diff --git a/CHANGELOG.md b/CHANGELOG.md
index 4fddff0291..34cf5c30c1 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,6 +7,65 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.58.0] - 2026-07-03
+
+The release where **GPU QLoRA fine-tuning becomes honest and enforced.** v0.57 made
+training *run correctly*; v0.58 makes the *validation* it reports trustworthy — the
+metric now reflects the model being trained, the default hyper-parameters no longer
+diverge, and the whole GPU correctness suite runs every night on **two silicons**
+(RTX 4090 sm_89 + GB10 Blackwell sm_121). The "GPU falsifiers aren't CI-enforced"
+doctrine gap named in v0.57's roadmap is closed.
+
+### Fixed
+
+- **NF4 QLoRA per-epoch `val_loss` reflects the trained adapters** (#2257) — on the CUDA
+  path `train_step` writes adapter deltas to the GPU-resident `cuda_blocks`, but
+  `evaluate()` forwarded the CPU `lora_layers`, refreshed only inside `save_checkpoint` —
+  never before the epoch-loop eval. So per-epoch `val_loss` was byte-identical every epoch,
+  freezing best-epoch selection at 0 and firing early-stopping on a phantom plateau.
+  `evaluate()` now syncs GPU→CPU adapters first. (`FALSIFY-CUDA-EVAL-ADAPTER-SYNC-001`)
+- **CPU LoRA forward dropped the Q/K/V projection biases** (#2260) — `forward_with_lora`
+  (the path every CPU LoRA train/eval forward runs) never applied `b_q/b_k/b_v`, so on
+  Qwen2-family models (`use_bias=true`) it computed a *different, bias-less* model than
+  inference — bit-exactly matching the parity probe's biases-DROPPED oracle (CE 4.49 vs
+  2.13; 14.53 — worse than uniform — on a longer sample). Fixed grad-preservingly via an
+  autograd-aware broadcast add. (`FALSIFY-CPU-LORA-QKV-BIAS-001/002`, CI-enforced, no GPU)
+- **`evaluate()` runs the GPU forward when the model is GPU-resident** (#2260) — it used the
+  CPU forward even under CUDA: ~9.9 s/sample at seq ~50 (minutes at 2048) with the GPU idle,
+  so a real `apr finetune -m qlora` run finished its epoch then timed out *inside* the
+  validation pass. Now 89 ms/sample — a **111× per-sample speedup** — measuring the NF4 model
+  actually being trained. (`FALSIFY-CUDA-EVAL-GPU-FORWARD-001`)
+
+### Changed
+
+- **Rank-aware default learning rate** (#2258) — the planner auto-selects the largest LoRA
+  rank that fits VRAM (up to 256), but the CLI hard-defaulted lr to 2e-4, which **diverges**
+  there (measured on a 4090: loss 4.31→1.44 then blows up to 11–18, epoch avg 11.18).
+  `OptimalConfig` now recommends a rank-aware lr (2e-4 at rank ≤ 32 down to ~2.5e-5 at 256;
+  e2e-validated: epoch avg **1.58** vs 11.18 at the old default); `--learning-rate` is now
+  optional. (`FALSIFY-QLORA-RANK-AWARE-LR-001..003`, CI-enforced)
+- **Roadmap/pillars re-grounded against HEAD** (#2261) — scoreboard counts re-pinned (all
+  were understated), P3 upgraded from equivalence-only to equivalence **+ working, now-enforced
+  training**, and case-file **CF-5** recorded (a falsifier's tolerance band firing against its
+  own reference caught the CPU bias-drop — *bisect before loosening*).
+
+### CI / Infrastructure
+
+- **Cross-silicon CUDA nightly lane** (#2262, #2263) — closes `CUDA-CI-NIGHTLY-001`. The six
+  GPU QLoRA training-correctness falsifiers, previously `#[ignore]`-gated and developer-run,
+  now execute every night on a self-hosted matrix: **ada-4090** (RTX 4090, sm_89, x86-64,
+  CUDA 12.8) and **blackwell-gb10** (GB10, sm_121, aarch64, CUDA 13.0, egress via a
+  subnet-restricted proxy over the wired link). Nightly (01:30 UTC), yields to overnight
+  training (GPU-busy check), never triggers on a PR.
+- **CI checkout self-heal** (#2259) — a pre-checkout ownership-restore step ends the recurring
+  root-owned `target/.rustc_info.json` EACCES flake a hard-killed docker job leaves behind.
+- **actions/checkout bumped 4 → 7** (#2180).
+
+### Verified
+
+- Both cross-silicon nightly legs went **green end-to-end** (checkout → cargo build → all six
+  falsifiers + loss-window suite) on 2026-07-03 — ada-4090 and blackwell-gb10, run 28661079713.
+
 ## [0.57.0] - 2026-07-03
 
 The release where **GPU QLoRA fine-tuning actually works**. `apr finetune -m qlora` went
diff --git a/Cargo.lock b/Cargo.lock
index 6d9122a139..6452fc2e2f 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -255,7 +255,7 @@ dependencies = [
 
 [[package]]
 name = "apr-cli"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "aprender-common",
@@ -323,7 +323,7 @@ dependencies = [
 
 [[package]]
 name = "apr-format"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "bincode",
  "cargo_metadata",
@@ -393,7 +393,7 @@ dependencies = [
 
 [[package]]
 name = "aprender"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "apr-cli",
  "aprender-core",
@@ -401,7 +401,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-bench-compute"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-core",
  "criterion 0.7.0",
@@ -410,7 +410,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-bench-tokenizer"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-core",
  "criterion 0.7.0",
@@ -419,7 +419,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-cbtop"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "aprender-common",
@@ -444,7 +444,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-cgp"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "aprender-gpu",
@@ -464,7 +464,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-common"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "lz4_flex 0.11.6",
  "zstd",
@@ -472,7 +472,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-compute"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "aprender-core",
@@ -521,7 +521,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-compute-xtask"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "bashrs",
@@ -534,7 +534,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-contracts"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-contracts-macros",
  "criterion 0.5.1",
@@ -550,7 +550,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-contracts-cli"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-contracts",
  "clap",
@@ -561,7 +561,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-contracts-macros"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -571,7 +571,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-core"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aes-gcm",
  "alsa",
@@ -623,7 +623,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-cuda-edge"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-gpu",
  "proptest",
@@ -634,7 +634,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-cupti"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "bindgen 0.71.1",
  "bitflags 2.13.0",
@@ -644,7 +644,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-data"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aes-gcm",
  "argon2",
@@ -696,7 +696,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-db"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "aprender-common",
@@ -741,7 +741,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-distribute"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-compute",
  "aprender-db",
@@ -772,7 +772,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-explain"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-gpu",
  "aprender-present-core",
@@ -790,7 +790,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-fft"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "criterion 0.7.0",
  "proptest",
@@ -799,7 +799,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-gemm-codegen"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -808,7 +808,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-gpu"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-common",
  "aprender-simulate",
@@ -827,7 +827,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-graph"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "aprender-compute",
@@ -849,7 +849,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-image"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "criterion 0.7.0",
  "proptest",
@@ -858,7 +858,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-mcp"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "assert_cmd",
@@ -873,9 +873,9 @@ dependencies = [
 
