AbstractPhil commited on
Commit
7c1aaf1
Β·
verified Β·
1 Parent(s): 331c05d

docs/technical/01_arm_program.md: paraphrase the program lead's directives (no inline quotations)

Browse files
Files changed (1) hide show
  1. docs/technical/01_arm_program.md +2 -2
docs/technical/01_arm_program.md CHANGED
@@ -42,7 +42,7 @@
42
 
43
  **stop and stopnl: the never-terminating graduate, and the symbol an arm cannot mint.** The graduate reads stop 0.0 and length 200 β€” the cap β€” at greedy, T = 0.7 and T = 1.0 (longest repeat 3 greedy, 2 and 1 sampled), no speaker drift; that row is the "Stephanie ramble" of Phil's screenshot. **stopnl-s1600** (default, chat rows with the newline pair at each assistant turn end, 1,600 steps, one seed) takes stop rate from 0/6 to **6/6**, mean reply from 200 to 109.17 bytes, longest repeat from 3 to 1, and holds the capabilities (p2 .533 = .533, p4 .40 β†’ .467, p8 .50 β†’ .433); paired with identity it stops at 1.0 at T = 0.7 and T = 1.0. Its twin **stop-s1200** targets NUL (0x00), a byte the prior never emits, and after 1,200 steps reads **P(NUL at turn end) = 0.0000**, unmoved from ~4e-8: a small hidden-space arm *steers* the core's prior and cannot *mint* a symbol the prior never emits, so control-symbol vocabulary enters at core training β€” which is how the successor's thirteen specials were born (section 3). One seed, one symbol, and the anchor names carry 1,200 against 1,600 steps where the statute says "same steps" β€” unreconciled β€” so **mint-versus-steer is a candidate**; the NUL anchor ships as a labeled refuted control, its number in no hub ledger.
44
 
45
- **The four frame arms are what the distillation lane could not do through index space, done through byte-state space.** Campaign B (08-18 night; Phil: "take both the t5-base and bert, teach them to a beatrix arm now") attaches one wide arm and trains it at the B9 last-byte anchors of codex C β€” 999 states, 55,192 sites, raw lines, no chat frame β€” against her own frozen identity map as preservation anchor plus unit-sphere MSE to bert-base layer-4 first-byte and vanilla t5-base layer-6 last-byte frame vectors, 6,000 steps, `descent_gate` first. The convention map ran before any bar was set: her B4 last-byte identification is .9991 overall and .988 on digits, *above* both teachers, so the teachable object was the teachers' frame *geometry*, never identity. **frameB-s0/s1** (v1, site-level MSE, two seeds agreeing to the third decimal) passed every bar: frame-meet .5151/.5353 and .5151/.5355 against parity .296/.3208 (+.22/+.21), B9 digit identification *improved* from .837 to .9713/.9741, held-codex next-byte top-1 up +.0226/+.0197 with the arm on, adapters-off bit-exact β€” the first positive teacher import into Beatrix, claimed as bias-import, never capability. **frameB2-s0/s1** (v2, state-table InfoNCE plus 0.5Γ— MSE) also passed 2/2 β€” meet .4754/.5061 and .4769/.5075, B9 digit .9759/.9778, LM +.0147/+.0175 β€” but **collapsed spread**: effective rank 139.4/139.8 against v1's 179.5/180.5 and the unadapted 239.2. Target granularity governs spread, not loss family, and v1 is the recommended arm. A near-miss rides with them: the v2 runner overwrote the v1 anchor files, and the true v1 arms survived only because ship-on-completion had put them on the hub. Canon still prints the v1 B9 digit as ".974/.978", a v1/v2 mix; the ledgers read .9713/.9741.
46
 
47
  **day1 and night stacks: pairs compose, three-stacks compress.** stopnl + identity (night, one seed) composes β€” stop 1.0, length 70.0, repeat 1, capabilities held. Add the third conditioner and the stack **compresses toward its shortest member**: stopnl (109 bytes) + identity (101) + defs (32) yields 13.17-byte replies, four of six empty, and the day replicate with defs-XL reads 19.67 β€” the same 19.67 as the day three-stack of capability arms. The night capability pair reached repeat 47 and stopnl could not rescue it. None of this is a new law; it is the byte-scale instance of the amplitude budget's conservation under the ungated always-on regime, and the audit below says what the session got wrong about the remedy. One absence should be stated: no untrained arm at matched count was ever attached as a capacity control, so every collective margin is against the bare core, not random capacity.
48
 
