It's been about a year since I bolted together my first printer and produced something that looked like a cube left out in the rain. The box of failed prints is now genuinely large, large enough that I've started feeding the worst of it back to a recycler. What's changed isn't that I stopped failing. It's that the failures got more interesting, and I learned which ones are worth chasing and which are just the machine being a machine.
If I could send one note back to the version of me unboxing that flat-pack, it would be this: almost nothing that goes wrong is a slicer setting. You'll spend weeks fiddling with retraction and temperature because those are the knobs in front of you, when the actual fault is mechanical and an inch lower down. Tightness, squareness, a belt with the tension of a wet shoelace. The software sits on top of the hardware and faithfully reproduces every bit of slop you've left in it.
the first layer, still the first layer
A year in and the first layer is still where most of my prints live or die. The difference now is I stopped treating it as a thing to get right once and started treating it as a thing that drifts. The bed isn't flat, not really. It's flat-ish, and it moves as it heats, and the spring-steel clips I'm levelling against have opinions of their own.
The single best change I made was a sheet of PEI on the bed. Before that I was wiping glass with isopropyl before every print like a surgeon scrubbing in, and getting maybe an eighty percent stick rate on PLA. With PEI and a clean wipe now and again, PLA just sticks, full stop, and lets go on its own when the bed cools. I went from babysitting the first few minutes of every print to walking away.
The boring habit that fixed the rest was writing the numbers down. I keep a plain text file with a block per filament:
hatchbox-pla-grey
nozzle 205
bed 60
first-layer 195 / 65
retraction 1.0mm @ 40
notes sticks fine on PEI, no glue, fan off layer 1
It sounds trivial. It is trivial. But it means a new spool of the same stuff doesn't send me back to a temperature tower from scratch, and when a print suddenly goes wrong I have a known-good baseline to compare against instead of a vague memory of "I think it was about 210".
the failures that taught me the most
The spaghetti-on-the-bed failures were never the educational ones. Those are just adhesion, and adhesion is mostly solved. The failures that taught me something were the ones that looked fine until they weren't.
Layer shifts were the first proper puzzle. A print marching along happily, then suddenly the top half offset by a few millimetres, every layer above the shift carrying the error forward. I chased the slicer for an evening before I accepted it was physical: a grub screw on a pulley that had backed off just enough to let the belt slip under load on fast travel moves. A drop of threadlock and a sensible acceleration limit, and it hasn't happened since. The lesson being that a sudden, repeatable, mechanical-looking fault is almost always mechanical.
Then there's the heartbreak class: the print that fails at ninety percent. Hours in, grams of plastic down, and a corner lifts or the part comes loose and turns the last layers into a bird's nest welded to the nozzle. These taught me to stop being precious about the first layer time. A brim costs you two minutes and a bit of cleanup, and it has saved me more eight-hour prints than I can count. I now brim anything tall, anything with a small footprint, and anything I actually care about.
PETG was the other education, exactly as everyone warned. It's stronger and it doesn't warp like ABS, but it sticks to the nozzle and strings like candyfloss if you run it the way you run PLA. Hotter, slower, more retraction, and crucially a bit more gap on the first layer because PETG bonds to PEI so well it'll take a divot out of the sheet if you print it too close. I learned that one the expensive way, with a glue-stick release layer now standard whenever PETG goes down.
what i'd actually buy again
Not much, is the honest answer. The upgrades that mattered were cheap and dull. The PEI sheet. A better hotend so I could push temperatures without the heat creep that was causing half my mid-print jams. A spool holder that doesn't let the filament bind. And an old webcam pointed at the bed so I can glance at my phone from the sofa and see whether the thing is printing or quietly producing a hairball.
The mods I didn't bother with are the ones the forums get excited about: exotic firmware, fancy probes, the endless calibration cubes. They're not wrong, they're just a long way down the list from "is the frame square and did you write down the temperature". A year of this has mostly taught me that the printer rewards patience and good notes far more than it rewards money, and that the satisfaction of needing a bracket, drawing it, and holding it twenty minutes later has not worn off even slightly.
Next year's project is a second printer, because apparently I've learned nothing.