OpenAI released ChatGPT Work, a workplace AI agent powered by GPT-5.6, with scheduled automation, business app connections, and Codex merged into the desktop client. Pro, Enterprise, and Edu users get first access.
California Governor Gavin Newsom has proposed extending the state’s 7.25% sales tax to SaaS and cloud software subscriptions. If approved, the measure would take effect in 2027 and generate $2 billion annually from year two.
Harness Engineering is emerging as a core discipline for AI Agent deployment, covering context, tools, permissions, memory, hooks, sub-agents and prompt caching.
This in-depth review examines Match2Pay, a specialized crypto payment infrastructure provider for brokers, prop trading firms, and fintech platforms, covering its product architecture, deployment models, security, compliance, and competitive positioning.
China’s largest internet companies are moving in the same direction on AI agents: fewer standalone products, more unified workplace entry points. Over the past month, Tencent folded the business of its QClaw product center and part of the team into WorkBuddy. Alibaba is preparing to launch "Qwen Office," combining QoderWork, Wukong and MuleRun under DingTalk CEO Chen Yusen. ByteDance, for its part, has renamed TRAE SOLO to TRAE Work, signaling a shift away from a standalone coding agent model. The article argues that this is more than routine internal restructuring. After months of running multiple overlapping agent projects, major companies are now pulling them into a single interface, reflecting a broader industry shift from experimentation to concentration. The core issue is cost, resource allocation and control over enterprise data and API orchestration. As open-source tools flatten technical barriers, the advantage no longer lies in launching more agents, but in owning the main workplace interface through which those agents operate. That shift also changes the target market. Coding remains an early AI use case, but the larger opportunity is general office work: email, meetings, documents, approvals and decision follow-up. In that model, agents become less visible as products and more embedded as infrastructure inside a "super workbench."
Metaplanet Inc., described in the source as Japan’s largest Bitcoin treasury company, said on July 21 that changes had occurred among its major shareholders and in its top shareholder position. U.S. investment adviser Capital Research and Management Company, or CRMC, became the company’s largest shareholder after increasing its holdings. According to the disclosure, CRMC’s directly held voting rights rose from 1.1865 million to 1.3606 million, while its share of total shareholder voting rights increased from 9.32% to 10.63%. The update was reported by BlockBeats.
A BlockTempo report uses what writer Joan Westenberg calls the “bread paradox” to argue that lower software creation costs do not automatically destroy the software-as-a-service business. The comparison is simple: bread machines are cheap, ingredients are inexpensive, and recipes have been available for thousands of years, yet consumers still buy commercially made bread at scale. The same logic, the article argues, applies to enterprise software. Companies do not pay vendors such as Notion, Jira, or Basecamp merely for code. They pay for maintenance, compliance work, security review, integrations, certifications, support systems, and the ability to offload operational risk. The report also points to evidence that AI-generated software is not a clean substitute for managed services. It says research shows major defects in AI-generated code appear at roughly 1.7 times the rate found in human-written code. Gartner, meanwhile, has observed enterprise SaaS renewal increases still commonly landing in the 10% to 20% range, suggesting buyers are continuing to pay even as AI coding tools improve. A January 2026 report from Avenir found that 63% of enterprise buyers expect existing software vendors to benefit from generative AI, while only 8% expect them to be harmed. In this view, the most exposed products are thin, single-function tools, while deeper software supply chains remain intact.
Anthropic unveils new features for Claude agents, including a 'Dream' tool for memory improvement, multi-agent coordination, and outcomes evaluation. Internal tests show a 10% increase in task success rate, aiming to enhance enterprise AI collaboration.
