CodexGameDev/experiments/008_catalyst_ecology/results/7b7c7a92-revision-2-preliminary-analysis.md

Experiment 008 Revision 2 Preliminary Analysis — Session 7b7c7a92

Status: complete.

Source: JSONL/catalyst-ecology-7b7c7a92-42cb-4ab2-8a81-d1316ea972c5.jsonl

Session Structure

• 5,338 events over 609.4 elapsed seconds, including a roughly 196-second break between Shoal and Bastion.
• All three eight-field expeditions completed once in displayed order.
• All 24 fields cleared first attempt; no deaths, restarts, abandoned runs, or replays.
• Twelve choices total and exactly three post-build transitions per expedition, confirming the corrective structure worked.
• Seven damage events across the session. Combat difficulty was not a meaningful constraint.
• One save occurred after all three expeditions.

The player reports that revision 2 made it “a bit easier to see how my choices impacted my play.” This validates the correction at the legibility/exposure level, not yet at the enjoyment or strategy-half-life level.

Builds

Expedition	Choice sequence	Complete build
Shoal	Bloom → Fork → Conduit → Fork	Bloom 1, Fork 2, Conduit 1
Bastion	Focus → Fork → Bloom → Fork	Focus 1, Fork 2, Bloom 1
Brood	Arc → Fork → Arc → Resonance	Arc 2, Fork 1, Resonance 1

Unlike revision 1, Bloom+Arc was not a universal core. Fork appeared everywhere, but at different levels and with different consumers. Each expedition specialized by repeating a module or using Resonance rather than sampling four distinct mechanisms.

The builds are causally distinct:

• Shoal used Fork and Bloom to create secondary hits, then Conduit converted them into 69 heavy lances. Its 237 kills were distributed across primary (46), fragments (73), sparks (70), and lances (48).
• Bastion used repeated primaries for 22 Focus ruptures while Fork produced 658 fragment hits. Bloom supplied 151 spark hits after kills began. Focus caused 13 kills and fragments 60.
• Brood omitted Bloom entirely. Fork generated hits for Arc; Arc was stacked, then Resonance amplified both. Arc triggered 114 times and caused 54 kills; fragments caused 83.

This is much stronger behavioral evidence of conditional composition than revision 1. It is still not proof of deliberate ecology reasoning because the player may have been comparing builds experimentally or avoiding repetition from the previous session.

Mature-Build Exposure

Each complete build persisted through fields five through eight:

Expedition	Mature field times	Main growth during mature fields
Shoal	9.8, 10.1, 9.9, 11.2 s	Conduit lances rose from 18 cumulative to 69; spark hits from 111 to 349
Bastion	16.8, 18.2, 19.5, 21.7 s	Fork hits rose from 198 to 658; Focus ruptures from 12 to 22
Brood	10.6, 12.9, 17.1, 15.1 s	Arc triggers rose from 44 to 114; spawned Motes from 24 to 69

The extra fields supplied approximately 41 seconds of mature Shoal, 76 seconds of mature Bastion, and 56 seconds of mature Brood, versus 7–22 seconds in revision 1. The player now reports clearer choice impact, so the first-run measurement defect was successfully repaired.

No field exhausted the secondary budget, though Bastion approached it most closely (observed minimum 18/300). Large effect chains were therefore bounded but not usually clipped.

Choice Deliberation

Deliberation did not simply decline with familiarity:

• Shoal Conduit took about 15.0 seconds.
• Bastion's first Focus took about 11.5 seconds and final Fork stack about 14.0 seconds.
• Brood's final Resonance took about 9.8 seconds.

Those pauses are consistent with real build consideration, but reading, distraction, and uncertainty remain alternate explanations. The long Conduit choice is particularly notable because the first session's Conduit selection was a misunderstanding; revision 2 placed it into a functioning secondary-hit engine.

Preliminary Interpretation

Revision 2 establishes that the system can visibly express multiple four-choice builds across different ecologies. It weakens the immediate crystallization interpretation from revision 1: Bloom+Arc was not universal, stacking occurred, and one expedition deliberately or experimentally omitted Bloom despite its known strength.

