I've been thinking about all the projects I have in progress. It's a long list. The good news is some of them are connected.
For instance: I have solar panels out in a field, and their controller is a box you just plug into the house. But plugged in at the house there's a lot of noise in the circuit — the micro-inverters on the panels are 150 to 200 feet away, and their signal arrives too buried for the controller to hear. It worked better connected directly at the fuse box, but that's outside, and the old controller got flooded out there. Bad design, poor installation, not my fault. So at the solar array I'm installing a fully waterproof box that connects directly to the hot lines — 220 volts, which I'll pull off as 110. I think there are four wires, so I can use ground, neutral, and one of the hots. Gosh, I hope there are four. Got to check.
But now there's power and a plug at the solar panels, which means I can install a TP-Link powerline adapter and get Ethernet out there without pulling cable. Which means I can install an outdoor Wi-Fi access point. Which means my rover project — a little kid's electric toy Jeep, the kind they sit in — gets Wi-Fi in the field. Not for telemetry; for communication. And while I'm out there I could drop a GPS antenna on the array with an ESP32 doing some work, and that plus the rover's own GPS gets me fairly precise positioning. Give it LIDAR, give it cameras, and this solar-powered child's Jeep patrols the open acre doing silly things. If it sees a chicken it freezes; if it sees a dog it chases it. I don't know. Projects don't always have to have a specific purpose. I build things to see if I can.
That's partly why I'm writing this. I realized something recently with Jerry's keyboard project, and with the Tesla numbers I promised years ago and finally delivered: future me reads these things and goes, oh yeah — I'd better finish that. So you get to hear about projects in flight, some of which have been in flight a very long time.
The rover has been in flight for years. Did I actually buy the electric Jeep two years ago? Yes. Have I torn it apart? Yes. Have I put a radio controller in it and driven it around manually? Yes. Have I tested the tank-drive concept? No. Do I have a Raspberry Pi ready to go into it? Yes. So it's getting close. Right now it's four really cheap motors on plastic gears driving each wheel independently — four-wheel drive, to be driven tank-style. But there are no encoders. So I went to my makerspace membership — laser cutters, mills, everything — drilled a ring of holes in a gear, dropped magnets in, and had a little circuit board made with Hall-effect sensors. If I hold everything in place by hand and spin the gear, it reads the rotation, forward or backward, precisely. The code works. The sensor works. Where I'm stuck: the board has to sit about a millimeter and a half from the spinning gear, so I need something that holds it there precisely. So I'm going back to the makerspace to get certified on the 3D printers, and we'll go from there. That project is long and has a lot of pieces, but I've carved it up and I know each chunk: get the encoders working, reassemble the gearboxes, test, validate, move to the next piece.
That's how I think about priority too. Wi-Fi and GPS at the solar panels matters less than getting the encoders working. But moving the solar controller out to the panels — closer to the data on the wire — runs in parallel, probably higher priority than the rover. And then on any given Tuesday I come up with something like "hey, I should write a blog article about the projects I have in flight," and that becomes a micro-project of its own.
Any others of interest? I want to build a general PCB I can use for all my projects — an ESP32, a precision clock, a precision microphone, a temperature-probe connector, an SD card, some LEDs. Any dev board has most of these to a degree, and I could buy one. That's not the idea. The idea is: can I design this board? Yes, with AI's help. And yes, I've found AI screws up a lot.
Everyone's talking about AI. I have one, I use it, and it's fantastic — it's a Life OS. But I've found I need to build a significant set of skills into the harness, basically a sub-harness that understands electronics. I don't understand electronics well enough to build that myself, perhaps. I'm trying anyway. What I like about it is that for me it's a never-ending game. Here's how the first AI-designed board came back: the real-time clock worked, we got the microphone working, the temperature probe failed, the high-precision ADC failed. The battery charging circuit worked — though if you plug and unplug USB with the batteries in a certain state, it fails. And the reset and boot buttons: I wired one of them straight to ground. A hard short, and the voltage regulator fries itself. I should have double- and triple-checked that one, but AI gives you so much confidence, right? Probably $300 of boards. Hard lessons. They're not wasted — they're perfectly capable ESP32s with all their GPIOs pulled out, and they run my bench now; all the encoder testing runs on them. But gosh darn it, it would have been great if they'd worked out of the gate. I understand the mistakes. I'm comfortable I could rewire a board manually, fix the schematic, and it would be fine. But I want my AI harness to do that for me. There's commercial software for building PCBs and it might be better than anything I'll ever build — but I don't want to pay for it, and I want to learn. If you don't bump up against walls and crash, you don't learn.
