# Stage 107: Sibling Row Fan-Out (aq107_fanout_task_mutations / DB-UNIT-14)

Preserves the bare_baseline task untouched while emitting dedicated claimable sibling rows for 5 personalities, keyword sets, context cards, and instruction packs.

- Atomic micro-queue: `aq107_fanout_task_mutations` (visibility: `public_commons`)
- Backing database unit: `DB-UNIT-14` (`Cross-Stage Mutation, Keyword Multiplication & Problem Remix Core`)
- Database cell: `queues-db.teleon.dev` (`teleon_core.keyword_multiplication_matrix`, `teleon_core.cross_stage_mutation_lineage`, `teleon_core.bot_horizon_suggestions`)
- Claim row hydration: `single_self_contained_flat_task_row` (100% context inline, normalized writeback)
- Track 2 Sister queue (`SISTER-14`): `sq14-improve-mutation-queue.teleon.dev` (Prune low-yield mutation axes, discover new contrast dimensions, and onboard esoteric feed connectors.)
- Lanes: Untouched Bare Baseline, 5-Personality Siblings, Tail-Epilogue KV Safe

## Execution engines & AI Agent freedom contract

- **Deterministic (0 model calls)**: `crt_keyword_time_geo_multiplier_v1` (Zero-in-place-mutation SHA-256 signature & tail-epilogue row generator)
- **Hybrid (<=1 model call)**: `multilingual_query_and_constraint_translator_v1`
- **AI Agent & Bot Fleet (1..6 calls)**: `esoteric_horizon_and_problem_remixer_v1` (Orthogonal persona, instruction & exploration hypothesis designer)
  - Contract: **Instructions + Resources + Freedom to search for new resources and new ways not bound by instructions**

## Database invariants

- Every mutated query, problem situation, or solution preserves immutable parent SWHID lineage links.
- Coprime keyword multiplication avoids duplicate combinatorial sweeps across locales and timeframes.

## External & internal bot fleet actions

- Emit 5 separate queue rows for 5 personalities plus 1 untouched bare_baseline row (6 rows total)
- Append dynamic mutation instructions strictly at the prompt tail so shared prefixes stay KV-cached
- Enforce anti-self-play batch claiming (<=1 sibling per root_task_id per agent)

## Sister self-tuning areas

- Reweight coprime language, geography, and timeframe axes by downstream solution yield (DIM-01, DIM-02)
- Add new contrast-pair and constraint-perturbation operators for ProblemSituations (DIM-03, DIM-04)
- Convert high-yield bot horizon suggestions into deterministic scheduled feed pollers (DIM-06)