@@ -92,7 +92,7 @@ The array as indexed, one row per anchor. Width is the state-dict count; results
92
 
93
  ### The laws and their audit
94
 
95
- **(2026-08-17, 20:29 β†’ 08-22.)** The campaign minted thirteen laws in one session, and Phil's response now governs minting: "you're disregarding a large amount of the mind's research to make these laws." Eight agents (956,694 tokens, 141 tool calls) graded every mint against the registries and found **two contradictions, four duplicates under new names, five uncited instances, two novelties and one blocked claim**, plus the process violations β€” no experiment-lines row for bdist-e001, no loss-manifest rows for byte-KL or alloyT4, plain-integer seeds, TF32 unasserted, no collinearity gate. The meta-law is the result: **a session may mint no law without a grep-audit first**, and the plan's law text was demoted to citations of registered statutes.
96
 
97
  **The capacity-data law was a contradiction, and the scope clause is the law.** The session's "rows β‰₯ ~5Γ— the arm's question space" collides with the v35 question-space law "|Q| β‰₯ 3Γ— draws" (rated 10, code-enforced): the two inequalities are jointly unsatisfiable. They are two regimes of one law β€” generalization-seeking experts obey the first; detachable overfit-by-design arms obey the second, templates varied and the guard expected-and-logged β€” and the 08-15 arm-B saturation (size the arm down) and the 08-17 day campaign (size the corpus up) are its two dials. The aliases "arm-capacity law" and "capacity-data law" are retired for the **conditioning-corpora law**. "Template competence is scale-invariant" was rescoped to what was tested β€” same-template corpora at 0.8B with a 6.26M adapter (train 1.000, held .270/.245) and at 112.5M with 3.2M/5.3M arms (sub3 .20 β†’ .00, wide .02, poly .00), with 24,000 varied rows transferring at the second scale β€” never unconditional; and "frame-disjoint" was renamed **surface-template-disjoint** because "frame" now carries four senses.
98
 
 
42
 
43
  **stop and stopnl: the never-terminating graduate, and the symbol an arm cannot mint.** The graduate reads stop 0.0 and length 200 β€” the cap β€” at greedy, T = 0.7 and T = 1.0 (longest repeat 3 greedy, 2 and 1 sampled), no speaker drift; that row is the "Stephanie ramble" of Phil's screenshot. **stopnl-s1600** (default, chat rows with the newline pair at each assistant turn end, 1,600 steps, one seed) takes stop rate from 0/6 to **6/6**, mean reply from 200 to 109.17 bytes, longest repeat from 3 to 1, and holds the capabilities (p2 .533 = .533, p4 .40 β†’ .467, p8 .50 β†’ .433); paired with identity it stops at 1.0 at T = 0.7 and T = 1.0. Its twin **stop-s1200** targets NUL (0x00), a byte the prior never emits, and after 1,200 steps reads **P(NUL at turn end) = 0.0000**, unmoved from ~4e-8: a small hidden-space arm *steers* the core's prior and cannot *mint* a symbol the prior never emits, so control-symbol vocabulary enters at core training β€” which is how the successor's thirteen specials were born (section 3). One seed, one symbol, and the anchor names carry 1,200 against 1,600 steps where the statute says "same steps" β€” unreconciled β€” so **mint-versus-steer is a candidate**; the NUL anchor ships as a labeled refuted control, its number in no hub ledger.
44
 
