BACKLOG.md

Red Moon Backlog

Aspirational. Nothing here is committed work. Promote an item to ROADMAP.md before building it.

• Offline parser for the DOTS ECS blobs (ContentArchives, EntityScenes), so item stats need no running game. Rejected for cycle 1: per-patch reverse-engineering cost against a moving binary format.
• Overlay surface rendered on top of the game.
• Multi-player or dedicated-server support beyond a single local host.
• Screenshot and OCR tier as a bridge fallback.
• Automated V Rising launch plus OBS capture, so a validation session starts and records itself instead of being driven by hand. Raised by the operator 2026-07-26 during cycle 2. Scope is not settled: at minimum an RM-* launch task for the client and the dedicated server, an OBS scene and profile pinned to the game window, and a capture start and stop hook. It is an ops concern, so it belongs with cycle 8 rather than in the middle of the bridge, and it was deliberately NOT started inside the cycle 2 spike chain.
• PROMOTED 2026-08-01 and no longer aspirational. A falsification path for Bloodforge output, so a computed DPS, EHP or time-to-kill can be checked against an observed kill rather than only against its own inputs. Raised 2026-08-01, settled the same day by docs/superpowers/specs/2026-08-01-bloodforge-falsification-design.md and now tracked as ROADMAP cycle 3 gap 7. The scope questions below WERE the open ones and are now answered: what gets recorded is a bridge-side boss Health.Value time series at whatever rate the host's frame rate yields - the 4 Hz written here was never measured, and a dedicated server under -batchMode -nographics gave 1.99 Hz, ROADMAP gap 12 - a run is identified by a manifest keyed on build, target, difficulty and an equipped-item guid MAP, and agreement is a per-hit median APE of 2 percent plus an active-time TTK within 15 percent. Retained here for the reasoning, which stands: Cycle 3's six acceptance criteria are all about SOURCING INPUTS - none of them requires the computed number to be right, so all six can pass with a confidently wrong TTK. That is the cycle 2 lesson ("a real measurement can answer the RIGHT question about the WRONG SUBJECT") one level up: sourcing every field correctly says nothing about whether the math over them is sound. V Rising has no rewind or replay file, so the anchor has to be a recorded combat log or a hand-timed kill against a known V Blood with a known loadout, n small but non-zero. Scope is not settled: what gets recorded, how a run is identified, and what tolerance counts as agreement are all open. This does NOT block cycle 3 phase 2, which is ingest rather than math, but it must be settled BEFORE the combat-math spec opens - a time-to-kill published against an unfalsifiable model is the items.tier mistake with a bigger blast radius.
• PROMOTED 2026-08-01. A declared and displayed default subject vector for Bloodforge, now specified in section E of the falsification spec (Dracula at level 91, Normal difficulty, a GreatSword T08 whose pick is labelled ARBITRARY because 72 weapon rows tie at the maximum PhysicalPower, gear as a guid map rather than a scalar because gear_score covers 0 of 205 weapons, and blood named but not priced). A THIRD axis was found later the same day and is ROADMAP gap 9: difficulty scales boss level, power and health, so every figure in vbloods.json is implicitly a Normal one. The display half belongs to cycle 4 and is concept F3 in docs/research/DASHBOARD_CONCEPTS.md. Raised 2026-08-01. A time-to-kill is meaningless without a target, so the moment the engine computes one it must pick a default V Blood, a default gear level and a default blood quality. Whatever it picks silently ranks every weapon and every build for every user who does not override it, which makes an unexamined default a structural bias rather than a cosmetic one. The 65 vbloods rows on disk mean this is a choice among knowns, not a guess. Two parts: write down the default vector and WHY each component was chosen, and have the consuming surface state the assumption on screen rather than burying it. Belongs with cycle 3's combat-math spec for the choice and cycle 4 for the display.