docs(adr): ADR 0010 → Implemented; Phase 3 superseded by ADR 0013

Phases 0/1/2 (lazy decode for 1:1 transforms — ALP, FoR, ZigZag) shipped
over 2026-06-14 — 2026-06-15. Update status from Proposed to Implemented
and add an Outcome section noting:

- Phase 1 — Array hierarchy refactor to non-sealed interfaces.
- Phase 2 — LazyAlpDouble/Float, LazyForLong/Int, LazyZigZagLong/Int
  records; ArraySegments.of(arr, arena) materialise fallback.
- Pattern generalised beyond 1:1 transforms: ADR 0012 (lazy
  Chunked / Dict / VarBin) plus vortex.runend, vortex.sparse, and
  fastlanes.rle adopted the same top-level record + materialise-fallback
  shape (see LazyRunEnd*, LazySparse*, LazyRle* in reader.array).

Phase 3 (compute pushdown) is superseded by ADR 0013 — that ADR lands
the masks-and-kernels framework Phase 3 sketched, plus the wider design
needed for pushdown to compose across the lazy storage types from
ADRs 0010 and 0012.

Also update docs/adr/ADR.md index entry from Proposed to Implemented.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

Author: dfa1

docs/adr/0010-lazy-decode.md

@@ -1,7 +1,8 @@
 # ADR 0010: Lazy decode for 1:1 transform encodings
 
-- **Status:** Proposed
+- **Status:** Implemented
 - **Date:** 2026-06-13
+- **Implemented:** 2026-06-14 — 2026-06-15
 - **Deciders:** project maintainer
 - **Supersedes:** —
 - **Superseded by:** —
@@ -606,6 +607,37 @@ Net negative.
 Rejected: only works if scan does not eagerly decode — which is exactly
 phase 1.
 
+## Outcome
+
+Phases 0/1/2 (lazy decode for the 1:1 transforms) shipped over 2026-06-14 — 2026-06-15:
+
+- **Phase 1 — Array hierarchy refactor.** Primitive `Array` interfaces (`LongArray`,
+  `IntArray`, `DoubleArray`, …) converted to non-sealed so new lazy variants can implement
+  them without touching the hierarchy.
+- **Phase 2 — Lazy ALP / FoR / ZigZag.** `LazyAlpDoubleArray`, `LazyAlpFloatArray`,
+  `LazyForLongArray`, `LazyForIntArray`, `LazyZigZagLongArray`, `LazyZigZagIntArray`
+  shipped. Each holds the encoded child + transform parameters; per-row dispatch applies
+  the transform on demand. `ArraySegments.of(arr, arena)` materialises into a fresh segment
+  only when a downstream caller demands a contiguous buffer.
+
+The lazy-storage pattern generalised beyond the 1:1 transform scope of this ADR. Adjacent
+encodings adopted the same top-level record shape:
+
+- [ADR 0012 — Lazy Chunked / Dict / VarBin layouts](0012-zero-copy-layout-decoding.md).
+- `vortex.runend`, `vortex.sparse`, `fastlanes.rle` (lazy lookup tables; not 1:1 transforms
+  but the same lazy-record + `ArraySegments` materialise pattern; see
+  `LazyRunEndXxxArray`, `LazySparseXxxArray`, `LazyRleXxxArray` in `reader.array`).
+
+`docs/compatibility.md` Decode shape table tracks per-encoding status.
+
+### Phase 3 — superseded
+
+The compute-pushdown phase (per-encoding `compareXxx` / `take` / `filter` operating on the
+encoded form) is **superseded by [ADR 0013 — Compute primitives: masks, kernels,
+no-materialise](0013-compute-primitives.md)**. ADR 0013 lands the masks-and-kernels
+framework that Phase 3 sketched, plus the wider design needed to make pushdown compose
+across the lazy storage types from ADRs 0010 and 0012.
+
 ## References
 
 - Rust reference: `https://github.com/spiraldb/vortex/tree/main/encodings/alp/src/alp/compute`

docs/adr/ADR.md

@@ -22,7 +22,7 @@ Each ADR is a Markdown file named `NNNN-short-title.md`. Use `template.md` as th
 | 0007 | Pure-Java vortex.pco encoder                  | Implemented |
 | 0008 | Domain primitives for unsigned integers       | Proposed  |
 | 0009 | Write API ergonomics                          | Completed |
-| 0010 | Lazy decode                                   | Proposed  |
+| 0010 | Lazy decode                                   | Implemented |
 | 0011 | Writer zero-copy MemorySegment overload       | Deferred  |
 | 0012 | Zero-copy layout decoding: lazy Chunked/Dict  | Implemented |
 | 0013 | Compute primitives: masks, kernels, no-materialise | Proposed  |