v4 has no Compudynamics beyond heat counters: no rolling window, decay, pipelining, inference, L8, heartbeat -- blocks alpha.5 #189

New Issue

2026-11-12T21:36:25Z

rajames commented

2026-11-12 21:36:25 +00:00

Found

v4 -- the system every acceptance boot runs -- has no Compudynamics beyond its heat counters. Found while wiring #188 (2026-11-12): the rolling window #188 gated is v3's alone; v4 never reaches it.

What v4 has (v4/src/heat.c): a count per opcode (32), the anti-clock, a count per call target (4096 slots -- a stand-in by its own account; a target at 4096 or above is not counted, so every capsule word, from 8192, gets nothing; ACL-HEAT@ returns 0).

What v4 does not have, each live in v3 whenever a word executes:

Loop	v3	v4
L1 execution heat per word	`physics_execution_heat_increment`	call-target count only (above)
L2 rolling window	`rolling_window_of_truth.c`	none
L3 linear decay	`physics_metadata.c`	none
L4 pipelining	`physics_pipelining_metrics.c`	none
L5 window inference	`inference_engine.c`	none
L6 decay inference	`inference_engine.c`	none
L7 adaptive heartrate	`HeartbeatState`	none on the node (the kernel's tick only)
L8 Jacquard selector	`ssm_jacquard.c`	none

Already recorded, never tracked as a release blocker: docs/v4.0.0/V3-PARITY.md section 1 row 3 and section 2.2 ("There is no decay, rolling window, pipelining, inference, L8, heartbeat or Stadium wire on the node", 2026-10-07); #4 (parity, milestone v4.0.0). The v4 design puts these in hardware registers -- HEAT-OP, HEAT-CALL, ANTICLOCK, HEARTBEAT, GOV-*, the governor as a multi-level Rolling Window of Truth (DECOMPOSITION.md sections 7 and ~995, JUSTIFICATION.md sections 4 and 7) -- not "as v3".

Consequence: acceptance passes on a system with none of the physics; #188's fix, and any fix to L1-L8, is seen only in host tests.

Ruled

Captain Bob, 2026-11-12: "That's a critical finding that blocks alpha.0" -- it blocks the alpha.5 release; a new critical issue.

Done when

Ruled: which loops v4 carries for alpha.5, and in what form (the golden model of the DECOMPOSITION.md section 7 registers, or v3's loops on the node)
Each, a test written first and seen to fail, then passing
Acceptance asserts it on the three architectures and the stick -- a boot without the physics fails
V3-PARITY.md section 2 and #4 say what is now built

🤖 Filed with Claude Code

### Found v4 -- the system every acceptance boot runs -- has no Compudynamics beyond its heat counters. Found while wiring #188 (2026-11-12): the rolling window #188 gated is v3's alone; v4 never reaches it. What v4 has (`v4/src/heat.c`): a count per opcode (32), the anti-clock, a count per call target (4096 slots -- a stand-in by its own account; a target at 4096 or above is not counted, so every capsule word, from 8192, gets nothing; `ACL-HEAT@` returns 0). What v4 does not have, each live in v3 whenever a word executes: | Loop | v3 | v4 | |---|---|---| | L1 execution heat per word | `physics_execution_heat_increment` | call-target count only (above) | | L2 rolling window | `rolling_window_of_truth.c` | none | | L3 linear decay | `physics_metadata.c` | none | | L4 pipelining | `physics_pipelining_metrics.c` | none | | L5 window inference | `inference_engine.c` | none | | L6 decay inference | `inference_engine.c` | none | | L7 adaptive heartrate | `HeartbeatState` | none on the node (the kernel's tick only) | | L8 Jacquard selector | `ssm_jacquard.c` | none | Already recorded, never tracked as a release blocker: `docs/v4.0.0/V3-PARITY.md` section 1 row 3 and section 2.2 ("There is no decay, rolling window, pipelining, inference, L8, heartbeat or Stadium wire on the node", 2026-10-07); #4 (parity, milestone v4.0.0). The v4 design puts these in hardware registers -- `HEAT-OP`, `HEAT-CALL`, `ANTICLOCK`, `HEARTBEAT`, `GOV-*`, the governor as a multi-level Rolling Window of Truth (`DECOMPOSITION.md` sections 7 and ~995, `JUSTIFICATION.md` sections 4 and 7) -- not "as v3". Consequence: acceptance passes on a system with none of the physics; #188's fix, and any fix to L1-L8, is seen only in host tests. ### Ruled Captain Bob, 2026-11-12: "That's a critical finding that blocks alpha.0" -- it blocks the **alpha.5 release**; a new critical issue. ### Done when - [ ] Ruled: which loops v4 carries for alpha.5, and in what form (the golden model of the `DECOMPOSITION.md` section 7 registers, or v3's loops on the node) - [ ] Each, a test written first and seen to fail, then passing - [ ] Acceptance asserts it on the three architectures and the stick -- a boot without the physics fails - [ ] `V3-PARITY.md` section 2 and #4 say what is now built 🤖 Filed with [Claude Code](https://claude.com/claude-code)

