Who contributed to Google on May 20, 2026?

3 developers shipped this update, including copybara-service[bot], chalmerlowe, and sakthivelmanii.

What were the notable Google updates?

Add philox4x32 PRNG impl, Add threefry4x32 PRNG impl, and [XLA:GPU] Fold should_autotune logic into the native & block-level emitters.

Google

JAX, the GenAI SDK, and the Cloud libs - Google's open source layer

Pick a date

Topics: AI / ML Python Full archive →

The Wire · Showcase

JAX ADDS STRONGER RANDOM NUMBER GENERATORS WHILE CLOUD PYTHON CUTS LEGACY SUPPORT

By RepoJournal · Filed 06:03 UTC on May 20, 2026 · About Google

3 people shipped this

copybara-service[bot] @copybara-service[bot] 4 cited

chalmerlowe @chalmerlowe 3 cited

sakthivelmanii @sakthivelmanii 1 cited

JAX shipped two new PRNG implementations with larger key and counter spaces, giving you better statistical properties for large-scale sampling - right as the Google Cloud Python suite drops Python 3.7 through 3.9 entirely.

The JAX team added philox4x32 and threefry4x32 PRNG implementations [1] [2], both offering 128-bit counter spaces compared to their smaller predecessors. Philox4x32 uses 64-bit keys with operations that accelerators can lower more efficiently, while threefry4x32 pushes to 128-bit keys for contexts generating many samples. These hit `jax.random.key` directly with the `impl` parameter, so you can switch without changing calling code. On the GPU side, XLA folded autotuning logic into native and block-level emitters [3], tightening compilation and reducing redundant kernel selection. Meanwhile, Google Cloud Python is officially ending support for Python 3.7, 3.8, and 3.9 [4] [5] [8], updating lower bounds across Storage and Spanner with 3.10 as the new floor. The Storage team also optimized zonal system tests for concurrency and speed [6], while Spanner fixed idempotent database restore retries by catching AlreadyExists and validating backup state [7]. Python-genai added Gemini 3.5 Flash to model options [9], ready for immediate use.

Action items

→ Verify your JAX random seeding works with new philox4x32 or threefry4x32 if you generate high-volume samples google/jax [plan]
→ Pin Python 3.10+ for google-cloud-storage and google-cloud-spanner before next dependency update googleapis/google-cloud-python [plan]
→ Update any restore database tests in Spanner to handle AlreadyExists gracefully on retry googleapis/google-cloud-python [monitor]

References

[1] Add philox4x32 PRNG impl ↗ google/jax
[2] Add threefry4x32 PRNG impl ↗ google/jax
[3] [XLA:GPU] Fold should_autotune logic into the native & block-level emitters. ↗ google/jax
[4] feat(storage): drop Python 3.7-3.9 support and regenerate ↗ googleapis/google-cloud-python
[5] feat(spanner): drop Python 3.7-3.9 support and regenerate ↗ googleapis/google-cloud-python
[6] chore(storage): optimize zonal system tests CloudBuild and make concurrency-safe (#17171) googleapis/google-cloud-python
[7] test(spanner): handle AlreadyExists idempotently on restore test retries ↗ googleapis/google-cloud-python
[8] feat: regenerate asset w/ dependencies using generator v1.32.0 ↗ googleapis/google-cloud-python
[9] feat: Add Gemini 3.5 Flash model to options ↗ googleapis/python-genai

Quick answers

What shipped in Google on May 20, 2026?: JAX shipped two new PRNG implementations with larger key and counter spaces, giving you better statistical properties for large-scale sampling - right as the Google Cloud Python suite drops Python 3.7 through 3.9 entirely. In total, 44 commits and 37 pull requests landed.
Who contributed to Google on May 20, 2026?: 3 developers shipped this update, including copybara-service[bot], chalmerlowe, and sakthivelmanii.
What were the notable Google updates?: Add philox4x32 PRNG impl, Add threefry4x32 PRNG impl, and [XLA:GPU] Fold should_autotune logic into the native & block-level emitters.

JAX FIXES REMAT3 CONSTANT HANDLING CRASH, GOOGLE CLOUD PYTHON CENTRALIZES REST TRANSCODING

Matthew Johnson patched a critical remat3 regression where closed-over constants were silently dropped, causing immediate crashes in production rematerialization pipelines [ref:1].

ai 50 shipped 1-min read

Google 1 day ago

JAX TIGHTENS XLA OPTIMIZATION CONTROLS, GOOGLE CLOUD PYTHON SETS UP FEATURE GATES

JAX makes XLA optimization levels human-readable while Google Cloud Python client libraries prepare a unified feature gating system for observability.

ai 54 shipped 1-min read

Google 2 days ago

GENAI AUDIO TRANSCRIPTION SHIPS, PROTO-PLUS FIXES RACE CONDITIONS

Google's AI SDK gains native audio processing while the core Python cloud libraries ship critical stability fixes across the stack.

ai 60 shipped 1-min read

Google 3 days ago

JAX CUTS ASYNC COLLECTIVE PRIMITIVES, GENAI SDK ADDS LIVE API VOCABULARY

JAX is simplifying its async collective operations while the Python GenAI SDK ships custom vocabulary support for the Live API, marking a shift toward leaner abstractions and richer real-time capabilities.

ai 100 shipped 1-min read

Elsewhere on the wire

AI Agents about 9 hours ago

CLAUDE OPUS 5 LANDS ACROSS THE STACK

The newest Anthropic model is now live in langchain, Cline, and llama-index, with native support for extended reasoning and 1M context windows.

ai-agents 28 shipped 1-min read

Local LLMs about 9 hours ago

OLLAMA LANDS LAGUNA SUPPORT AND CRUSHES MEMORY LEAKS WHILE SGLANG HITS V0.5.16 WITH CONFIDENCE-DRIVEN SPECULATIVE DECODING

Ollama shipped three critical performance and reliability fixes for Metal residency and concurrent access patterns, while SGL-Lang released 0.5.16 with a new speculative algorithm hitting 383.7 tok/s on DeepSeek-V4.