 [[package]]
 name = "aprender-monte-carlo"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
- "aprender 0.57.0",
+ "aprender 0.58.0",
  "assert_cmd",
  "chrono",
  "clap",
@@ -892,7 +892,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-orchestrate"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "aprender-common",
@@ -966,7 +966,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-present-cli"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-present-yaml",
  "clap",
@@ -979,7 +979,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-present-core"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "criterion 0.7.0",
  "proptest",
@@ -991,7 +991,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-present-layout"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-present-core",
  "criterion 0.7.0",
@@ -1002,7 +1002,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-present-lib"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-present-core",
  "aprender-present-layout",
@@ -1031,7 +1031,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-present-terminal"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-present-core",
  "bitvec",
@@ -1052,7 +1052,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-present-test"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-present-core",
  "aprender-present-test-macros",
@@ -1063,7 +1063,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-present-test-macros"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -1072,7 +1072,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-present-widgets"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-present-core",
  "aprender-present-yaml",
@@ -1084,7 +1084,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-present-yaml"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-present-core",
  "proptest",
@@ -1094,7 +1094,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-profile"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "addr2line 0.25.1",
  "anyhow",
@@ -1147,7 +1147,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-profile-core"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "hex",
  "serde",
@@ -1157,7 +1157,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-ptx-debug"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "proptest",
  "thiserror 2.0.18",
@@ -1165,7 +1165,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-qa-certify"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "chrono",
  "proptest",
@@ -1175,7 +1175,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-qa-cli"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-qa-certify",
  "aprender-qa-gen",
@@ -1194,7 +1194,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-qa-gen"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "chrono",
@@ -1211,7 +1211,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-qa-report"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "aprender-qa-gen",
@@ -1229,7 +1229,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-qa-runner"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "aprender-qa-gen",
@@ -1254,14 +1254,14 @@ dependencies = [
 
 [[package]]
 name = "aprender-quant"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "half",
 ]
 
 [[package]]
 name = "aprender-rag"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-common",
  "aprender-compute",
@@ -1309,7 +1309,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-rand"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "criterion 0.7.0",
  "proptest",
@@ -1318,7 +1318,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-registry"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "aprender-core",
@@ -1348,7 +1348,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-serve"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "anyhow",
  "approx",
@@ -1415,7 +1415,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-shell"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-core",
  "assert_cmd",
@@ -1432,7 +1432,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-simulate"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-present-core",
  "aprender-present-terminal",
@@ -1471,7 +1471,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-solve"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "criterion 0.7.0",
  "half",
@@ -1481,7 +1481,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-sparse"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "criterion 0.7.0",
  "proptest",
@@ -1490,7 +1490,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-tensor"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "criterion 0.7.0",
  "proptest",
@@ -1499,7 +1499,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-test-cli"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-test-lib",
  "assert_cmd",
@@ -1527,7 +1527,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-test-derive"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -1537,7 +1537,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-test-js-gen"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "blake3",
  "chrono",
@@ -1553,7 +1553,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-test-lib"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-compute",
  "aprender-present-core",
@@ -1597,7 +1597,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-test-showcase"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-present-terminal",
  "console_error_panic_hook",
@@ -1614,7 +1614,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-train"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "approx",
  "aprender-common",
@@ -1683,7 +1683,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-train-bench"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-train",
  "aprender-train-common",
@@ -1700,7 +1700,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-train-common"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-viz",
  "clap",
@@ -1712,7 +1712,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-train-distill"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-train",
  "aprender-train-common",
@@ -1733,7 +1733,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-train-inspect"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-train",
  "aprender-train-common",
@@ -1749,7 +1749,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-train-lora"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-train",
  "aprender-train-common",
@@ -1765,7 +1765,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-train-shell"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-train",
  "aprender-train-common",
@@ -1781,7 +1781,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-train-wasm"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "proptest",
  "serde",
@@ -1792,9 +1792,9 @@ dependencies = [
 
 [[package]]
 name = "aprender-tsp"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
- "aprender 0.57.0",
+ "aprender 0.58.0",
  "assert_cmd",
  "clap",
  "crc32fast",
@@ -1807,7 +1807,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-verify"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "chrono",
  "criterion 0.5.1",
@@ -1819,7 +1819,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-verify-ml"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-compute",
  "aprender-core",
@@ -1849,7 +1849,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-viz"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "approx",
  "aprender-common",
@@ -1872,7 +1872,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-zram"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-zram-adaptive",
  "aprender-zram-core",
@@ -1882,7 +1882,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-zram-adaptive"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-zram-core",
  "proptest",
@@ -1891,7 +1891,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-zram-cli"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-zram-adaptive",
  "aprender-zram-core",
@@ -1905,7 +1905,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-zram-core"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-gpu",
  "criterion 0.7.0",
@@ -1917,7 +1917,7 @@ dependencies = [
 
 [[package]]
 name = "aprender-zram-generator"
-version = "0.57.0"
+version = "0.58.0"
 dependencies = [
  "aprender-zram-core",
  "serde",
diff --git a/Cargo.toml b/Cargo.toml
index d5b3bfebf9..55b4044ffe 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -109,7 +109,7 @@ exclude = [
 resolver = "2"
 
 [workspace.package]
-version = "0.57.0"
+version = "0.58.0"
 edition = "2021"
 license = "MIT"
 repository = "https://github.com/paiml/aprender"
@@ -175,7 +175,7 @@ minijinja = { version = "2.14", features = ["loader", "serde"] }
 # Contracts
 provable-contracts = "0.3"
 provable-contracts-macros = "0.3"
-aprender-contracts-macros = { path = "crates/aprender-contracts-macros", version = "0.57.0" }
+aprender-contracts-macros = { path = "crates/aprender-contracts-macros", version = "0.58.0" }
 
 # YAML
 serde_yaml = "0.9"
@@ -187,43 +187,43 @@ hf-hub = { version = "0.4", default-features = false, features = ["ureq"] }
 hf-xet = "1.5.1"
 
