I guess I should have guessed what all that empty space on the original Raspberry Pi Pico board was for. They put a WiFi 802.11N chip there. CYW43439 from Infineon. This will be a nice board for IoT. And it's cheap. $6.
https://www.raspberrypi.com/news/raspberry-pi-pico-w-your-6-iot-platform/
Just peeked at the Pico-W datasheet, they are using a RT6154 buck-boost (PDF datasheet), which looks good, but zero stock everywhere.
I’m still just using the old 8266 WeMos d1 mini boards for my sensors, and they’re surprisingly reliable at super low prices. No BLE on those though. I did get some esp32 boards to experiment with over the winter, but renovations and other physical labour projects kept me too busy. I’m hoping the BLE allows for a nicer wifi provisioning experience.
Please don’t get me wrong though - I can sound (very 
) negative when it’s really just my musings out loud. I’m excited to see this new affordable wifi dev board option from the Pi foundation - I expect some new-product issues but I have faith that they’ll ultimately have a sweet little easy to program wifi board at easy to afford pricing.
There's a weird trend to locate components in rows etc, like this:
Not saying that's the cause (sometimes it's worse, this layout might be acceptable) but there isn't any technical reason for laying out the components like this, and that just makes me wonder if the design is suboptimal in other ways too.
I used the same LDO regulator as in the photo above for a project, but laid it out like this, no neat row of components, and a bit of heatsinking at least.
Ah, that might be the reason, I believe some of the PnP machines have multiple heads to pick up more than one part for the same vicinity of the board as you say.
Generally I'm trying to dimension the board and place connectors etc first, at nice millimetre positions, so that I can cut enclosures easier, and then roughly place the components where they are close to others in the same sub-circuit, and attempt to just route in the sub-circuit and then occasionally move some of the sub-circuits around on the board, sometimes change the schematic to make the routing easier (e.g. I almost always change connector pinouts, that's low-hanging fruit to make routing simpler), and then once it is all routed, personally I un-route and re-route in sections, rotating components and fine-positioning them at this point, because I can't get it done well in the first go. So for me, I end up routing things almost twice. I'm no PCB designer either, so I don't know if the experienced people manage to do it right in a more efficient way.
With the ICs like the LDO regulator above, the datasheet is usually followed if it has guidelines or example layouts, or for more complex chips I'll take a look at the reference board for them. The decoupling capacitors I'll try to get one end of them (the positive end) as close to the pin on the IC as possible, and then make the other end go to ground through a couple of vias (the other side of the board is easiest if it's almost entirely ground). If I can't achieve an entire ground on the underside, then I'll aim for only short traces on the underside if possible.
For instance for the PCB in the example above, the underside looks like this (only very short traces where I couldn't manage to route on the top side): (The split is only because this board had an isolated ground side, that's not normally present on most boards):
I rarely make things very compact, so I'm more likely to aim for a board that's possibly larger, and easy to work with and easy to fit in a standard case (I don't have the ability to do CAD enclosure design and 3D print). For some boards, I might not aim to fit into a case, but I'll put holes on the corners so that it can be (say) sandwiched between perspex sheets if needed.
Also, I'll aim for placing extra components, like 0-ohm resistors, to join sub-circuits, in case I need to disconnect a portion while testing. I also won't try to cost-reduce by removing decoupling capacitors etc., since I'm not making anything in high volume (commercial products also don't always cost-reduce, for high-end products they will not try to reduce decoupling capacitors either, or pick the cheapest components, it's not worth the risk to product reliability, at least not unless there is a cost reduction exercise later during the product life-cycle).
Before microcontroller breakout or dev-boards were cheap : ) we would design our own boards each time we wanted to try a new processor. The manufacturer boards were too expensive unless you were 100% sure you were going to use the chip in a design, because the manufacturer board often had exotic in-circuit emulator on-board, and sometimes even fancy IC sockets for the target chip (even for surface-mount ICs - the sockets cost $$$) and the manufacturer dev board could cost thousands (today there are still dev-boards that cost thousands of course, but nowadays often they are for more high-end parts, sometimes lower volume devices).
Here's a dev-board for a Renesas/Hitachi processor, I designed the schematic (and as usual messed up the RS232 pin directions on the MAX202 chip, there are two blue bodge wires there : ) and just made sure to place some components that could be handy for testing or perhaps useful for the actual end product, but mostly just pin header connections. I also put LEDs onto eight or so pins (visible near the bottom of the board), to use as debug. I didn't do the PCB layout, that was done by a professional PCB designer.
