Tools & models
Alibaba’s newest Qwen model combines one million tokens of context, multimodal input and agent support across Codex and Claude Code. Its own benchmark table puts it close to the leading frontier models, while QwenCloud’s preview plan starts at $6 a month.
Read moreAI at work
Holland’s sequence is prototype, plan, build and review. If the agent can run commands without approval, keep it away from production systems, live data and unrestricted credentials.
Read moreAI at work
From 2 August 2026, new transparency duties cover AI interactions, synthetic content, deepfakes and some text on matters of public interest. The practical first step is to know where AI enters the work and who takes responsibility before it reaches an audience.
Read moreAI at work
OpenAI’s new enterprise product starts with a defined job, scoped permissions, testing and a route back to a person.
Read moreTools & models
OpenAI’s new agent can stay with a longer task, use connected files and apps, and produce finished documents. This puts more weight on the quality of the brief.
Read moreTools & models
Anthropic’s newest Sonnet model can plan, use tools and complete longer tasks at a lower price than its largest models. For smaller teams, that reduces the cost of testing a custom AI tool.
Read more
