prettier

Signed-off-by: Will Manning <will@willmanning.io>

Author: lwwmanning

proposed/0033-block-turboquant.md

@@ -231,12 +231,12 @@ computation.
 
 **QJL strategy options** (to be experimentally compared):
 
-| Strategy | Theoretical | Variance | Padding waste | Storage | Speed (per block) |
-| --- | --- | --- | --- | --- | --- |
-| Per-block Gaussian | Correct (Lemma 4) | (π/(2B))·‖y‖² | None | k×B²×4 bytes | O(B²) |
-| Per-block SORF | Approximate | ~(π/(2B))·‖y‖² | None | k×3×B bits | O(B log B) |
-| Full-dim padded SORF | Approximate | ~(π/(2d))·‖y‖² | (pad-d)/pad | 3×padded_d bits | O(d log d) |
-| MSE-only | N/A | N/A | N/A | None | 0 |
+| Strategy             | Theoretical       | Variance       | Padding waste | Storage         | Speed (per block) |
+| -------------------- | ----------------- | -------------- | ------------- | --------------- | ----------------- |
+| Per-block Gaussian   | Correct (Lemma 4) | (π/(2B))·‖y‖²  | None          | k×B²×4 bytes    | O(B²)             |
+| Per-block SORF       | Approximate       | ~(π/(2B))·‖y‖² | None          | k×3×B bits      | O(B log B)        |
+| Full-dim padded SORF | Approximate       | ~(π/(2d))·‖y‖² | (pad-d)/pad   | 3×padded_d bits | O(d log d)        |
+| MSE-only             | N/A               | N/A            | N/A           | None            | 0                 |
 
 #### Norm architecture
 
@@ -556,6 +556,7 @@ Compare all four strategies at d=768 with B ∈ {64, 128, 256}:
 - **MSE-only** (no QJL): measure inner product bias to establish baseline.
 
 Key metrics:
+
 - Mean signed relative error on inner products (bias)
 - Variance of inner product estimates
 - ANN recall@k on standard benchmarks (e.g., SIFT, GloVe)