ve ever read, lol.
Some stuff, it’s okay to let the robot do.
Finding and filling in a slick little BOM display component, for example: people don’t just come up with extra 30+ minutes to do that when they’re spending hours on a post.
one of the coolest things I've ever seen
We just did this https://news.ycombinator.com/item?id=49478426 this is starting to become the risc v of fpga resumes. It's certainly work and making anything that exists and works even somewhat is commendable but it's getting played out
website is down
at least https://news.ycombinator.com/item?id=49478426 shared a github.
Author here. One thing that didn't make it into the post: there are no hand-drawn traces on this board. I find mouse-driven PCB layout maddening, so the whole card is a Python script driving KiCad's API—antenna coil, lettering, everything. Retuning the antenna meant changing one number and re-running the script, which as a hardware novice was the difference between finishing this and giving up.
The chip is NXP's NTAG I2C Plus. I picked it for the energy-harvest pin (that's what lights the LED) and only realized later that it ships without NDEF formatting—which is why every consumer NFC app refused to touch it. Sending it raw write commands through Core NFC turned out to be much less scary than I expected: the part that matters is about 60 lines of Swift.
I still can't get over JLCPCB assembling ten of these for about $75.
I can understand because I find MCAD drafting to be a chore and like that I can describe a script to generate geometry now for most things.
Even still, it's difficult for me to fully get it. There is so much art in ECAD layout. There are many dimensions to weigh and high levels of thought and care can result in a higher quality, lower cost product. Many nuances are not properly captured by simplistic automations or chatbots and won't be for some time. Things that accelerate schematic capture, layout capture, library management, and simulation are all useful tools. They're most effective when used with care and thoughtfulness.
> I still can't get over JLCPCB assembling ten of these for about $75.
I remember years and years ago I made some business cards out of 304 stainless with a laser etched design and bottle opener on the side (before that whole idea became mass marketed). It cost me over a hundred bucks for ten, wild that you can get whole pcb assembly done for less now days.
That made the post? Everything in your comment is in the post, AI summaries aren't very popular.
> I designed the whole board as code, which I’d recommend to anyone who, like me, finds dragging traces around with a mouse maddening. KiCad—the free PCB tool—has a Python API, so the entire card is generated by a script.