feat(cellular): region ticks of up to 4,096 lives for testing (schema 68)
Foundation / checks (checks) (push) Successful in 5s
Foundation / checks (client) (push) Successful in 5s
Foundation / checks (ecs) (push) Successful in 4s
Foundation / checks (deploy) (push) Successful in 5s
Foundation / checks (network) (push) Successful in 4s
Foundation / checks (postgres) (push) Successful in 5s
Foundation / checks (tsan) (push) Successful in 4s
Foundation / postgres-tools (push) Successful in 31s
Foundation / server-image (push) Successful in 4m29s
Foundation / image (push) Successful in 5m43s

ADR-0182 section 17. The plan's 2,000-organism cap became the rules-12
key region.organism_cap, which was parsed but never enforced; the real
bound was max_cellular_tick_lives = 256 (ADR-0106), which rules 13's
organism cap could not exceed. The bound is now 4,096, for testing limits
only; the DEV tuning file keeps 256.

- max_cellular_tick_lives (4,096) bounds the raw region path: step, owner
  state, batch trace (cellular_batch_bytes_for) and current checkpoint
  traces. max_cellular_frozen_lives (256) keeps cohort populations,
  population batches, migration traces and pre-v3 cellular traces, whose
  database checks are fixed.
- Tuning: reproduction.organism_cap and wildlife_cap accept up to 4,096.
- Schema 68: the trace checker under the current publishers
  (cellular_births_v45) accepts 4,096 intents, births and mutations per
  tick; spawn_wildlife takes up to 256 cells a command.
- Region view and portal: the lineage archive's fill (births of 65,536,
  genomes of 4,096), at which a region stops ticking.
- PostgreSQL suite: 256 wildlife spawned in one command, then a 289-life
  tick published and recovered exactly.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
dingo
2026-10-28 17:30:22 -04:00
co-authored by Claude Opus 5.5
parent 8005e7c4ae
commit 3c753970c8
29 changed files with 301 additions and 49 deletions
+27 -4
View File
@@ -375,10 +375,33 @@ Playtesting region 16 (230 spawned wildlife under the 256-life cap) found three
to the cell inspector, Remove all wildlife to Actions, and marking on the map (shift+click,
shift+drag a box) to remove the marked cells, one `remove_cell` each.
Not changed: the organism cap stays at most 256 (`max_cellular_tick_lives`). That bound sizes
the batch trace, checkpoints, migration traces and the database's trace checks, and region 16
already runs at about a quarter of real time with about 250 lives. Raising it is a separate
format and performance change.
The organism cap itself is section 17.
### 17. Larger region ticks for testing (delivered 2026-10-28, schema 68)
The region/movement plan set a 2,000-organism region cap, but the rules-12 key it became
(`region.organism_cap`) was parsed and never enforced. The real bound was the engine's
`max_cellular_tick_lives = 256`, a per-tick safety bound from ADR-0106 that the batch trace,
checkpoints and database trace checks were sized from; rules 13's `reproduction.organism_cap`
was capped at it.
- **Decision (2026-10-28).** The tick bound is 4,096 lives, for testing limits only; regions
are not meant to run that many. The DEV tuning file keeps 256.
- **Split bound.** `max_cellular_tick_lives` (4,096) bounds the raw region path: the step, the
owner state, the batch trace (`cellular_batch_bytes_for`) and current checkpoint traces.
`max_cellular_frozen_lives` (256) keeps the formats whose database checks are fixed: cohort
populations, population batches, migration traces and pre-v3 cellular traces. Rules 12 and
13 never use those paths.
- **Tuning.** `reproduction.organism_cap` and `reproduction.wildlife_cap` accept up to 4,096.
- **Schema 68.** The trace checker under the current publishers (schema 43's, renamed
`cellular_births_v45` by schema 46) accepts up to 4,096 intents, births and mutation choices
per tick; `spawn_wildlife` takes up to 256 cells a command (sixteen commands fill a tick).
- **Archive.** The lineage archive never forgets: a region stops ticking at 65,536 recorded
births or 4,096 genomes. The region view reports both (`archive`) and the portal's
Population shows the fuller one, warning at 80%. A crowded region reaches them sooner.
- **Not changed.** Performance: region 16 ran at about 5 of 20 ticks per second with about
250 lives, so a region near the bound will run far below real time. The client snapshot is
budgeted (nearest changes first), so it degrades rather than fails.
## Delivery order