Quick question
I did a small string of ws2812 lights (15 pixels) connected to 3aaa batteries and controlled by a bare attiny85. The first pixel is generally dimmer (or ‘flakier’) than the others.
I seem to recall someone talking about this before…
I have run into similar problems and I figured out that I had too much heat for too long when soldering leads on the first led. I’m not sure how much this applies to your situation. This was the case on 9 out of 15 strips where I didn’t pay attention to the heat and time. And 0 out of 44 (so far) after changing to a smaller tip, lowering the temp and watching a clock to limit the time.
Hmm, it’s happened consistently on many strips so I think that it’s
something else…
I’ve read about significant tolerance variation in 2811/2812 specs. 3xAAA is 4.5V fully charged (assuming non-rechargeable, lower if NiMH!), & only goes down from there as they discharge. the 2812 spec sheet from Worldsemi on Adafruit says both Vcc & Vdd should be between 6.0 to 7.0V (huh - I’d never noticed this before!), although their typical application example shows 5V for both. at any rate, 4.5V is already 10% off 5V, so even with fully charged batteries you’re pushing your luck, and maybe it just happens to be that first 2812’s tolerances that’s making it suffer first (then others as batt V drops). I’d add a 4th AAA at least for the LED string, & tap off 4.5V from the +ve of the 3rd cell to power the ATtiny (or power it all from the 4xAAA but drop it through a couple of diodes first for the ATtiny), then see what happens.
or you might just have a dud 2812 
did you try to add resistor between attiny and led string?
Also i have a questions: how much pixels you can drive with attiny85? And did you faced with some kind of stuck attiny85. I’ve tried to make code with button: blink some leds while button is pressed… and it’s stuck sometimes 