 # Workspace-internal crates (paths within the monorepo)
-aprender-compute = { path = "crates/aprender-compute", version = "0.57.0" }
-aprender-gpu = { path = "crates/aprender-gpu", version = "0.57.0" }
-aprender-quant = { path = "crates/aprender-quant", version = "0.57.0" }
-aprender-serve = { path = "crates/aprender-serve", version = "0.57.0" }
-realizar = { path = "crates/aprender-serve", version = "0.57.0", package = "aprender-serve" }
-alimentar = { path = "crates/aprender-data", version = "0.57.0", package = "aprender-data" }
-pacha = { path = "crates/aprender-registry", version = "0.57.0", package = "aprender-registry" }
+aprender-compute = { path = "crates/aprender-compute", version = "0.58.0" }
+aprender-gpu = { path = "crates/aprender-gpu", version = "0.58.0" }
+aprender-quant = { path = "crates/aprender-quant", version = "0.58.0" }
+aprender-serve = { path = "crates/aprender-serve", version = "0.58.0" }
+realizar = { path = "crates/aprender-serve", version = "0.58.0", package = "aprender-serve" }
+alimentar = { path = "crates/aprender-data", version = "0.58.0", package = "aprender-data" }
+pacha = { path = "crates/aprender-registry", version = "0.58.0", package = "aprender-registry" }
 # APR-MONO self-containment: legacy sibling lib-name aliases -> in-monorepo crates (2026-06-12)
-trueno-quant = { path = "crates/aprender-quant", version = "0.57.0", package = "aprender-quant" }
-trueno-db = { path = "crates/aprender-db", version = "0.57.0", package = "aprender-db" }
-trueno-viz = { path = "crates/aprender-viz", version = "0.57.0", package = "aprender-viz" }
-trueno-rag = { path = "crates/aprender-rag", version = "0.57.0", package = "aprender-rag" }
-trueno-cuda-edge = { path = "crates/aprender-cuda-edge", version = "0.57.0", package = "aprender-cuda-edge" }
-renacer-core = { path = "crates/aprender-profile-core", version = "0.57.0", package = "aprender-profile-core" }
-repartir = { path = "crates/aprender-distribute", version = "0.57.0", package = "aprender-distribute" }
-simular = { path = "crates/aprender-simulate", version = "0.57.0", package = "aprender-simulate" }
-probar = { path = "crates/aprender-test-lib", version = "0.57.0", package = "aprender-test-lib" }
-aprender-train = { path = "crates/aprender-train", version = "0.57.0" }
-entrenar = { path = "crates/aprender-train", version = "0.57.0", package = "aprender-train" }
-aprender-train-common = { path = "crates/aprender-train-common", version = "0.57.0" }
-aprender-train-distill = { path = "crates/aprender-train-distill", version = "0.57.0" }
-aprender-orchestrate = { path = "crates/aprender-orchestrate", version = "0.57.0" }
-aprender-contracts = { path = "crates/aprender-contracts", version = "0.57.0" }
-aprender-profile = { path = "crates/aprender-profile", version = "0.57.0" }
-aprender-profile-core = { path = "crates/aprender-profile-core", version = "0.57.0" }
-aprender-zram-core = { path = "crates/aprender-zram-core", version = "0.57.0" }
-aprender-zram-adaptive = { path = "crates/aprender-zram-adaptive", version = "0.57.0" }
-aprender-present-core = { path = "crates/aprender-present-core", version = "0.57.0" }
-aprender-present-terminal = { path = "crates/aprender-present-terminal", version = "0.57.0" }
-aprender-present-widgets = { path = "crates/aprender-present-widgets", version = "0.57.0" }
-aprender-present-layout = { path = "crates/aprender-present-layout", version = "0.57.0" }
-aprender-present-yaml = { path = "crates/aprender-present-yaml", version = "0.57.0" }
-aprender-present-test = { path = "crates/aprender-present-test", version = "0.57.0" }
-aprender-db = { path = "crates/aprender-db", version = "0.57.0" }
-aprender-graph = { path = "crates/aprender-graph", version = "0.57.0" }
-aprender-rag = { path = "crates/aprender-rag", version = "0.57.0" }
-aprender-data = { path = "crates/aprender-data", version = "0.57.0" }
+trueno-quant = { path = "crates/aprender-quant", version = "0.58.0", package = "aprender-quant" }
+trueno-db = { path = "crates/aprender-db", version = "0.58.0", package = "aprender-db" }
+trueno-viz = { path = "crates/aprender-viz", version = "0.58.0", package = "aprender-viz" }
+trueno-rag = { path = "crates/aprender-rag", version = "0.58.0", package = "aprender-rag" }
+trueno-cuda-edge = { path = "crates/aprender-cuda-edge", version = "0.58.0", package = "aprender-cuda-edge" }
+renacer-core = { path = "crates/aprender-profile-core", version = "0.58.0", package = "aprender-profile-core" }
+repartir = { path = "crates/aprender-distribute", version = "0.58.0", package = "aprender-distribute" }
+simular = { path = "crates/aprender-simulate", version = "0.58.0", package = "aprender-simulate" }
+probar = { path = "crates/aprender-test-lib", version = "0.58.0", package = "aprender-test-lib" }
+aprender-train = { path = "crates/aprender-train", version = "0.58.0" }
+entrenar = { path = "crates/aprender-train", version = "0.58.0", package = "aprender-train" }
+aprender-train-common = { path = "crates/aprender-train-common", version = "0.58.0" }
+aprender-train-distill = { path = "crates/aprender-train-distill", version = "0.58.0" }
+aprender-orchestrate = { path = "crates/aprender-orchestrate", version = "0.58.0" }
+aprender-contracts = { path = "crates/aprender-contracts", version = "0.58.0" }
+aprender-profile = { path = "crates/aprender-profile", version = "0.58.0" }
+aprender-profile-core = { path = "crates/aprender-profile-core", version = "0.58.0" }
+aprender-zram-core = { path = "crates/aprender-zram-core", version = "0.58.0" }
+aprender-zram-adaptive = { path = "crates/aprender-zram-adaptive", version = "0.58.0" }
+aprender-present-core = { path = "crates/aprender-present-core", version = "0.58.0" }
+aprender-present-terminal = { path = "crates/aprender-present-terminal", version = "0.58.0" }
+aprender-present-widgets = { path = "crates/aprender-present-widgets", version = "0.58.0" }
+aprender-present-layout = { path = "crates/aprender-present-layout", version = "0.58.0" }
+aprender-present-yaml = { path = "crates/aprender-present-yaml", version = "0.58.0" }
+aprender-present-test = { path = "crates/aprender-present-test", version = "0.58.0" }
+aprender-db = { path = "crates/aprender-db", version = "0.58.0" }
+aprender-graph = { path = "crates/aprender-graph", version = "0.58.0" }
+aprender-rag = { path = "crates/aprender-rag", version = "0.58.0" }
+aprender-data = { path = "crates/aprender-data", version = "0.58.0" }
 
 # Additional external deps needed by sub-crates
 serde_yaml_ng = "0.10"
@@ -268,25 +268,25 @@ wasmtime = "43"
 bollard = "0.17"
 
 # Old crate name aliases (workspace = true refs in migrated sub-crates)
-trueno-zram-core = { path = "crates/aprender-zram-core", version = "0.57.0", package = "aprender-zram-core" }
-trueno-zram-adaptive = { path = "crates/aprender-zram-adaptive", version = "0.57.0", package = "aprender-zram-adaptive" }
-trueno = { path = "crates/aprender-compute", version = "0.57.0", package = "aprender-compute" }
-presentar-core = { path = "crates/aprender-present-core", version = "0.57.0", package = "aprender-present-core" }
-presentar-terminal = { path = "crates/aprender-present-terminal", version = "0.57.0", package = "aprender-present-terminal" }
-presentar-widgets = { path = "crates/aprender-present-widgets", version = "0.57.0", package = "aprender-present-widgets" }
-presentar-layout = { path = "crates/aprender-present-layout", version = "0.57.0", package = "aprender-present-layout" }
-presentar-yaml = { path = "crates/aprender-present-yaml", version = "0.57.0", package = "aprender-present-yaml" }
-presentar-test = { path = "crates/aprender-present-test", version = "0.57.0", package = "aprender-present-test" }
-presentar-test-macros = { path = "crates/aprender-present-test-macros", version = "0.57.0", package = "aprender-present-test-macros" }
-jugar-probar-derive = { path = "crates/aprender-test-derive", version = "0.57.0", package = "aprender-test-derive" }
-batuta-common = { path = "crates/aprender-common", version = "0.57.0", package = "aprender-common" }
+trueno-zram-core = { path = "crates/aprender-zram-core", version = "0.58.0", package = "aprender-zram-core" }
+trueno-zram-adaptive = { path = "crates/aprender-zram-adaptive", version = "0.58.0", package = "aprender-zram-adaptive" }
+trueno = { path = "crates/aprender-compute", version = "0.58.0", package = "aprender-compute" }
+presentar-core = { path = "crates/aprender-present-core", version = "0.58.0", package = "aprender-present-core" }
+presentar-terminal = { path = "crates/aprender-present-terminal", version = "0.58.0", package = "aprender-present-terminal" }
+presentar-widgets = { path = "crates/aprender-present-widgets", version = "0.58.0", package = "aprender-present-widgets" }
+presentar-layout = { path = "crates/aprender-present-layout", version = "0.58.0", package = "aprender-present-layout" }
+presentar-yaml = { path = "crates/aprender-present-yaml", version = "0.58.0", package = "aprender-present-yaml" }
+presentar-test = { path = "crates/aprender-present-test", version = "0.58.0", package = "aprender-present-test" }
+presentar-test-macros = { path = "crates/aprender-present-test-macros", version = "0.58.0", package = "aprender-present-test-macros" }
+jugar-probar-derive = { path = "crates/aprender-test-derive", version = "0.58.0", package = "aprender-test-derive" }
+batuta-common = { path = "crates/aprender-common", version = "0.58.0", package = "aprender-common" }
 # APR-MONO sibling aliases — MUST use workspace refs, NEVER crates.io versions.
 # Hard-pinning these to crates.io creates a diamond: sub-crates pull older
 # aprender versions → multiple `aprender` crates in Cargo.lock = MUDA.
-renacer = { path = "crates/aprender-profile", version = "0.57.0", package = "aprender-profile" }
-entrenar-lora = { path = "crates/aprender-train-lora", version = "0.57.0", package = "aprender-train-lora" }
-batuta = { path = "crates/aprender-orchestrate", version = "0.57.0", package = "aprender-orchestrate" }
-trueno-graph = { path = "crates/aprender-graph", version = "0.57.0", package = "aprender-graph" }
+renacer = { path = "crates/aprender-profile", version = "0.58.0", package = "aprender-profile" }
+entrenar-lora = { path = "crates/aprender-train-lora", version = "0.58.0", package = "aprender-train-lora" }
+batuta = { path = "crates/aprender-orchestrate", version = "0.58.0", package = "aprender-orchestrate" }
+trueno-graph = { path = "crates/aprender-graph", version = "0.58.0", package = "aprender-graph" }
 
