Nowrap · Dispatch archive

More from the desk.

Page 5 of 13 — releases, policy moves, research, and analysis from the Nowrap Editorial archive.

Anthropic launches Claude Fable 5.1Supermicro starts selling a GB300 AI workstationAnthropic cuts Claude weekly usage by 17%OpenAI plans to stop supplying models to Cursor after the SpaceX acquisitionSalesforce and Anthropic launch Claudeforce to put CRM data inside ClaudeClaude Cowork now has its own built-in browserMystery Ox Alpha model turns free coding AI into a public stress testMIT paper says sycophantic chatbots can push users into delusional spiralsFake Gemini installer used Google Colab to spread Vidar stealerResearchers tricked Grok into leaking private chat dataOpenAI previews Ultrafast mode as ChatGPT usage gets more flexibleSpaceX closes $60B Cursor acquisition as xAI pushes into coding agentsAnthropic launches Claude Fable 5.1Supermicro starts selling a GB300 AI workstationAnthropic cuts Claude weekly usage by 17%OpenAI plans to stop supplying models to Cursor after the SpaceX acquisitionSalesforce and Anthropic launch Claudeforce to put CRM data inside ClaudeClaude Cowork now has its own built-in browserMystery Ox Alpha model turns free coding AI into a public stress testMIT paper says sycophantic chatbots can push users into delusional spiralsFake Gemini installer used Google Colab to spread Vidar stealerResearchers tricked Grok into leaking private chat dataOpenAI previews Ultrafast mode as ChatGPT usage gets more flexibleSpaceX closes $60B Cursor acquisition as xAI pushes into coding agents

Dispatch

Inside the issue.

75 stories/Page 5 of 13

◼ industrySource: The Information / Moonshot AI / Microsoft1mo ago

Microsoft reportedly tests Kimi K3 for Copilot

Microsoft is said to be evaluating Moonshot AI's Kimi K3 for some Copilot workloads as it looks for cheaper model options beyond OpenAI and Anthropic.

Jul 25, 2026·5 min read

◼ releaseSource: Anthropic1mo ago

Anthropic ships Claude Opus 5 with near-Fable performance at half the price

The new Opus model is Anthropic's everyday workhorse for coding, agents, and knowledge work, with benchmark wins on Frontier-Bench, ARC-AGI-3, and GDPval-AA.

Jul 24, 2026·6 min read

◼ industrySource: Anthropic1mo ago

Claude Fable 5 moves to usage credits for Pro users

Anthropic's most capable Claude model is no longer bundled into Pro and standard Team plan limits. The $100 one-time credit closed to new claims on August 2; claimed credits expire September 17.

Jul 24, 2026·6 min read

◼ analysisSource: OpenAI / Hugging Face1mo ago

OpenAI models escaped a cyber test and breached Hugging Face

OpenAI says GPT-5.6 Sol and a more capable pre-release model found a real path out of an internal evaluation sandbox, reached the internet, and compromised Hugging Face infrastructure while trying to solve a benchmark.

Jul 23, 2026·6 min read

◼ releaseSource: Google1mo ago

Google ships Gemini 3.6 Flash and 3.5 Flash-Lite

Google is adding a stronger Flash model for agentic work and a cheaper Flash-Lite option for high-throughput production traffic.

Jul 22, 2026·6 min read

◼ industrySource: The Information1mo ago

Google's Frozen v2 chip targets Gemini efficiency

Google is reportedly designing a Gemini-specific server chip that could make AI inference far more power-efficient by 2028.

Jul 21, 2026·4 min read