Then there's the project born of a deep desire to never again buy something on Amazon that I know I already own but can't find. It's sad, I know. We had the perfect kitchen spatula — red, silicone, flips eggs, slides around on things, just great — and then one random day it's: where's the spatula? Have you seen the spatula? Swipe Amazon, past orders, spatula, click. It shows up the next day, and five minutes after you open the package, you look down and there's the spatula you thought you'd lost. Now, fair warning: the spatula is a bad analogy — if you've truly lost something, no inventory system is going to find it. But the camping stove that's been in the garage for a year, and it's summer, and you're going camping now? That it can solve. That's ThingDB2. There's a ThingDB on GitHub — buyer beware, I'm radically changing it into a separate project called ThingDB2 — and the point is getting a better handle on my stuff. Type "camping gear" into the search and semantic search — not AI per se — connects it to the stove, the sleeping bag, the tent, the cooking utensils, the three boxes on the third shelf in the garage. It might even tell you the stove isn't in the house at all — it's in the storage locker, back wall. Grab it on your way out of town. It runs on a Raspberry Pi in your closet, on your home Wi-Fi, and the new piece is an iPhone app that talks to the Pi — a richer interface, decoupled, while the Pi just sits there. I love QR codes, so it has those too; they're not necessary, I just love them. That's a lot of words for one project, but it's a premier one. Top of mind.
Home automation is in the pile too. ESPHome is super cool, Home Assistant is super, super cool. I used to run another system on a Mac, but keeping a Mac dedicated to that isn't sustainable, so an old x86 runs Home Assistant just fine now — and I have a strong affinity for the ESP32 family, so that's the direction. Which brings me to the Mighty Mule gate on the driveway. It needs telemetry, man. There's some Wi-Fi board you can buy for it, but it's easier to tell an ESP32 with ESPHome: here's 5 volts off the 12-volt battery, and when this command comes in, toggle the gate. And see, that's the thing — toggle means change state, closed to open or open to closed, which is only useful if you know the current state. There are relay inputs you can easily use, but nothing immediately obvious that says the gate has completed its opening cycle. When I search, nobody has solved it with the actual controller — they bolt on reed switches, some of them literally point cameras at the gate. There's got to be a better way. So: project number 9,226 — reverse-engineer that board and find the signal that says the gate is now open. It's got to be there, man.
When my AI reads this and I say post it, it's going to tell me I forgot about this project and that one — there are easily fifty. Every website is a project. Some are live projects; many, many others are someday-maybes. If you'd read the raw text I dictated, you'd think I'm crazy.
That's okay. I'm allowed to be.
Because here's what I'm actually saying: none of these projects are late, because none of them have deadlines. They're not stalled — they're in flight. Some have been at altitude for years, waiting on a connecting piece I haven't built yet. The waterproof box feeds the powerline adapter that feeds the access point that feeds the rover that needs the encoders that need a 3D printer certification I don't have yet. It's all one machine, and I'm a part in it.
Thirteen years ago I built a keyboard helper for Jerry, a kid with spinal muscular atrophy who ran his whole computer with a mouse and not much else — saying he couldn't use a keyboard doesn't begin to cover it. I wrote it up and forgot about it. Last week I read my own article and put the project back on the front burner. The helper is still working, thirteen years and zero iterations later, and Jerry games on Twitch like a pro with a mouse and a face camera. That's why I write these things down. Not for you — sorry — for me. Future me is the only reader I can count on, and someday he'll skim this, laugh at the spatula, and go finish the rover.
If you've got your own list of projects in flight, write it down somewhere public enough to embarrass you. It works. The Tesla numbers took me fourteen years to deliver, and I still got there.
See you at the makerspace.