 [workspace.lints.rust]
 # Safety
@@ -468,7 +468,7 @@ semicolon_if_nothing_returned = "allow"  # Style preference
 
 [package]
 name = "aprender"
-version = "0.57.0"
+version = "0.58.0"
 edition = "2021"
 rust-version = "1.89"
 authors = ["Noah Gift <noah@paiml.com>"]
@@ -498,7 +498,7 @@ required-features = ["cli"]
 # (e.g. apr-cookbook, downstream apps) don't pull apr-cli's heavy transitive
 # dep graph (renacer/profile/inference/zram/visualization). `cargo install
 # aprender` still works because `cli` is in default features. See GH-1599.
-apr-cli = { path = "crates/apr-cli", version = "0.57.0", default-features = true, optional = true }
+apr-cli = { path = "crates/apr-cli", version = "0.58.0", default-features = true, optional = true }
 # Core ML library (lib name = "aprender")
 aprender_ml = { path = "crates/aprender-core", version = ">=0.29", package = "aprender-core" }
 
diff --git a/ROADMAP.md b/ROADMAP.md
index f9bf2aeb12..e505d4a7ab 100644
--- a/ROADMAP.md
+++ b/ROADMAP.md
@@ -34,7 +34,7 @@ canonical, issue-linked roadmap is
 | v0.42–v0.49 | ✅ Released | **Four-pillar BEAT campaign** — CI-gated falsifiable wins vs sklearn / PyTorch / Unsloth / Ollama (below) |
 | v0.50–v0.54 | ✅ Released | Correctness wave — sampling params, Blackwell-GPU coherence, Gemma CPU, APR-format safety, RoPE/LoRA/MCP, `apr-format` leaf extraction (each contract-backed) |
 | **v0.55–v0.57** | ✅ **Current** | **GPU QLoRA training actually works** — a 6-defect cascade fixed (deadlock → loss-window → cross-stream NaN → wrong-model → stale-adapter eval → CPU-only eval), each with a mutation-verified falsifier + contract; runnable `--merge`; rank-aware auto-lr (default no longer diverges) |
-| _in flight_ | 🔧 Hardening | CUDA CI lane so the GPU correctness falsifiers are actually enforced (see doctrine gap below); apr-code tool-call flip envelope measurement |
+| _in flight_ | 🔧 Hardening | apr-code tool-call flip envelope measurement; promote the cross-silicon CUDA nightly lane to a blocking training-path gate (now green — see below) |
 | _next_ | 📋 Planned | tracked in [`docs/roadmaps/roadmap.yaml`](docs/roadmaps/roadmap.yaml) |
 
 ### 🎯 The mission (north star)
@@ -48,7 +48,7 @@ wedge none of the four have: **provable, contract-gated correctness.**
 |--------|-----------|--------------------|----------------------------------|
 | **P1** | scikit-learn | Classical-ML breadth & ergonomics | LinearRegression **2.0× faster** (LAPACK-free O(nd)), iris-RF accuracy. *(LAPACK-bound Ridge/Lasso/KMeans/PCA honestly conceded.)* |
 | **P2** | PyTorch | Tensors + autograd + training | Autograd gradients **≡ PyTorch, max\|Δ\|=5e-7**. *(Training speed conceded — ~11× MKL gap; the win is provable gradient correctness.)* |
-| **P3** | Unsloth | Fast low-VRAM PEFT (LoRA/QLoRA) | NF4 quant **≡ bitsandbytes (4.9e-7)** + LoRA-merge forward-**equivalence (1.5e-8)**; **GPU QLoRA training runs correct end-to-end** (v0.55–v0.57: 6-defect cascade fixed, trains+validates at GPU speed on RTX 4090 & GB10 Blackwell). *(GPU-Triton tok/s conceded; training-correctness falsifiers are developer-run pending a CUDA CI lane.)* |
+| **P3** | Unsloth | Fast low-VRAM PEFT (LoRA/QLoRA) | NF4 quant **≡ bitsandbytes (4.9e-7)** + LoRA-merge forward-**equivalence (1.5e-8)**; **GPU QLoRA training runs correct end-to-end** (6-defect cascade fixed, trains+validates at GPU speed), **CI-enforced nightly on two silicons** (RTX 4090 sm_89 + GB10 Blackwell sm_121, v0.58). *(GPU-Triton tok/s conceded.)* |
 | **P4** | Ollama / llama.cpp | Fast local quantized inference | **Fail-closed correctness** — `apr` rejects 10/10 semantically-broken models that Ollama/llama.cpp silently run; decode **1.2–1.37× on RTX 4090**. |
 
 All four beats are **adversarially mutation-verified** — a deliberately injected
@@ -64,17 +64,23 @@ fix ships a named `proof_obligation` + a falsifier verified RED-on-bug / GREEN-o
 a `pv`-validated contract bump (a bug shipped green ⇔ its falsifier was missing or too
 weak).
 
-### ⚠️ Honest doctrine gap: GPU falsifiers are not yet CI-enforced
-
-The v0.55–v0.57 QLoRA training-correctness falsifiers (deadlock, cross-stream NaN,
-wrong-model rope/bias, stale-adapter eval, GPU-vs-CPU eval-forward) require a CUDA
-device and are `#[ignore]`-gated — the CI fleet has **no CUDA-labeled runner**, so
-today they only run developer-side on the pre-authorized RTX 4090. By the "gates or
-theater" rule this is a real hole: a correctness gate that never runs unattended is
-weaker than one that does. **The highest-EV hardening item is a nightly CUDA lane on
-the 4090** that executes these falsifiers as a scheduled (eventually blocking) check.
-Until it lands, these wins are stated as *developer-verified*, not *CI-enforced* — the
-same discipline §6 applies to conceded speed.
+### ✅ Doctrine gap CLOSED (2026-07-03): GPU falsifiers are now CI-enforced on two silicons
+
+*Was a named gap in v0.57; closed in v0.58.* The v0.55–v0.57 QLoRA training-correctness
+falsifiers (deadlock, cross-stream NaN, wrong-model rope/bias, stale-adapter eval,
+GPU-vs-CPU eval-forward) require a CUDA device and are `#[ignore]`-gated, so they used to
+run developer-side only. As of v0.58 they run every night on a **cross-silicon
+self-hosted matrix** (`.github/workflows/cuda-nightly.yml`, `CUDA-CI-NIGHTLY-001`):
+
+- **ada-4090** — RTX 4090, sm_89, x86-64, CUDA 12.8 (reference platform)
+- **blackwell-gb10** — GB10, sm_121, aarch64, CUDA 13.0 (higher-risk JIT target; egress
+  via a subnet-restricted forward proxy on the wired link, no routing change on the box)
+
+Both legs went **green end-to-end** (checkout → cargo build → all six falsifiers +
+loss-window) on 2026-07-03. The lane is nightly (01:30 UTC ≈ 03:30 Madrid), yields to
+overnight training (GPU-busy check), and never triggers on a PR. These wins are now
+stated as **CI-enforced (nightly, 2 silicons)**, not developer-verified. Next step:
+promote to a blocking gate on training-path PRs once it has a green track record.
 
