Stocks Waver as Mixed Earnings and Chipmaker Rebound Offset Software Weakness

Hirundo, an AI safety lab specializing in machine unlearning, released Westernized versions of Alibaba's Qwen open-weight models with Chinese Communist Party political alignment removed directly from the model weights. On Hirundo's evaluation, the original Qwen3.6-35B-A3B produced CCP-aligned censorship, propaganda-aligned framing or political bias in 89.8% of responses across a 500-prompt benchmark, while the Westernized model did so in 2.8%, with reasoning, coding and instruction-following performance essentially unchanged. The reduction held on two external benchmarks, with refusals on DECCP falling from 65.26% to 3.16% and non-compliance on ChinaBench falling from 96.67% to 6.67%, and the same method cut CCP-aligned responses in the much smaller Qwen3.5-4B from 89.2% to 1.2%. On GPQA, IFBench, LiveCodeBench and MMLU-Pro, the Westernized model's scores stayed within 0.72 points of the original model's on average, and its safety and harmfulness benchmark scores also held. Both models are available now on Hugging Face as Qwen3.6-35B-A3B-Westernized and Qwen3.5-4B-Westernized, and Hirundo intends to release its CCPC-500 benchmark publicly.

Pegasystems announced Launchpad.io's AI Assistant, a conversational development tool that lets enterprises design, build, and evolve applications using natural language while preserving governance, reliability, and scalability. The assistant turns plain-language business requirements directly into application logic, workflows, and data models, which could widen participation in enterprise software development across non-technical teams. The launch follows the general availability of Pega Infinity 26 in July 2026, which introduced broader AI assistants, agentic workflows, and no token pricing across the core platform, with Launchpad.io extending those ideas into more accessible, conversational app creation. Pegasystems' narrative projects $2.3 billion in revenue and $334.9 million in earnings by 2029, requiring 10.1% yearly revenue growth and an earnings increase of about $10.9 million from $324.0 million today, while some of the most optimistic analysts already expected about US$2.4 billion in revenue and roughly US$397.9 million in earnings. The company's forecasts yield a $40.00 fair value, an 18% upside to its current price.

ZoomInfo announced Agent Teams on October 1, a native AI agent orchestration layer built directly into its existing GTM Studio environment at no additional cost, making it the first major GTM data vendor to embed agent orchestration as a standard platform capability rather than a premium tier. The feature requires no new SKU, no separate contract, and no additional purchase, and ships with more than 50 ready-made agents that run against ZoomInfo's GTM Context Graph, a data layer connecting more than 100 million companies, 500 million contacts, and billions of buying signals. The architecture comes from DoubleO.ai, which ZoomInfo acquired on September 30 for undisclosed financial terms, one day before the launch. Every play operates through a Playbook configuration layer with audit trails and credit estimates, addressing a trust gap highlighted by Cisco research at RSA 2026 showing 85% of organizations experimenting with agentic AI but only 5% in broad production, and by Gartner's projection that more than 40% of agentic AI projects will be canceled by the end of 2027. Agent Teams ships with native integrations into Salesforce, HubSpot, Outreach, Salesloft, Slack, and Gmail, and connects to Salesforce Agentforce, HubSpot Breeze, Microsoft Copilot Studio, Claude, and ChatGPT. ZoomInfo has not published reliability benchmarks or performance metrics for Agent Teams, and the specific plays described are vendor-reported capabilities.