It does not yet establish why the builds differed or whether the extra exposure was enjoyable. Three readings remain:

1. The player transferred module knowledge and selected conditional causal engines for each population.
2. The player intentionally sampled three distinct builds to make the experiment informative, regardless of preference.
3. More mature fields made effects readable, but the run still lacked enough decision points or surprise to motivate replay.

Fork may be emerging as a reusable abstraction rather than a complete answer. It appeared in all builds because extra hits feed Bloom, Arc, and Conduit, while its partner determined the resulting capability. That would match “reuse components, not answers.” Alternatively, Fork×1/2 may simply be the strongest generic throughput option and thus an early universal core.

No replay followed the clearer mature exposure. As always, this is ambiguous: three eight-field runs may have provided enough evidence, may have become repetitive, or may not have produced a new specific question. The corrective test now permits that question to be asked meaningfully.

Follow-up Needed

1. Were the three specialized builds chosen in response to each population, to test different combinations, or for another reason? What was the intended role of Fork in all three?
2. During fields five through eight, did any interaction become surprising, satisfying, or fun rather than merely easier to read? When, if at all, did the extra fields become repetition?
3. Did the results suggest a specific alternative build worth replaying, and did the four-slot limit now feel productive or still premature/restrictive?

Player Report

Build selection combined remembered knowledge from the first 008 run with some intuitive experimentation. Fork was chosen across all three expeditions because it seemed like a generally decent way to put more projectiles on screen. The player also observed what appeared to be double or triple hits on large bodies, reducing the number of primary shots needed to kill them.

The additional mature fields permitted a small amount of tactical refinement. The player developed a consistent movement strategy in each expedition. They did not report a large strategic change or a new build interaction that demanded another run.

Power-up selection felt more meaningful than in revision 1, but not significantly so. The player believed essentially any four-module combination—and possibly no modules—would still complete the encounters. Because every choice was simply beneficial under permissive combat, the four-choice cap did not feel like either a productive tradeoff or a frustrating restriction. It was strategically inert.

Revised Interpretation

Revision 2 repaired observation time but exposed a second measurement problem: the problem space did not discriminate among compositions. A component budget matters only when inclusion changes credible capabilities and exclusion creates a relevant limitation. Here baseline output appeared sufficient, no run failed, damage was negligible, and later population growth never forced a build-specific response.

This does not mean higher difficulty is automatically the answer. Raising health/count until weak builds fail could create compulsory throughput and repeat Experiment 002's pressure-driven persistence. The more useful conclusion is:

Compositional choices become meaningful when their causal differences change what the player can effectively do—not merely how many additional effects appear while every route already succeeds.

Fork may be a reusable abstraction rather than a whole answer. It increased the hit surface available to different consumers and appeared to overlap larger targets. But it may also simply be overtuned generic throughput. The player's causal interpretation of multi-hits is report evidence; telemetry confirms many fragment hits but does not resolve whether multiple fragments from one impact hit the same large target as perceived.

The consistent movement strategies show that longer exposure supported execution learning. They do not establish fun: players can optimize serviceable controls while completing an experiment. One final report distinction should establish whether movement/build refinement itself was enjoyable or merely the clearest available way to finish.

Final Enjoyment Distinction

The player describes the building as enjoyable but meaningless. This is the clearest summary of 008 and resolves the remaining ambiguity.

The positive activity was real: selecting causal components, seeing more projectiles and trigger chains, and learning how a completed build behaved had some intrinsic enjoyment. The negative was not implementation friction or lack of legibility after revision 2. The environment placed so little demand on capability that the chosen architecture did not determine success, access, recovery, or a valued payoff. Refinement therefore lacked weight.

This is stronger and more specific than saying the game was easy. Difficulty matters only insofar as it makes different capabilities consequential. More enemy health alone could preserve the same universal throughput problem. The next experiment should use enemy capabilities that outgrow baseline fire through several possible causal routes, while exact qualitative choices continue over one escalating run.

Experiment 008 is closed. Do not add more fields or tune the same expeditions again. Its contribution is:

Building can be enjoyable for this player, but it becomes meaningless when the problem is insensitive to what was built.