diff --git a/CHANGELOG.md b/CHANGELOG.md
index 0089906..96540fa 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,6 +7,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Added
+
+- **`nn.SaveTo`/`nn.LoadFrom`** — streaming Save/Load over `io.Writer`/`io.ReadSeeker` for HTTP responses, network connections, and in-memory buffers without touching disk. `nn.SaveToBytes` added as in-memory convenience. `LoadFromBytes` refactored as thin wrapper over `LoadFrom`. Serializer writer inverted onto `io.Writer`; `WriteTo` deduplication (~90 lines removed). Statement coverage 69.6% → 81.8% ([#133](https://github.com/born-ml/born/pull/133) by [@amery](https://github.com/amery))
+
 ## [0.9.17] - 2026-07-12
 
 ### Added
diff --git a/README.md b/README.md
index ac4300d..44879ec 100644
--- a/README.md
+++ b/README.md
@@ -86,7 +86,7 @@ prediction := model.Predict(image)
 - **GGUF Import** - llama.cpp format with K-quant dequantization (Q4_K, Q5_K, Q6_K, Q8_0)
 - **LLaMA** - `models/llama.LoadGGUF()` for end-to-end LLaMA inference; verified on TinyLlama 1.1B Q8_0 and Q4_K_M
 - **Injectable Attention** - swap attention implementation at model load time for research experiments
-- **Native Format** - `.born` format with `nn.Save()` / `nn.Load()` / `nn.LoadFromBytes()` (`go:embed` support)
+- **Native Format** - `.born` format with `nn.Save()` / `nn.Load()` / `nn.SaveTo()` / `nn.LoadFrom()` (file, stream, or `[]byte`)
 - **Checkpoints** - Resume training with optimizer state preservation
 - **SafeTensors** - HuggingFace compatible export
 - **Reproducibility** - `nn.SetSeed()` for deterministic weight initialization