 **Case-file lesson (CF-5, 2026-07-03):** the GPU eval-forward falsifier's tolerance band
 *fired against its own CPU reference* (Δ=12.29 ≫ 0.5). Rather than loosen the band, the
diff --git a/crates/apr-cli/Cargo.toml b/crates/apr-cli/Cargo.toml
index ac90a13e73..211647b1ed 100644
--- a/crates/apr-cli/Cargo.toml
+++ b/crates/apr-cli/Cargo.toml
@@ -103,12 +103,12 @@ libc = "0.2"
 
 [dependencies]
 # Core aprender library
-aprender = { path = "../aprender-core", version = "0.57.0", package = "aprender-core", default-features = true, features = ["format-compression"] }
+aprender = { path = "../aprender-core", version = "0.58.0", package = "aprender-core", default-features = true, features = ["format-compression"] }
 # Provable-contracts engine — `apr beat-run` parses + evaluates BeatBenchmark contracts (PMAT-741)
 aprender-contracts = { workspace = true }
 
 # MCP (Model Context Protocol) server — `apr mcp` subcommand
-aprender-mcp = { path = "../aprender-mcp", version = "0.57.0" }
+aprender-mcp = { path = "../aprender-mcp", version = "0.58.0" }
 async-stream = "0.3"
 provable-contracts-macros = "0.3"
 
@@ -119,7 +119,7 @@ batuta-common = { workspace = true }
 # Non-optional since claude-code-parity-apr M150 (outcome-parity Phase 3
 # requires apr code in default builds). GH-344: path dep via
 # .cargo/config.toml.dev-overrides; CI uses checkout+symlink.
-batuta = { version = "0.57.0", path = "../aprender-orchestrate", package = "aprender-orchestrate", default-features = false, features = ["agents", "agents-inference", "agents-mcp", "rag"] }
+batuta = { version = "0.58.0", path = "../aprender-orchestrate", package = "aprender-orchestrate", default-features = false, features = ["agents", "agents-inference", "agents-mcp", "rag"] }
 
 # Inference engine (workspace path dep — enables cuda/gpu feature forwarding)
 # GH-573: shim (0.9.1) didn't forward cuda feature → 0.7 tok/s instead of 273 tok/s
diff --git a/crates/aprender-bench-compute/Cargo.toml b/crates/aprender-bench-compute/Cargo.toml
index 689d4295be..11985daccf 100644
--- a/crates/aprender-bench-compute/Cargo.toml
+++ b/crates/aprender-bench-compute/Cargo.toml
@@ -11,7 +11,7 @@ disallowed_methods = "allow"
 
 [dependencies]
 # The implementation under test
-aprender = { path = "../aprender-core", version = "0.57.0", package = "aprender-core", features = ["format-quantize"] }
+aprender = { path = "../aprender-core", version = "0.58.0", package = "aprender-core", features = ["format-quantize"] }
 
 # Reference implementation for A/B comparison
 # Pure Rust (no BLAS) to keep comparison fair — both are pure Rust SIMD
diff --git a/crates/aprender-bench-tokenizer/Cargo.toml b/crates/aprender-bench-tokenizer/Cargo.toml
index 4ec8ba7550..c5fe152432 100644
--- a/crates/aprender-bench-tokenizer/Cargo.toml
+++ b/crates/aprender-bench-tokenizer/Cargo.toml
@@ -11,7 +11,7 @@ disallowed_methods = "allow"
 
 [dependencies]
 # The implementation under test
-aprender = { path = "../aprender-core", version = "0.57.0", package = "aprender-core" }
+aprender = { path = "../aprender-core", version = "0.58.0", package = "aprender-core" }
 
 # HuggingFace reference for A/B comparison
 # Regular dep (not dev-dep) because [[bench]] harness=false binaries can't see dev-deps
diff --git a/crates/aprender-cbtop/Cargo.toml b/crates/aprender-cbtop/Cargo.toml
index 43c700c6ca..a1ef545ec4 100644
--- a/crates/aprender-cbtop/Cargo.toml
+++ b/crates/aprender-cbtop/Cargo.toml
@@ -24,7 +24,7 @@ presentar-terminal = { workspace = true }
 
 # Compute backends
 trueno = { workspace = true }
-trueno-gpu = { version = "0.57.0", path = "../aprender-gpu", package = "aprender-gpu", optional = true }
+trueno-gpu = { version = "0.58.0", path = "../aprender-gpu", package = "aprender-gpu", optional = true }
 
 # Terminal handling
 crossterm = "0.28"
diff --git a/crates/aprender-cgp/Cargo.toml b/crates/aprender-cgp/Cargo.toml
index 326005a5b3..f73a587449 100644
--- a/crates/aprender-cgp/Cargo.toml
+++ b/crates/aprender-cgp/Cargo.toml
@@ -36,7 +36,7 @@ num_cpus = "1.17"
 comfy-table = "7"
 colored = "3"
 # GPU profiling: direct cuBLAS measurement (OWN THE STACK)
-trueno-gpu = { path = "../aprender-gpu", version = "0.57.0", features = ["cuda"], optional = true, package = "aprender-gpu" }
+trueno-gpu = { path = "../aprender-gpu", version = "0.58.0", features = ["cuda"], optional = true, package = "aprender-gpu" }
 
 [features]
 default = []
diff --git a/crates/aprender-compute/Cargo.toml b/crates/aprender-compute/Cargo.toml
index 1e3738c02b..cc2ff393fd 100644
--- a/crates/aprender-compute/Cargo.toml
+++ b/crates/aprender-compute/Cargo.toml
@@ -41,11 +41,11 @@ futures-intrusive = { version = "0.5", optional = true }
 wasm-bindgen-futures = { version = "0.4", optional = true }
 wasm-bindgen = { version = "0.2", optional = true }
 # Native CUDA monitoring via trueno-gpu (TRUENO-SPEC-010)
-trueno-gpu = { version = "0.57.0", path = "../aprender-gpu", optional = true, package = "aprender-gpu" }
+trueno-gpu = { version = "0.58.0", path = "../aprender-gpu", optional = true, package = "aprender-gpu" }
 # PMAT-336: Provable contracts compile-time enforcement
 provable-contracts-macros = "0.3"
 # P1a: Sovereign GEMM microkernel codegen (compile-time shape specialization)
-trueno-gemm-codegen = { version = "0.57.0", path = "../aprender-gemm-codegen", package = "aprender-gemm-codegen" }
+trueno-gemm-codegen = { version = "0.58.0", path = "../aprender-gemm-codegen", package = "aprender-gemm-codegen" }
 # Tracing/profiling support (renacer integration)
 tracing = { version = "0.1", optional = true }
 tracing-subscriber = { version = "0.3", features = ["env-filter", "fmt"], optional = true }
diff --git a/crates/aprender-contracts-cli/Cargo.toml b/crates/aprender-contracts-cli/Cargo.toml
index df3cc38e96..128b4fcfbc 100644
--- a/crates/aprender-contracts-cli/Cargo.toml
+++ b/crates/aprender-contracts-cli/Cargo.toml
@@ -13,7 +13,7 @@ name = "pv"
 path = "src/main.rs"
 
 [dependencies]
-aprender-contracts = { path = "../aprender-contracts", version = "0.57.0" }
+aprender-contracts = { path = "../aprender-contracts", version = "0.58.0" }
 clap = { workspace = true }
 serde_json = { workspace = true }
 serde_yaml = { workspace = true }
diff --git a/crates/aprender-contracts/Cargo.toml b/crates/aprender-contracts/Cargo.toml
index fb9edc8127..f0b92c202e 100644
--- a/crates/aprender-contracts/Cargo.toml
+++ b/crates/aprender-contracts/Cargo.toml
@@ -19,7 +19,7 @@ kani = []
 name = "provable_contracts"
 
