Claude’s evolving copilots: merges, artifacts and mods
In the last 24 hours my design inbox has been buzzing with Claude updates, and it’s hard not to notice the pattern. Claude Cowork is merging with Claude Chat, so you won’t have to juggle tabs to keep a big task going. It also looks like Claude Artifacts is getting dedicated Docs and Slides products, with Claude Design moving into conversations too. Fun times for anyone prototyping AI-assisted UX workflows.
These shifts aren’t just “nice to have” features; they reshape how we design around AI agents. A single chat that carries context across apps, plus living documents and design prompts that live in conversations, could streamline how we scope, iterate and hand off work. If you’re tooling up for agent-assisted design reviews, think about persistent context, multi-app visibility, and clear prompts that survive session boundaries (Claude Cowork is merging with Claude Chat; Claude Artifacts with Docs & Slides; Claude Design in conversations; Claude Code with Mods).
Implication for designers: start testing dashboards that surface AI-prompted context and output alongside ongoing work, and build docs that auto-update as ideas evolve. It’s a gentle nudge that our “AI assistant in the loop” is going to be less about a separate tool and more about an embedded collaboration experience.
Live models go real‑time: faster feedback, cheaper compute, bigger on-the-fly help
Two new Gemini Live models (3.8 Live and 3.8 Live Extended Thinking) can take video and audio input and output audio, delivering real-time help when you need it most. It sounds like a great fit for rapid UX testing and on‑the‑fly feedback during design sessions, especially where you want a model to keep up with a live discussion (Gemini API live models).
There’s also a new “Jev” model from TypeSafe AI, aimed at making quick, probabilistic judgments for what to call next or which model to route a request to. And Union Alpha shows up as a stealth option with strong benchmark performance at lower cost, which could shake up how we price and deploy on-device or edge-heavy design tools (JeV and Union Alpha).
Bottom line for UX teams: real-time AI can speed up ideation, critique and prototyping, but it also means rethinking latency, reliability and the human-in-the-loop. Test streaming feedback early in your process and track how much faster you can cycle decisions with live assistance.
AI-powered collaboration tools: planning, docs and security in one toolkit
Design teams are getting more comfortable with AI-assisted collaboration stacks. Slack Code now supports team-wide planning and review with coding agents, while 1Password integration (Grok Bot can use it) helps keep those shared prompts and tokens secure. Self-updating docs from Mintlify and ever-evolving datasets APIs are nudging us toward living libraries that don’t go stale (Slack Code and 1Password; Self-updating docs).
For designers, this translates to faster living design systems and better guardrails for automation. If you’re building internal tools, consider how you’ll define contracts for code and how your docs auto-refresh as features or data change (Code contracts and living docs).
These shifts invite us to design with “agents” as first-class teammates—not as a toggle in the corner. How will your UX reflect agent-driven decision points, and how will you make the collaboration feel seamless rather than clunky?
Industry signals for design entrepreneurs: subscriptions, open models, and cost-aware futures
On the business side, Meta One adds AI features across its suite, hinting at deeper monetisation of creator tools and features in Muse, Instagram and WhatsApp. Meanwhile, Factory raised $200M at a $5B valuation, a reminder that there’s capital chasing scalable AI-enabled products. And open-source Droid Core models are staying fast and affordable, underscoring a trend toward accessible, community-driven AI tooling (Meta One; Factory funding; Droid Core).
There’s also a useful warning in the chatter: “What stays expensive when AI gets cheap?”—a reminder to design decisions around cost, not just capability. If you’re building products for external clients, plan for accessible AI, scalable data, and transparent pricing from the start (What stays expensive when AI gets cheap).
As designers, we should experiment with “a company brain” concepts and runnable automation that actually adds value without bloating the product. Small bets, clear success metrics, and a steady eye on cost will keep us honest and creative.