rajames added this to the v4.0.0-alpha.5: SDK milestone 2026-11-12 21:36:25 +00:00

rajames added the area/kernel

area/v4 labels 2026-11-12 21:36:26 +00:00

rajames commented

2026-11-12 21:46:47 +00:00

The Stadium: part of it is there -- and a v4 boot never starts it

Captain Bob, 2026-11-12: "Isn't all that shit in the stadium already?"

What the Stadium has (kernel/src/vm/stadium.c, stadium_words.c, stadium_blocks.c): conserved heat per cell (Q48.16, shares of 1.0), TTL, pin, admission and eviction; per-VM quotas and a reservoir; heat per word (stadium_word_dispatch) and per block, each decaying on heartbeat ticks. That is roughly L1 (heat) and L3 (decay), in the fabric's conserved form rather than v3's counters.

What it does not have: the rolling window, pipelining, the two inferences, the heartrate loop, the Jacquard selector -- none in the Stadium sources; they live only in v3's code (v3/src/, v3's dispatch in kernel/src/vm/).

Why v4 has none of it: kernel_main.c:647 hands the boot to sk_v4_run(), whose REPL does not return. stadium_boot_init() (:663) and stadium_birth_hera() (:702) come after it, so a v4 boot never starts the Stadium; the only thing that feeds it is v3's dispatch (kernel/src/vm/vm_core.c:783). The amd64 acceptance log of 2026-11-12 (logs/20261112-210607/amd64/) does not mention the Stadium. As V3-PARITY.md section 1 has it: rows 1 (the Stadium) and 9 (its conservation checks), missing on v4.

A third way for the first Done-when: start the Stadium on a v4 boot and feed it from v4's dispatch, as vm_core.c:783 does for v3 -- L1 and L3 then in the fabric's form -- leaving to rule only what L2 and L4-L8 become: in the Stadium, the design's GOV-* registers (DECOMPOSITION.md section 7), or none. A concrete proposal, read from the v4 design, comes before anything is built.

🤖 Added with Claude Code

### The Stadium: part of it is there -- and a v4 boot never starts it Captain Bob, 2026-11-12: "Isn't all that shit in the stadium already?" **What the Stadium has** (`kernel/src/vm/stadium.c`, `stadium_words.c`, `stadium_blocks.c`): conserved heat per cell (Q48.16, shares of 1.0), TTL, pin, admission and eviction; per-VM quotas and a reservoir; heat per word (`stadium_word_dispatch`) and per block, each decaying on heartbeat ticks. That is roughly L1 (heat) and L3 (decay), in the fabric's conserved form rather than v3's counters. **What it does not have:** the rolling window, pipelining, the two inferences, the heartrate loop, the Jacquard selector -- none in the Stadium sources; they live only in v3's code (`v3/src/`, v3's dispatch in `kernel/src/vm/`). **Why v4 has none of it:** `kernel_main.c:647` hands the boot to `sk_v4_run()`, whose REPL does not return. `stadium_boot_init()` (`:663`) and `stadium_birth_hera()` (`:702`) come after it, so a v4 boot never starts the Stadium; the only thing that feeds it is v3's dispatch (`kernel/src/vm/vm_core.c:783`). The amd64 acceptance log of 2026-11-12 (`logs/20261112-210607/amd64/`) does not mention the Stadium. As `V3-PARITY.md` section 1 has it: rows 1 (the Stadium) and 9 (its conservation checks), missing on v4. **A third way for the first Done-when:** start the Stadium on a v4 boot and feed it from v4's dispatch, as `vm_core.c:783` does for v3 -- L1 and L3 then in the fabric's form -- leaving to rule only what L2 and L4-L8 become: in the Stadium, the design's `GOV-*` registers (`DECOMPOSITION.md` section 7), or none. A concrete proposal, read from the v4 design, comes before anything is built. 🤖 Added with [Claude Code](https://claude.com/claude-code)

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: rajames/LithosAnanke#189