@Jinx Synesthesia perhaps?

It wasn't meant to look like anything in particular, it's just a sketch solution of how to efficiently generate random points uniformly on a circle.

The points are generated using polar coordinates and it uses the fact that you can take the maximum of two random numbers to get the appropriate distribution for r (which is faster than using SQR RND(1)).

I ran @rheolism's program and got this.
Source: https://bbcmic.ro/?t=b0Xiu #bbcbasic

I ran @rheolism's program and got this.
Source: https://bbcmic.ro/?t=b0dsP #bbcbasic

I ran @rheolism's program and got this.
Source: https://bbcmic.ro/?t=b0Bf2 #bbcbasic

#bbcmicrobot
MO.1:F.C=0TO3:V.28;C*8+7,39,C*8:C.NOTC:C.C:CLS:P.;3-C:N.:V.26,20:F.C=16TO31:GC.C,0:MOVE(C-14)*64,32:PL.97,60,@%:V.31,C*2-28,31:P.;C;:N.:V.5,21
This shows that pattern number AND 15 is actually still interpreted as a logical operation.
16->0 so replaces, 17->1 ORs the pattern bits, 18->2 ANDs, 19->3 EORs and 20->4 simply inverts the pixels already there (ignoring the pattern just like GCOL4,X ignores X).
21 to 31 -> 5 to 15 which have undefined effects.

#bbcmicrobot MO.1:F.P=2TO5:V.23,P:F.I=0TO7:V.I?(&C078+P*8):N.,:S=128:s=4:F.C=0TO79:F.X=0TO1:F.Y=0TO1:GC.C,X+Y+Y:MOVE(C DIV8+X/2)*S,((7A.C)+Y/2)*S:PL.97,60,60:GC.C,3:P.TAB(C DIV8*s,28-(7A.C)*s);C;CHR$(42A.C>0A.(15A.C)=0):N.,,:V.5,21
This redefines each of the 4 patterns using the character data of digit 1 to 4 respectively.
It shows most "extra" patterns depend on the data of the defined pattern which is the set pattern number AND &70

#bbcmicrobot @geoffl
MO.1:S=128:s=4:F.C=0TO79:F.X=0TO1:F.Y=0TO1:GC.C,X+Y+Y:MOVE(C DIV8+X/2)*S,((7A.C)+Y/2)*S:PL.97,60,60:GC.C,3:P.TAB(C DIV8*s,28-(7A.C)*s);C;CHR$(42A.C>0A.(15A.C)=0):N.,,:V.5,21

Only *ed patterns seem documented.
In the left two columns the first GCOL arg is (partly?) interpreted as a logical op, and the second arg has some influence.
I suspect other patterns are accidental due to reading outside of the built-in tables for 4-colour modes.
They seem the same on a master though.

I ran @rheolism's program and got this.
Source: https://bbcmic.ro/?t=aLZ6g #bbcbasic

I ran @rheolism's program and got this.
Source: https://bbcmic.ro/?t=aAQiI #bbcbasic

@markmoxon Very nice!

This has me wondering about stereo 3d teletext Elite - doing it directly with anaglyph 3d seems unfeasible, but the teletext screen is only 1000 bytes so having a whole different screen image per eye is possible.

Bank switching plus active shutter glasses is one approach. Similar tech seems to have been around in the 1980s, but probably wasn't mainstream enough to easily get. Perhaps bank switching with one colour in each bank could achieve anaglyph 3d in teletext...

I ran @rheolism's program and got this.
Source: https://bbcmic.ro/?t=axwGL #bbcbasic

I ran @rheolism's program and got this.
Source: https://bbcmic.ro/?t=apQfZ #bbcbasic