Ah that feeling when you find out the 2N3347 is not a PNP dual transistor. It’s an NPN one. The one data sheet on the internet is wrong. And of course that will be why my circuit is not working
Edit: it now works. The findings so far:
1. The 2N3347 is a dual NPN transistor not a dual PNP transistor.
2. The emitter pins are NOT at the same end as the tab unlike every other TO18/TO5 package out there.
Now those problems are out the way and the circuit turned upside down basically, it works nicely.
Performance data has been tested.
settability = +/- 0.2uA
+/- 1V supply swing = +/- 1uA (1%)
0-4.7K load swing = +/-0uA (not measurable with the Fluke 25 in uA anyway)
This is NOT temperature stable. The circuit is CTAT. To compensate we'd need to sum with a PTAT source at the junction which I can't be arsed with. I'm happy with 5% or so here.