Compare lens runs

Any two frozen runs — one model on two prompts, two checkpoints of one model, or two entirely different architectures on the same question.

ASteve Agentopen run

The capital of France is

2026-08-18 13:58 · MonumentalSystems/harmonic-gpt-128m-byte-agent-multiparty-lens@74c7dc09a1f9

positions: 26 (+2 gen)
sites: 2
lens agreement: —

completes: t

Where predictions settle — depth of commitment

Block 16 post-residual20

Block 17 post-residual6

BAlpha 60M Chatopen run

Once upon a time in Paris

2026-08-18 12:37 · ajaxdavis/alpha-chat-jlens@0b9850e4d06c

positions: 54 (+48 gen)
sites: 16
lens agreement: 52%

completes: , the Frenchman, who had been in the French army, had been in the army. He was in the army, and he was in the army. He was in the army, and he was in the army. He was in

Where predictions settle — depth of commitment

Decoder block 1 post-residual7

Decoder block 6 post-residual1

Decoder block 12 post-residual4

Decoder block 13 post-residual8

Decoder block 14 post-residual5

Decoder block 15 post-residual12

Decoder block 16 post-residual17

Final readings, position by position

#	A token	A reads	B token	B reads
0	T	h	Once	the
1	h	e	upon	the
2	e		a	time
3		s	time	,
4	c	o	in	the
5	a	l	Paris	,
6	p	a	,	the
7	i	t	the	French
8	t	a	French	man
9	a	l	man	,
10	l		,	who
11		o	who	had
12	o	f	had	been
13	f		been	in
14		t	in	the
15	F	r	the	French
16	r	a	French	army
17	a	n	army	,
18	n	c	,	had
19	c	e	had	been
20	e		been	in
21		w	in	the
22	i	s	the	army
23	s		army	.
24		1	.	He
25	t	h	He	was
26		·	was	in
27		·	in	the
28		·	the	army
29		·	army	,
30		·	,	and
31		·	and	he
32		·	he	was
33		·	was	in
34		·	in	the
35		·	the	army
36		·	army	.
37		·	.	He
38		·	He	was
39		·	was	in
40		·	in	the
41		·	the	army
42		·	army	,
43		·	,	and
44		·	and	he
45		·	he	was
46		·	was	in
47		·	in	the
48		·	the	army
49		·	army	.
50		·	.	He
51		·	He	was
52		·	was	in
53		·	in	the