Who contributed to Google on May 19, 2026?

2 developers shipped this update, including copybara-service[bot] and jskeet.

What were the notable Google updates?

[Mosaic GPU][NFC] Split swizzle and memref transform extraction from transform attributes, [Mosaic GPU][NFC] Delete unused `TransposeTransformAttr`, and [NFC] Remove unnecessary calls to `swizzle_and_transforms_from_transforms_attr`.

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 REFACTORS GPU TRANSFORM PIPELINE, PALLAS FUSER GAINS STATEFUL KERNEL SUPPORT

By RepoJournal · Filed 06:03 UTC on May 19, 2026 · About Google

2 people shipped this

copybara-service[bot] @copybara-service[bot] 3 cited

jskeet @jskeet 1 cited

JAX's Mosaic GPU backend is undergoing significant internal restructuring to support stateful operations, while the Pallas Fuser now handles read-write Refs for fusion kernels.

The JAX team shipped a substantial refactor across the Mosaic GPU stack [1], splitting the monolithic `swizzle_and_transforms_from_transforms_attr` function into two specialized functions: `swizzle_from_transforms_attr` and `memref_transforms_from_transforms_attr`. This change removes an implicit requirement that swizzled references must have tiling transforms, opening the door for more flexible memory layouts. The refactor also cleaned up unused attributes [2] and eliminated unnecessary function calls [3], reducing cognitive overhead across the codebase. On the runtime side, the Pallas Fuser shipped support for stateful operations [4], allowing fusible kernels to read and write through Refs. This means you can now write fusions that blend computation with state updates like adding bias mid-kernel without breaking the pure-function contract. The Google Cloud Python libraries saw maintenance across multiple services [5], including datacatalog-lineage regeneration and fixes to Spanner's backup restore logic [6] that address timeout and premature instance cleanup bugs.

Action items

→ Review JAX Mosaic GPU transform changes if you maintain downstream GPU compilation jax-ml/jax [plan]
→ Test Pallas Fuser stateful kernel support in your fusion workflows jax-ml/jax [plan]
→ Pull google-cloud-python updates if you use Spanner backup restore googleapis/google-cloud-python [monitor]

References

[1] [Mosaic GPU][NFC] Split swizzle and memref transform extraction from transform attributes. ↗ google/jax
[2] [Mosaic GPU][NFC] Delete unused `TransposeTransformAttr`. ↗ google/jax
[3] [NFC] Remove unnecessary calls to `swizzle_and_transforms_from_transforms_attr`. google/jax
[4] [Pallas Fuser] Support fusions that read and write Refs. ↗ google/jax
[5] feat: regenerate google-cloud-datacatalog-lineage ↗ googleapis/google-cloud-python
[6] test(spanner): fix backup restore TimeoutError and premature instance cleanup (#17150) googleapis/google-cloud-python

Quick answers

What shipped in Google on May 19, 2026?: JAX's Mosaic GPU backend is undergoing significant internal restructuring to support stateful operations, while the Pallas Fuser now handles read-write Refs for fusion kernels. In total, 36 commits and 33 pull requests landed.
Who contributed to Google on May 19, 2026?: 2 developers shipped this update, including copybara-service[bot] and jskeet.
What were the notable Google updates?: [Mosaic GPU][NFC] Split swizzle and memref transform extraction from transform attributes, [Mosaic GPU][NFC] Delete unused `TransposeTransformAttr`, and [NFC] Remove unnecessary calls to `swizzle_and_transforms_from_transforms_attr`.

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.