 [dependencies]
-provable-contracts-macros = { path = "../aprender-contracts-macros", version = "0.57.0", package = "aprender-contracts-macros" }
+provable-contracts-macros = { path = "../aprender-contracts-macros", version = "0.58.0", package = "aprender-contracts-macros" }
 regex = "1"
 serde = { workspace = true }
 serde_json = { workspace = true }
diff --git a/crates/aprender-core/Cargo.toml b/crates/aprender-core/Cargo.toml
index b14a62c6f9..69b8f33660 100644
--- a/crates/aprender-core/Cargo.toml
+++ b/crates/aprender-core/Cargo.toml
@@ -96,7 +96,7 @@ trueno-quant = { workspace = true }
 # APR-2231: sovereign `.apr` container leaf. aprender-core From-wraps its
 # AprFormatError into AprenderError and re-exports it (no API break). Leaf has
 # no dependency back on core, so this introduces no publish cycle.
-apr-format = { path = "../apr-format", version = "0.57" }
+apr-format = { path = "../apr-format", version = "0.58" }
 
 # NOTE: trueno-rag (=aprender-rag) removed (APR-MONO self-containment): aprender-rag
 # depends on aprender-core + aprender-serve, so core→rag closed a core→rag→serve→core
diff --git a/crates/aprender-cuda-edge/Cargo.toml b/crates/aprender-cuda-edge/Cargo.toml
index 0d9efd6a65..a446fe176a 100644
--- a/crates/aprender-cuda-edge/Cargo.toml
+++ b/crates/aprender-cuda-edge/Cargo.toml
@@ -15,7 +15,7 @@ name = "trueno_cuda_edge"
 [dependencies]
 thiserror = "2.0"
 serde = { version = "1.0", features = ["derive"] }
-trueno-gpu = { version = "0.57.0", path = "../aprender-gpu", package = "aprender-gpu", optional = true }
+trueno-gpu = { version = "0.58.0", path = "../aprender-gpu", package = "aprender-gpu", optional = true }
 
 [dev-dependencies]
 proptest = "1.6"
diff --git a/crates/aprender-explain/Cargo.toml b/crates/aprender-explain/Cargo.toml
index 3cbf9f3ab0..6497d37923 100644
--- a/crates/aprender-explain/Cargo.toml
+++ b/crates/aprender-explain/Cargo.toml
@@ -25,7 +25,7 @@ presentar-terminal = { workspace = true }
 crossterm = "0.28"
 
 # Internal dependency for PTX generation
-trueno-gpu = { version = "0.57.0", path = "../aprender-gpu", package = "aprender-gpu", features = ["cuda"] }
+trueno-gpu = { version = "0.58.0", path = "../aprender-gpu", package = "aprender-gpu", features = ["cuda"] }
 
 [dev-dependencies]
 proptest = "1.4"
diff --git a/crates/aprender-graph/Cargo.toml b/crates/aprender-graph/Cargo.toml
index e50645f821..2e283ec45c 100644
--- a/crates/aprender-graph/Cargo.toml
+++ b/crates/aprender-graph/Cargo.toml
@@ -28,7 +28,7 @@ anyhow = "1"
 thiserror = "2"
 
 # Phase 2+ dependencies
-aprender = { path = "../aprender-core", version = "0.57.0", package = "aprender-core" }                                        # Sparse matrix ops, PageRank
+aprender = { path = "../aprender-core", version = "0.58.0", package = "aprender-core" }                                        # Sparse matrix ops, PageRank
 trueno = { workspace = true }
 trueno-db = { workspace = true }
 
diff --git a/crates/aprender-monte-carlo/Cargo.toml b/crates/aprender-monte-carlo/Cargo.toml
index 0dbd5492f5..6c4aa2316a 100644
--- a/crates/aprender-monte-carlo/Cargo.toml
+++ b/crates/aprender-monte-carlo/Cargo.toml
@@ -18,7 +18,7 @@ path = "src/lib.rs"
 name = "aprender-monte-carlo"
 
 [dependencies]
-aprender = { path = "../../", version = "0.57.0" }
+aprender = { path = "../../", version = "0.58.0" }
 clap = { version = "4", features = ["derive"] }
 rand = "0.9"
 rand_chacha = "0.9"
diff --git a/crates/aprender-orchestrate/Cargo.toml b/crates/aprender-orchestrate/Cargo.toml
index 534fef9898..118e2812f6 100644
--- a/crates/aprender-orchestrate/Cargo.toml
+++ b/crates/aprender-orchestrate/Cargo.toml
@@ -120,7 +120,7 @@ pacha = { workspace = true, optional = true }
 realizar = { workspace = true, optional = true }
 
 # ML algorithms and training (native-only)
-aprender = { path = "../aprender-core", version = "0.57.0", package = "aprender-core", optional = true }
+aprender = { path = "../aprender-core", version = "0.58.0", package = "aprender-core", optional = true }
 entrenar = { workspace = true, optional = true }
 alimentar = { workspace = true, optional = true }
 
diff --git a/crates/aprender-profile/Cargo.toml b/crates/aprender-profile/Cargo.toml
index 8cfef2e572..abc7e3ff71 100644
--- a/crates/aprender-profile/Cargo.toml
+++ b/crates/aprender-profile/Cargo.toml
@@ -74,7 +74,7 @@ trueno = { workspace = true }
 
 # Machine learning for anomaly detection (Sprint 23)
 # GH-581: Use path dep to avoid trueno version skew with local workspace
-aprender = { path = "../aprender-core", version = "0.57.0", package = "aprender-core" }
+aprender = { path = "../aprender-core", version = "0.58.0", package = "aprender-core" }
 
 # Lock-free data structures for ring buffer (Sprint 40)
 crossbeam = { version = "0.8", default-features = false, features = ["crossbeam-channel", "crossbeam-queue", "std"] }
@@ -112,7 +112,7 @@ crossterm = "0.28"
 
 # SIMD-accelerated visualization primitives (sister project)
 # NOTE: `monitor` feature removed upstream (#1979); this crate uses only viz primitives.
-trueno-viz = { version = "0.57.0", path = "../aprender-viz", package = "aprender-viz", default-features = false }
+trueno-viz = { version = "0.58.0", path = "../aprender-viz", package = "aprender-viz", default-features = false }
 
 # OpenTelemetry for distributed tracing (Sprint 30)
 opentelemetry = { version = "0.31.0", optional = true }
diff --git a/crates/aprender-qa-cli/Cargo.toml b/crates/aprender-qa-cli/Cargo.toml
index ef81073875..1ccb646d28 100644
--- a/crates/aprender-qa-cli/Cargo.toml
+++ b/crates/aprender-qa-cli/Cargo.toml
@@ -23,10 +23,10 @@ serde_json = { workspace = true }
 serde_yaml = { workspace = true }
 clap = { workspace = true }
 chrono = { workspace = true }
-aprender-qa-gen = { version = "0.57.0", path = "../aprender-qa-gen" }
-aprender-qa-runner = { version = "0.57.0", path = "../aprender-qa-runner" }
-aprender-qa-report = { version = "0.57.0", path = "../aprender-qa-report" }
-aprender-qa-certify = { version = "0.57.0", path = "../aprender-qa-certify" }
+aprender-qa-gen = { version = "0.58.0", path = "../aprender-qa-gen" }
+aprender-qa-runner = { version = "0.58.0", path = "../aprender-qa-runner" }
+aprender-qa-report = { version = "0.58.0", path = "../aprender-qa-report" }
+aprender-qa-certify = { version = "0.58.0", path = "../aprender-qa-certify" }
 ctrlc = "3.4"
 safetensors = { workspace = true }
 colored = "2.2"
diff --git a/crates/aprender-qa-report/Cargo.toml b/crates/aprender-qa-report/Cargo.toml
index 2d5fc30121..0a099eec59 100644
--- a/crates/aprender-qa-report/Cargo.toml
+++ b/crates/aprender-qa-report/Cargo.toml
@@ -18,8 +18,8 @@ anyhow = { workspace = true }
 chrono = { workspace = true }
 colored = "2.2"
 csv = "1.3"
-aprender-qa-gen = { version = "0.57.0", path = "../aprender-qa-gen" }
-aprender-qa-runner = { version = "0.57.0", path = "../aprender-qa-runner" }
+aprender-qa-gen = { version = "0.58.0", path = "../aprender-qa-gen" }
+aprender-qa-runner = { version = "0.58.0", path = "../aprender-qa-runner" }
 
