Tech • AI • Robotics • Game
AI News: Claude 2.0 Dropped, Grok Bot Voice & Claude Code Projects (+ more)
8/10TL;DR
Anthropic, Higsfield, Grockbot, Meta Muse, and OpenAI rolled out a wave of AI product updates centered on unified workspaces, lower-cost media generation, voice-controlled agents, and more specialized professional tools.
KEY POINTS
Claude merges chat and work modes
Anthropic is combining Claude Chat and Claude Co-work into a single interface designed to remove the need to choose between modes before starting a task. The updated layout adds an output area where users can generate and edit materials directly inside Claude, while co-work settings still allow project selection and approval controls such as manual, automatic, or skip all approvals.
Claude Docs and Slides arrive inside the workspace
New tools called Claude Docs and Claude Slides let users turn uploaded files into editable documents and presentations without leaving the Claude environment. A CSV file can be transformed into a structured document with summaries, tables, and status fields that can be manually edited, converted into checklists, or split into multiple tabs for different workflows.
Presentations can be customized and exported
Claude Slides can convert the same source data into a slideshow and apply preconfigured visual systems from Claude Design. The resulting slides remain editable, can include images and speaker notes, and can be shared by link or exported to PDF, PowerPoint, HTML, and third-party apps including Canva, Gamma, and Miro.
Claude Projects adds parallel task handling
Anthropic also introduced an update to Claude Projects in Claude Code that allows multiple task threads to run in parallel within a single project. The feature creates separate working threads for subtasks while keeping them tied to one broader request, though access is currently a limited beta for select Claude Pro and Max users using cloud sessions and meeting specific account conditions.
Higsfield opens its API and pitches lower costs
Higsfield has opened its platform and made a public API available for image and video generation. The main appeal is cost: instead of paying for a recurring subscription, users can access models on a pay-as-you-go basis, with listed discounts such as 15% off models including Kling 3.0 and Seedance 2.5, and up to 50% savings when selecting certain model bundles.
One key gives access to multiple media models
The Higsfield model marketplace bundles access to several image and video systems under a single API key, reducing the need to manage separate providers. Available options include Seedance 2.5, Seedance 2.0, Kling 3.0, Kling 2.6, Veo 3.0 Prime, and Marketing Studio Image, with a dashboard for tracking spend and comparing prices against rivals such as FAL AI.
Video generation is cheaper, but still costly at scale
In one test case, a 5-second, 720p, 16:9 video generated with Seedance 2.5 came to roughly $1.60 after discount, with two total generations costing $3.24. That makes the API attractive for teams building generation into apps or avoiding another monthly subscription, but heavy usage could still add up quickly.
Grockbot rolls out voice mode for AI agents
Grockbot introduced a new voice mode on desktop and mobile, allowing users to speak with individual AI agents rather than only typing commands. Each agent can be assigned a distinct voice, playback speed can be increased to 1.25x or 1.5x, and the system supports multiple languages with an auto-detect option.
Voice agents can coordinate tools and computer actions
Grockbot’s voice-enabled agents can interact with connected apps and, in some cases, perform computer-use actions such as opening websites or handling inbox workflows. The platform supports around 24 connected applications including Gmail, ClickUp, Clay, X, and Granola, while Zapier MCP can bridge unsupported tools such as Drip for email campaign data retrieval.
Meta Muse and OpenAI expand into specialist workflows
Meta’s Muse is now available on Mac with access to local folders and files, enabling tasks such as organizing downloads, finding documents, and summarizing notes or messages. OpenAI also launched ChatGPT Law, a legal-focused offering currently limited to certain law firms, signaling a broader push toward industry-specific AI models that could later extend into fields such as finance.
CONCLUSION
This week’s releases point to a clear industry shift: AI companies are trying to make assistants more useful inside real workflows by combining interfaces, connecting more tools, and tailoring models to specific jobs. The competition is increasingly about execution, integration, and price rather than raw model access alone.
Ask a question