Record the first M11.3 cellular probe run
Foundation / checks (checks) (push) Successful in 5s
Foundation / checks (client) (push) Successful in 5s
Foundation / checks (ecs) (push) Successful in 4s
Foundation / checks (network) (push) Successful in 4s
Foundation / checks (postgres) (push) Successful in 5s
Foundation / checks (tsan) (push) Successful in 4s
Foundation / checks (deploy) (push) Successful in 3m44s
Foundation / postgres-tools (push) Successful in 5s
Foundation / server-image (push) Successful in 5s
Foundation / image (push) Successful in 5s

The 2026-10-13 probe in DEV region 100/3 showed steering and feeding
working end to end after the session fixes, and not yet enough game to
judge fun. Record its findings and inputs for 11.4-11.8, and mark the
remaining 11.3 work (engulfment and corpse absorption) in the roadmap.
Also refresh the dev session window (3 h).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
dingo
2026-10-13 18:40:28 -04:00
co-authored by Claude Opus 5.5
parent 5c92cf137a
commit 4751b23857
3 changed files with 67 additions and 8 deletions
@@ -12,7 +12,7 @@ spec:
env:
- {name: REGION_SESSION, value: "dev"}
- {name: REGION_SESSION_LISTEN, value: "0.0.0.0:7400"}
- {name: REGION_SESSION_WINDOW, value: "dev-20261013T215247-52993:1791939167"}
- {name: REGION_SESSION_WINDOW, value: "dev-20261013T224028-49591:1791942028"}
- {name: REGION_M10_MAX_GRANT_LIFETIME, value: "14400"}
ports:
# QUIC (ADR-0090): UDP only; nothing listens on TCP 7400.
+4 -7
View File
@@ -605,13 +605,10 @@ is evidence that the cellular product is compelling.
Status:
- 11.1's design decisions and per-second rate migration are complete. The 11.3
engineering probe and its checklist are prepared. The maintainer reports that
`evo-region` is running and a client connects; the workspace's network client
binary predates the current cellular-trial changes, so this does not yet prove
the probe works end to end. The maintainer now reports there have been no
successful server image builds since yesterday, so the current image cannot be
deployed for the probe. Record movement/feeding notes after a deployable server
image is available and the current cellular client can connect to it.
exploratory probe ran on 2026-10-13 in DEV region 100/3 (ADR-0123); notes are
in [the probe notes](playtests/M11_3_PROBE_NOTES_2026-10-13.md). Steering and
feeding work end to end, but there is not yet enough game to judge fun.
Engulfment and corpse absorption (the predation part of 11.3) remain.
- 11.4a and 11.4b are implemented. The 11.4c reservation gate is green in CI
runs 274 and 275, but durable selection publication and restart recovery are
still outstanding. The mutation-choice playtest becomes possible in 11.4d,
@@ -0,0 +1,62 @@
# M11.3 cellular probe notes — 2026-10-13
The first hands-on run of the [M11.3 cellular probe](M11_3_CELLULAR_PROBE.md).
One maintainer, one DEV region (world 100, region 3), the engineering client
with `--cellular`, several short sessions (each grant lasts at most 300 s).
These are exploratory observations, not balance decisions (ADR-0098: tunables
change only with evidence).
## Build and region
- Server image `5e13e009` (source `e9b010e`) after the fixes below; region 100/3
cellular-v1 opening (ADR-0123): 10 × 10 mm arena, 4 rich (200M pJ), 3 lean
(50M pJ) and 1 empty patch, spawn at (2.5, 2.5) mm.
- Client: `build/client-network` from `532f45b` onward (cellular camera framing).
## What the session showed
1. **The cell could not be found at first.** The camera used the metre-scale
lineage default, so the whole arena rendered inside half a pixel. Fixed by
framing about 12 mm around the own cell (`532f45b`). Patches now draw their
true 1 mm footprint.
2. **Feeding is readable once the footprint is visible.** The player steered onto
a patch, pressed **Feed this patch** once, and saw intake committed, patch
stock fall and reserve rise (to about 106M pJ). Feeding from three different
patches worked.
3. **The first opening starved the cell while it was feeding.** Only patch
energy feeds the reserve; the 4M and 16M pJ patches emptied in seconds at
founder intake (1,148,000 pJ/s). Region 3 raised patch energy (ADR-0123).
4. **Half of all feeding actions were refused** ("too many requests") because
admission allowed 4 actions per 8 ticks while the client sends one per tick.
Fixed in `b735860`.
5. **Death was abrupt.** The founder reserve is about 30 s at full speed; the
first cell died within a minute when it did not reach a stocked patch, and
the client gave no warning before it disappeared.
6. **Session plumbing dominated the experience.** 300-second grants, an
account rate limit tripped by client auto-retry, a region livelock from a
departed account's queued birth (`f3aecaf`), and a refused life request that
broke every fresh login (`e9b010e`). All fixed except the grant lifetime,
which is a separate security decision.
## Player notes
- First confusing moment: finding the cell at all, then understanding why the
cell starved while "feeding acquired" was shown (energy vs matter).
- First satisfying moment: moving between patches and seeing the reserve climb.
- Verdict: the interaction works as an engineering slice, but there is not yet
enough game to judge fun. "We are going to have to get much further along
before real playtesting to determine if it is fun." Continue to 11.4–11.9.
## Inputs for later steps
- 11.4 eligibility (ADR-0120): a competent player gained about 100M pJ of
retained energy in a few minutes on rich patches; revisit the 60 s fed-energy
trigger against this rate before widening offers.
- 11.6/11.8: death needs a visible cause and warning (starvation came with no
feedback).
- 11.7: disconnects were frequent; the linger/freeze contract must tolerate
ordinary network drops.
- 11.8: patches as footprints, cell framing and an energy readout are the
minimum readable view; resources should distinguish energy from matter.
- Open operational items: cellular grant ceiling (300 s), the client's retry
while rate-limited, and the flaky CI client-lifecycle smoke test.