 [dev-dependencies]
 proptest = { workspace = true }
diff --git a/crates/aprender-qa-runner/Cargo.toml b/crates/aprender-qa-runner/Cargo.toml
index f32f6ba589..4e5c68c2b5 100644
--- a/crates/aprender-qa-runner/Cargo.toml
+++ b/crates/aprender-qa-runner/Cargo.toml
@@ -22,14 +22,14 @@ anyhow = { workspace = true }
 tokio = { workspace = true }
 chrono = { workspace = true, features = ["serde"] }
 tracing = { workspace = true }
-aprender-qa-gen = { version = "0.57.0", path = "../aprender-qa-gen" }
+aprender-qa-gen = { version = "0.58.0", path = "../aprender-qa-gen" }
 hostname = "0.4"
 rayon = "1.10"
 sha2 = "0.10"
 regex = { workspace = true }
 num_cpus = "1.16"
 safetensors = { workspace = true }
-jugar-probar = { version = "0.57.0", path = "../aprender-test-lib", package = "aprender-test-lib", features = ["llm-types"] }
+jugar-probar = { version = "0.58.0", path = "../aprender-test-lib", package = "aprender-test-lib", features = ["llm-types"] }
 # Test fixtures dependencies (optional)
 tempfile = { version = "3", optional = true }
 half = { version = "2.4", optional = true }
diff --git a/crates/aprender-rag/Cargo.toml b/crates/aprender-rag/Cargo.toml
index 57860a05ab..e14c1d60d1 100644
--- a/crates/aprender-rag/Cargo.toml
+++ b/crates/aprender-rag/Cargo.toml
@@ -64,7 +64,7 @@ fastembed = { version = "5", optional = true }
 
 # NVIDIA Embed Nemotron 8B via realizar GGUF infrastructure (GH-3)
 # Large model (~4-32GB depending on quantization), requires GGUF file
-realizar = { version = "0.57.0", path = "../aprender-serve", package = "aprender-serve", default-features = false, optional = true }
+realizar = { version = "0.58.0", path = "../aprender-serve", package = "aprender-serve", default-features = false, optional = true }
 
 # Speech-to-text via whisper-apr (Whisper ASR, .apr model format)
 # Sovereign stack: whisper-apr + aprender + trueno for pure-Rust transcription
diff --git a/crates/aprender-registry/Cargo.toml b/crates/aprender-registry/Cargo.toml
index 0770f0bcb1..0443454b8b 100644
--- a/crates/aprender-registry/Cargo.toml
+++ b/crates/aprender-registry/Cargo.toml
@@ -42,7 +42,7 @@ zstd = { version = "0.13", optional = true }
 clap = { version = "4", features = ["derive"], optional = true }
 
 # Ecosystem integration
-aprender = { path = "../aprender-core", version = "0.57.0", package = "aprender-core", optional = true }
+aprender = { path = "../aprender-core", version = "0.58.0", package = "aprender-core", optional = true }
 alimentar = { workspace = true, optional = true }
 
 # HTTP client (optional)
diff --git a/crates/aprender-serve/Cargo.toml b/crates/aprender-serve/Cargo.toml
index 093810a78e..cdc7c1bcfa 100644
--- a/crates/aprender-serve/Cargo.toml
+++ b/crates/aprender-serve/Cargo.toml
@@ -47,7 +47,7 @@ name = "realizar"
 trueno = { workspace = true, features = ["gpu"] }
 # CUDA PTX generation and runtime for NVIDIA GPUs (optional, pure Rust, no LLVM/nvcc)
 # v0.4.12: BatchedQ6K, MultiWarp, RopeNeox, FlashDecoding kernels
-trueno-gpu = { version = "0.57.0", path = "../aprender-gpu", package = "aprender-gpu", optional = true, features = ["cuda"] }  # APR-MONO single-source-of-truth
+trueno-gpu = { version = "0.58.0", path = "../aprender-gpu", package = "aprender-gpu", optional = true, features = ["cuda"] }  # APR-MONO single-source-of-truth
 # KV cache storage (for attention caching)
 trueno-db = { workspace = true, optional = true }
 # K-quantization formats (Q4_K, Q5_K, Q6_K) - Toyota Way: ONE source of truth
@@ -61,7 +61,7 @@ renacer-core = { workspace = true }
 
 # ML algorithms and .apr model format (optional for aprender model serving)
 # v0.24.1: chat templates, GGUF tokenizer preservation (local may be newer)
-aprender = { path = "../aprender-core", version = "0.57.0", package = "aprender-core", optional = true }
+aprender = { path = "../aprender-core", version = "0.58.0", package = "aprender-core", optional = true }
 
 # Data loading library (optional for data pipeline examples)
 alimentar = { workspace = true, optional = true }
diff --git a/crates/aprender-shell/Cargo.toml b/crates/aprender-shell/Cargo.toml
index e8e20fe40c..b13f5563a2 100644
--- a/crates/aprender-shell/Cargo.toml
+++ b/crates/aprender-shell/Cargo.toml
@@ -24,7 +24,7 @@ harness = false
 
 [dependencies]
 # Use path for local dev, crates.io for release
-aprender = { path = "../aprender-core", version = "0.57.0", package = "aprender-core", features = ["format-encryption", "format-compression", "format-homomorphic"] }
+aprender = { path = "../aprender-core", version = "0.58.0", package = "aprender-core", features = ["format-encryption", "format-compression", "format-homomorphic"] }
 clap = { version = "4", features = ["derive"] }
 dirs = "5"
 serde = { version = "1", features = ["derive"] }
diff --git a/crates/aprender-simulate/Cargo.toml b/crates/aprender-simulate/Cargo.toml
index d8a5de7ffc..f82e7971a5 100644
--- a/crates/aprender-simulate/Cargo.toml
+++ b/crates/aprender-simulate/Cargo.toml
@@ -53,7 +53,7 @@ crossterm = { version = "0.28", optional = true }
 # ComputeBlocks from presentar-terminal (SIMD-optimized TUI widgets)
 # Path deps to avoid jugar-probar prod dep chain (CB-081)
 presentar-terminal = { workspace = true, optional = true }
-presentar-core = { version = "0.57.0", path = "../aprender-present-core", package = "aprender-present-core", default-features = false, optional = true }
+presentar-core = { version = "0.58.0", path = "../aprender-present-core", package = "aprender-present-core", default-features = false, optional = true }
 
 # WASM (optional)
 wasm-bindgen = { version = "0.2", optional = true }
diff --git a/crates/aprender-test-cli/Cargo.toml b/crates/aprender-test-cli/Cargo.toml
index c7c909b015..140bc0ba9c 100644
--- a/crates/aprender-test-cli/Cargo.toml
+++ b/crates/aprender-test-cli/Cargo.toml
@@ -24,7 +24,7 @@ path = "src/main.rs"
 
 [dependencies]
 # Core library (was: jugar-probar) — dep key preserves Rust identifier
-jugar-probar = { path = "../aprender-test-lib", version = "0.57.0", package = "aprender-test-lib" }
+jugar-probar = { path = "../aprender-test-lib", version = "0.58.0", package = "aprender-test-lib" }
 
 # CLI framework
 clap = { workspace = true }
diff --git a/crates/aprender-train-bench/Cargo.toml b/crates/aprender-train-bench/Cargo.toml
index 123c0c2f3d..8f0ff6500e 100644
--- a/crates/aprender-train-bench/Cargo.toml
+++ b/crates/aprender-train-bench/Cargo.toml
@@ -20,7 +20,7 @@ workspace = true
 