45
+ **The four frame arms are what the distillation lane could not do through index space, done through byte-state space.** Campaign B (08-18 night; Phil's directive that both the t5-base and bert frames be taught to a Beatrix arm at once) attaches one wide arm and trains it at the B9 last-byte anchors of codex C β€” 999 states, 55,192 sites, raw lines, no chat frame β€” against her own frozen identity map as preservation anchor plus unit-sphere MSE to bert-base layer-4 first-byte and vanilla t5-base layer-6 last-byte frame vectors, 6,000 steps, `descent_gate` first. The convention map ran before any bar was set: her B4 last-byte identification is .9991 overall and .988 on digits, *above* both teachers, so the teachable object was the teachers' frame *geometry*, never identity. **frameB-s0/s1** (v1, site-level MSE, two seeds agreeing to the third decimal) passed every bar: frame-meet .5151/.5353 and .5151/.5355 against parity .296/.3208 (+.22/+.21), B9 digit identification *improved* from .837 to .9713/.9741, held-codex next-byte top-1 up +.0226/+.0197 with the arm on, adapters-off bit-exact β€” the first positive teacher import into Beatrix, claimed as bias-import, never capability. **frameB2-s0/s1** (v2, state-table InfoNCE plus 0.5Γ— MSE) also passed 2/2 β€” meet .4754/.5061 and .4769/.5075, B9 digit .9759/.9778, LM +.0147/+.0175 β€” but **collapsed spread**: effective rank 139.4/139.8 against v1's 179.5/180.5 and the unadapted 239.2. Target granularity governs spread, not loss family, and v1 is the recommended arm. A near-miss rides with them: the v2 runner overwrote the v1 anchor files, and the true v1 arms survived only because ship-on-completion had put them on the hub. Canon still prints the v1 B9 digit as ".974/.978", a v1/v2 mix; the ledgers read .9713/.9741.
46
 
47
  **day1 and night stacks: pairs compose, three-stacks compress.** stopnl + identity (night, one seed) composes β€” stop 1.0, length 70.0, repeat 1, capabilities held. Add the third conditioner and the stack **compresses toward its shortest member**: stopnl (109 bytes) + identity (101) + defs (32) yields 13.17-byte replies, four of six empty, and the day replicate with defs-XL reads 19.67 β€” the same 19.67 as the day three-stack of capability arms. The night capability pair reached repeat 47 and stopnl could not rescue it. None of this is a new law; it is the byte-scale instance of the amplitude budget's conservation under the ungated always-on regime, and the audit below says what the session got wrong about the remedy. One absence should be stated: no untrained arm at matched count was ever attached as a capacity control, so every collective margin is against the bare core, not random capacity.
48
 
 
92
 
93
  ### The laws and their audit
94
 
95
+ **(2026-08-17, 20:29 β†’ 08-22.)** The campaign minted thirteen laws in one session, and Phil's response β€” that the laws had been minted without regard for the research already on record β€” now governs minting. Eight agents (956,694 tokens, 141 tool calls) graded every mint against the registries and found **two contradictions, four duplicates under new names, five uncited instances, two novelties and one blocked claim**, plus the process violations β€” no experiment-lines row for bdist-e001, no loss-manifest rows for byte-KL or alloyT4, plain-integer seeds, TF32 unasserted, no collinearity gate. The meta-law is the result: **a session may mint no law without a grep-audit first**, and the plan's law text was demoted to citations of registered statutes.
96
 
97
  **The capacity-data law was a contradiction, and the scope clause is the law.** The session's "rows β‰₯ ~5Γ— the arm's question space" collides with the v35 question-space law "|Q| β‰₯ 3Γ— draws" (rated 10, code-enforced): the two inequalities are jointly unsatisfiable. They are two regimes of one law β€” generalization-seeking experts obey the first; detachable overfit-by-design arms obey the second, templates varied and the guard expected-and-logged β€” and the 08-15 arm-B saturation (size the arm down) and the 08-17 day campaign (size the corpus up) are its two dials. The aliases "arm-capacity law" and "capacity-data law" are retired for the **conditioning-corpora law**. "Template competence is scale-invariant" was rescoped to what was tested β€” same-template corpora at 0.8B with a 6.26M adapter (train 1.000, held .270/.245) and at 112.5M with 3.2M/5.3M arms (sub3 .20 β†’ .00, wide .02, poly .00), with 24,000 varied rows transferring at the second scale β€” never unconditional; and "frame-disjoint" was renamed **surface-template-disjoint** because "frame" now carries four senses.
98