Who contributed to Google on July 31, 2026?

9 developers shipped this update, including kodlan, copybara-service, jax authors, release-please[bot], Amir Hardon, Anthonios Partheniou, sakthivelmanii, and sycai, and 1 more.

What were the notable Google updates?

Fix cudnn fp8 attention batching rules, [JAX] Lower cumulative reductions to chlo.ScanOp on TPU, and v2.16.0.

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 FIXES FP8 ATTENTION, LANDS TPU CUMULATIVE REDUCTIONS

By RepoJournal · Filed 06:59 UTC on July 31, 2026 · About Google

9 people shipped this

Anthonios Partheniou @parthea 2 cited

kodlan @kodlan 1 cited

copybara-service @copybara-service 1 cited

release-please[bot] @release-please[bot] 1 cited

sakthivelmanii @sakthivelmanii 1 cited

jax authors @Google-ML-Automation

Amir Hardon @amirh

sycai @sycai

hebaalazzeh @hebaalazzeh

JAX shipped critical fixes for fp8 batching rules and native TPU support for cumulative operations, while the Gemini SDK crossed v2.16.0 with new robotics models and expanded grounding capabilities.

The JAX team merged a follow-up fix for fp8 attention batching that applies the same normalization logic as the earlier non-fp8 correction [1], addressing a root cause that affected multiple code paths. Simultaneously, cumulative reductions (cumsum, cumprod, cummax, cummin) now lower to native chlo.ScanOp on TPU [2], mirroring GPU behavior while preserving backward compatibility through the TPU compiler's rewrite pipeline. On the SDK side, python-genai v2.16.0 landed with environment resource support, native GoogleMaps Tool grounding types, and the Gemini Robotics ER 2 Preview model [3], expanding multimodal capabilities for production workloads. The google-cloud-python ecosystem saw Protobuf 6.33.5 become a hard requirement [6] to unblock client library modernization, while SQLAlchemy-Spanner gained native UUID data type support with opt-in backward compatibility [5]. One caveat: google-cloud-bigquery was temporarily dropped from bulk release due to a socket leak in the dbapi connection tests [4], requiring isolated investigation before reintegration.

Action items

→ Pull JAX fp8 and TPU cumulative fixes immediately if running fp8 attention or cumulative reductions on TPU google/jax [immediate]
→ Upgrade python-genai to v2.16.0 and integrate Gemini Robotics ER 2 if deploying multimodal applications googleapis/python-genai [plan]
→ Enforce Protobuf 6.33.5+ in google-cloud-python dependencies before next release cycle googleapis/google-cloud-python [plan]
→ Hold google-cloud-bigquery deployments until socket leak fix lands; monitor PR for resolution googleapis/google-cloud-python [monitor]

References

[1] Fix cudnn fp8 attention batching rules ↗ google/jax
[2] [JAX] Lower cumulative reductions to chlo.ScanOp on TPU. ↗ google/jax
[3] v2.16.0 ↗ googleapis/python-genai
[4] chore: remove google-cloud-bigquery from bulk release due to failing system test (#17944) googleapis/google-cloud-python
[5] feat(sqlalchemy-spanner): native UUID data type support in dialect ↗ googleapis/google-cloud-python
[6] fix: require Protobuf 6.33.5+ ↗ googleapis/google-cloud-python

Quick answers

What shipped in Google on July 31, 2026?: JAX shipped critical fixes for fp8 batching rules and native TPU support for cumulative operations, while the Gemini SDK crossed v2.16.0 with new robotics models and expanded grounding capabilities. In total, 20 commits, 16 pull requests, and 2 releases landed.
Who contributed to Google on July 31, 2026?: 9 developers shipped this update, including kodlan, copybara-service, jax authors, release-please[bot], Amir Hardon, Anthonios Partheniou, sakthivelmanii, and sycai, and 1 more.
What were the notable Google updates?: Fix cudnn fp8 attention batching rules, [JAX] Lower cumulative reductions to chlo.ScanOp on TPU, and v2.16.0.

JAX PINS PYTHON 3.14T, GENAI GAINS ROBOTICS MODEL AND ERROR HANDLING

JAX disabled profiler tests on free-threaded Python to unblock CI, while python-genai 2.17.0 ships the Gemini Robotics ER 2 Preview and fixes tool call limits.

ai 90 shipped

Google Aug 5

JAX ADDS NUMPY 2.6 TOP_K, INTEL GPU SUPPORT LANDS IN CI

JAX shipped native top_k alignment with NumPy 2.6.0 while simultaneously landing full OneAPI CI infrastructure for Intel GPU testing across the entire stack.

ai 59 shipped

Google Aug 4

JAX HARDENS GPU KERNELS WHILE CLOUD PYTHON PATCHES SUPPLY CHAIN RISK

JAX shipped critical Mosaic GPU optimizations that squeeze performance out of transposed memory transfers, while the cloud SDK locked down six open security vulnerabilities in sqlalchemy-spanner.

ai 100 shipped

Google Aug 3

JAX SYNCS XLA DEPENDENCY TO LATEST REVISION

JAX pinned its XLA dependency to a fresh upstream revision overnight, pulling in compiler optimizations and fixes across the ML stack.

ai 1 shipped

Elsewhere on the wire

AI Agents Aug 6

LANGCHAIN HARDENS ANTHROPIC TOOL CALLING AND GIT CREDENTIALS AS OPENHANDS SHIPS 1.10.0

LangChain fixed a critical credential leak in evaluation runs while Anthropic tool schemas get proper fallback handling, as three major agent frameworks push releases across the ecosystem.

ai-agents 54 shipped

Local LLMs Aug 6

OLLAMA FIXES KV CACHE CORRUPTION, LLAMA.CPP HARDENS FILE SEARCH, VLLM KILLS MOE LEGACY CODE

Ollama patched a critical cache corruption bug that corrupts lazy snapshots during multi-request prefills, while llama.cpp hardened Windows junction traversal and vLLM removed 14 hardcoded MoE shards across quantization methods.