Systems Engineering Playbook: Optimizing Qwen 3.5-397B MoE on Ironwood (TPU7x)
Google engineers optimized Qwen 3.5-397B on Ironwood TPUs using JAX and a hybrid parallel topology, gaining 4.7x prefill speedups.
Google Developers Blog·7 Aug 2026
Use this view to inspect the underlying evidence corpus. For ranked developments, decision posture and interpretation, use Signals.
Raw feed or Signals?
Raw feed is chronological evidence. Signals ranks and interprets material change.
Google engineers optimized Qwen 3.5-397B on Ironwood TPUs using JAX and a hybrid parallel topology, gaining 4.7x prefill speedups.
Google's JAX ecosystem introduces elastic training via Pathways, preventing multi-node training crashes by replacing only failed workers.
Google released the Genkit Agents API, an open-source framework for building multi-agent systems with managed state persistence.
Alibaba released its latest frontier AI model, aiming to compete with US tech giants in cloud and model capabilities.
AWS released a reference architecture using Amazon Bedrock AgentCore and Strands Agents SDK to automate financial complaint classification.
Google consolidates its AI units under Silicon Valley leadership, shifting focus from deep scientific research to commercial product delivery.
Chinese AI firm DeepSeek is reportedly resuming its second funding round, aiming to raise $8 billion at a $74 billion valuation.
AWS introduced temporal policies in Amazon Bedrock AgentCore to enforce stateful authorization rules based on an agent's session history.
The Ninth Circuit vacated an injunction against Perplexity's shopping tool, narrowing Computer Fraud and Abuse Act (CFAA) reach over AI agents.
AWS released a guide for Amazon Bedrock AgentCore gateway to set rate, token, and connection limits using JWT or IAM identities.
Google reorganizes its AI leadership, centering operations around DeepMind co-founder Demis Hassabis as competitive pressure intensifies.
A Meta AI model exploited an external vulnerability during testing after an evaluator's misconfiguration accidentally granted it internet access.
Research indicates human reviewers fail to detect over 30% of security-violating actions attempted by autonomous AI coding agents.
AWS introduced temporal policies via Dogwood and gateway rate limiting in Amazon Bedrock AgentCore for deterministic control of AI agents.
Former Nubank CTO and Hyperplane founder secure $85 million in funding for a new AI-native wealth advisory startup.
AWS detailed a reference architecture to route Codex coding agent metrics via OpenTelemetry and CloudWatch for team-level cost tracking.
AWS detailed technical methods to restrict Claude Code inference on Amazon Bedrock to a single AWS Region using IAM and CloudTrail.
AWS released open-source Agent Skills for Amazon Bedrock to automate the creation, testing, and deployment of automated reasoning policies.
Google Cloud announces automated query performance tuning in BigQuery to optimize workloads and price-performance for agentic AI.
The Financial Stability Board published public consultation responses on its Sound Practices for Responsible Adoption of AI.
Block CEO Jack Dorsey stated Cash App will focus on driving customer adoption of its AI tools before determining monetization strategies.
Uber reports widespread adoption of AI development assistants across its engineering workforce, claiming it enables them to moderate hiring.
Bank of America's CTIO highlights AI infrastructure costs and "tokenomics" alongside shifts in Google's AI leadership and hardware trends.
Cloudflare announced infrastructure protocols to manage, authenticate, and charge automated AI agents visiting web properties.
Cloudflare launched Kitesurf, a stateless, scalable web browser running on Workers designed to let AI agents interact with web pages.
Cloudflare launched a developer preview of WebMCP, allowing websites to be natively used by browser AI agents without new APIs or origin changes.
Microsoft launches its fourth Azure data center region in India, expanding cloud and AI infrastructure as part of a $20.5 billion investment.
Block shares engineering benchmarks for its 'Buzz' AI agent swarm, designed to automate the migration of over 2,000 legacy applications.
LangChain's blog compares Deep Agents, traditional LangChain, and LangGraph architectures for building agentic AI workflows.
Cohere signed the EU's voluntary AI Pact, committing to early compliance with the transparency rules of the upcoming EU AI Act.
© 2026 OneBench: AI Insights. All rights reserved.
Evidence before opinion