 [dependencies]
 entrenar = { workspace = true }
-entrenar-common = { version = "0.57.0", path = "../aprender-train-common", package = "aprender-train-common" }
+entrenar-common = { version = "0.58.0", path = "../aprender-train-common", package = "aprender-train-common" }
 clap = { version = "4.5", features = ["derive"] }
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
diff --git a/crates/aprender-train-distill/Cargo.toml b/crates/aprender-train-distill/Cargo.toml
index bed58dc287..295c615860 100644
--- a/crates/aprender-train-distill/Cargo.toml
+++ b/crates/aprender-train-distill/Cargo.toml
@@ -33,7 +33,7 @@ shard-batch-source = []
 
 [dependencies]
 entrenar = { workspace = true }
-entrenar-common = { version = "0.57.0", path = "../aprender-train-common", package = "aprender-train-common" }
+entrenar-common = { version = "0.58.0", path = "../aprender-train-common", package = "aprender-train-common" }
 clap = { version = "4.5", features = ["derive"] }
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
diff --git a/crates/aprender-train-inspect/Cargo.toml b/crates/aprender-train-inspect/Cargo.toml
index 6df1229bc2..35c9780770 100644
--- a/crates/aprender-train-inspect/Cargo.toml
+++ b/crates/aprender-train-inspect/Cargo.toml
@@ -20,7 +20,7 @@ workspace = true
 
 [dependencies]
 entrenar = { workspace = true }
-entrenar-common = { version = "0.57.0", path = "../aprender-train-common", package = "aprender-train-common" }
+entrenar-common = { version = "0.58.0", path = "../aprender-train-common", package = "aprender-train-common" }
 clap = { version = "4.5", features = ["derive"] }
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
diff --git a/crates/aprender-train-lora/Cargo.toml b/crates/aprender-train-lora/Cargo.toml
index e013f545f8..e9a2c8f08f 100644
--- a/crates/aprender-train-lora/Cargo.toml
+++ b/crates/aprender-train-lora/Cargo.toml
@@ -20,7 +20,7 @@ workspace = true
 
 [dependencies]
 entrenar = { workspace = true }
-entrenar-common = { version = "0.57.0", path = "../aprender-train-common", package = "aprender-train-common" }
+entrenar-common = { version = "0.58.0", path = "../aprender-train-common", package = "aprender-train-common" }
 clap = { version = "4.5", features = ["derive"] }
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
diff --git a/crates/aprender-train-shell/Cargo.toml b/crates/aprender-train-shell/Cargo.toml
index 7f15fbde12..f0d7bed703 100644
--- a/crates/aprender-train-shell/Cargo.toml
+++ b/crates/aprender-train-shell/Cargo.toml
@@ -20,7 +20,7 @@ workspace = true
 
 [dependencies]
 entrenar = { workspace = true }
-entrenar-common = { version = "0.57.0", path = "../aprender-train-common", package = "aprender-train-common" }
+entrenar-common = { version = "0.58.0", path = "../aprender-train-common", package = "aprender-train-common" }
 clap = { version = "4.5", features = ["derive"] }
 rustyline = "14"
 serde = { version = "1.0", features = ["derive"] }
diff --git a/crates/aprender-train/Cargo.toml b/crates/aprender-train/Cargo.toml
index 963bc21348..67495c5f40 100644
--- a/crates/aprender-train/Cargo.toml
+++ b/crates/aprender-train/Cargo.toml
@@ -78,9 +78,9 @@ provable-contracts-macros = { version = "0.2" }
 # Core PAIML stack
 ndarray = "0.16"
 trueno = { workspace = true, features = ["parallel"] }
-trueno-gpu = { version = "0.57.0", path = "../aprender-gpu", package = "aprender-gpu", optional = true }  # CUDA compute (APR-MONO single-source-of-truth)
+trueno-gpu = { version = "0.58.0", path = "../aprender-gpu", package = "aprender-gpu", optional = true }  # CUDA compute (APR-MONO single-source-of-truth)
 realizar = { workspace = true, optional = true }
-aprender = { path = "../aprender-core", version = "0.57.0", package = "aprender-core" }  # ML interpret module + .apr format + tokenizers
+aprender = { path = "../aprender-core", version = "0.58.0", package = "aprender-core" }  # ML interpret module + .apr format + tokenizers
 trueno-viz = { workspace = true, features = ["terminal"] }
 batuta-common = { workspace = true, optional = true }  # WithDimensions trait for `viz` feature (APR-MONO §S #1978)
 presentar-terminal = { workspace = true, optional = true }
diff --git a/crates/aprender-tsp/Cargo.toml b/crates/aprender-tsp/Cargo.toml
index ef4fbec3d4..3a222f809b 100644
--- a/crates/aprender-tsp/Cargo.toml
+++ b/crates/aprender-tsp/Cargo.toml
@@ -18,7 +18,7 @@ path = "src/lib.rs"
 name = "aprender-tsp"
 
 [dependencies]
-aprender = { path = "../../", version = "0.57.0" }
+aprender = { path = "../../", version = "0.58.0" }
 clap = { version = "4", features = ["derive"] }
 rand = "0.9"
 crc32fast = "1.4"
diff --git a/crates/aprender-verify-ml/Cargo.toml b/crates/aprender-verify-ml/Cargo.toml
index 081589b27e..0f359d6f5b 100644
--- a/crates/aprender-verify-ml/Cargo.toml
+++ b/crates/aprender-verify-ml/Cargo.toml
@@ -50,7 +50,7 @@ uuid = { version = "1.11", features = ["v4", "serde"] }
 trueno = { workspace = true }
 
 # Machine learning for bug prediction (optional)
-aprender = { path = "../aprender-core", version = "0.57.0", package = "aprender-core", optional = true }
+aprender = { path = "../aprender-core", version = "0.58.0", package = "aprender-core", optional = true }
 
 # LLM fine-tuning integration (optional)
 entrenar = { workspace = true, optional = true }
diff --git a/crates/aprender-viz/Cargo.toml b/crates/aprender-viz/Cargo.toml
index 2dba9d8712..796644f77e 100644
--- a/crates/aprender-viz/Cargo.toml
+++ b/crates/aprender-viz/Cargo.toml
@@ -14,7 +14,7 @@ categories = ["visualization", "graphics", "science", "wasm"]
 
 [dependencies]
 # Core SIMD/GPU acceleration
-trueno = { version = "0.57.0", path = "../aprender-compute", package = "aprender-compute", default-features = false }
+trueno = { version = "0.58.0", path = "../aprender-compute", package = "aprender-compute", default-features = false }
 
 # Shared Batuta stack utilities (display traits, formatting)
 batuta-common = { workspace = true }
@@ -29,10 +29,10 @@ base64 = "0.22"
 thiserror = "2.0"
 
 # Optional: ML integration
-aprender = { path = "../aprender-core", version = "0.57.0", package = "aprender-core", optional = true }
+aprender = { path = "../aprender-core", version = "0.58.0", package = "aprender-core", optional = true }
 
 # Optional: Graph integration
-trueno-graph = { version = "0.57.0", path = "../aprender-graph", package = "aprender-graph", default-features = false, optional = true }
+trueno-graph = { version = "0.58.0", path = "../aprender-graph", package = "aprender-graph", default-features = false, optional = true }
 
 # Optional: Database integration
 trueno-db = { workspace = true, optional = true }
diff --git a/docs/roadmaps/roadmap.yaml b/docs/roadmaps/roadmap.yaml
index 1f1af304cb..2cf5c6ca80 100644
--- a/docs/roadmaps/roadmap.yaml
+++ b/docs/roadmaps/roadmap.yaml
@@ -56,7 +56,7 @@ roadmap:
   github_issue: null
   item_type: task
   title: 'CUDA nightly CI lane — enforce the GPU QLoRA training-correctness falsifiers'
-  status: pending
+  status: completed
   priority: high
   assigned_to: null
   created: 2026-07-03T00:00:00.000000000+00:00