Anthropic launched Claude Opus 4.6, an upgrade to its flagship model. A key improvement: better performance when working with large amounts of information. Previous models sometimes lost track of details or got "lost in the middle" of long tasks; Opus 4.6 is designed to handle that more reliably. For coding, this means Claude Code can now navigate larger codebases and make smarter decisions about where to add new code. Beyond coding, Anthropic says the model improves performance on everyday work tasks such as financial analysis, research, and creating spreadsheets. The release comes as competition between Anthropic and OpenAI intensifies, particularly around AI-powered coding tools.

Anthropic's Claude Cowork (available on macOS) moves Claude beyond the chat window. Combined with two related features, Skills and Plugins, Cowork creates a system that lets Claude read your local files, run multi-step workflows, create polished documents (Excel, PowerPoint, Word, PDF), and even control your browser.

Charlie Hills walks through the full setup, including how to build your first Skill, ten ready-to-use Cowork prompts, and six real workflows across content creation, sales prep, contract review, and data visualization.

Matt Davey published an open letter to his team arguing that "we're using AI" has become as meaningless as "we use the internet." He maintains that the gap between teams that are integrating AI into real workflows and those just producing more stuff with it is widening fast. To address this, he created an AI Product Design Maturity Model (available free on Figma Community) that gives teams a shared language for assessing where they are and where they need to go. The model spans five levels—from Limited (AI as a personal productivity aid) through Reactive (ad-hoc experimentation with inconsistent results) to Leading (where AI capability compounds across tooling, culture, strategy, and product quality). It covers six dimensions, including leadership, strategy & budgeting, culture & talent, learning & enablement, agents & automation, and product design.

If your team is feeling pressure to "do something with AI" but lacks a clear direction, this model is a solid starting point for having a more structured conversation about what progress actually looks like.

John Cutler (of Dotwork) recently shared an internal document he wrote for his team that identifies ten "forever problems" in product and work management: issues such as messy integrations, eroding data quality, cognitive overload, and the tension between centralization and local autonomy. His core argument is that AI might finally give us a wedge on these longstanding challenges, but only if we're careful and intentional about where we apply it.

Cutler suggests that every team write a similar document internally, mapping AI opportunities to your organization's persistent, specific problems.

Thanks for reading, and see you next Wednesday with more curated AI/UX news and insights. 👋

All the best, Heidi