Hello from The AI Night,
Today in AI:
You Can Now Remote-Control Claude Code From Your Phone
Perplexity Just Dropped Their Own OpenClaw For Browsers
Claude Cowork Now Runs Your Recurring Tasks Automatically
Here's the deal: Anthropic launched Remote Control inside Claude Code, letting developers start a coding session in their terminal and continue it from a phone, tablet or any browser through claude.ai/code or the Claude mobile app.
The Breakdown:
Sessions run entirely on your local machine. The web and mobile interfaces act as a window into that local environment not a cloud replica.
Your full local setup stays available remotely: filesystem, MCP servers, tools and project configuration.
Conversations sync across all connected devices. You can send messages from your terminal, browser and phone interchangeably.
If your laptop sleeps or your network drops, the session reconnects automatically when the machine comes back online.
Available as a research preview on Pro and Max plans only. Not supported on Team or Enterprise plans.
Each Claude Code instance supports one remote session at a time. Closing the terminal ends the session.
Extended network outages beyond roughly 10 minutes will time out the session.
The bigger picture: This removes the "stuck at my desk" constraint from local-first AI coding workflows. Developers can now step away from their machine without abandoning context, which keeps Claude Code competitive against cloud-native alternatives while preserving the security and tooling advantages of local execution.
Image Source: Perplexity blog
Here's the deal: Perplexity introduced Perplexity Computer a general purpose digital worker that breaks user-defined goals into tasks and subtasks, spawning sub-agents to execute them across isolated compute environments with real browsers, filesystems and tool integrations. It is available now for Perplexity Max subscribers with Enterprise Max access coming soon.
The Breakdown:
Operates as an orchestration layer not a single model. It delegates work to specialized frontier models; Opus 4.6 for core reasoning, Gemini for deep research, Grok for lightweight speed tasks, ChatGPT 5.2 for long-context recall, Nano Banana for images and Veo 3.1 for video.
Workflows can run for hours or months, asynchronously. Users can run multiple Computers in parallel.
Sub-agents handle web research, document generation, data processing and API calls. When blocked, the system spawns additional sub-agents to resolve issues autonomously.
Model assignments are swappable as capabilities evolve. Users can also manually assign models to specific subtasks.
The bigger picture: This positions Perplexity as a model agnostic orchestration platform rather than a wrapper around any single provider. For power users managing complex, multi-step projects, the value proposition shifts from "better answers" to "delegated execution," a meaningful distinction as agent frameworks become the next competitive frontier.
Image Source: Anthropic blog
Here's the deal: Anthropic announced new capabilities for Claude Cowork, its desktop automation tool. Cowork can now execute recurring tasks at set times automatically and gains a plugin system that adds specialized expertise across multiple work domains.
The Breakdown:
Scheduled tasks let Claude handle recurring work automatically at specific times, such as morning briefs, weekly spreadsheet updates, or Friday team presentations
Plugins give Cowork domain expertise across design, engineering, operations, and other fields, available at claude.com/plugins
A new Customize tab in the Cowork sidebar centralizes management of plugins, skills, and connectors in one place
This builds on Cowork's existing role as a desktop tool for non-developers to automate file and task management
The bigger picture: Scheduled tasks move Cowork from reactive (user triggers a task) to proactive (Claude runs tasks independently on a timer). For teams and individual operators, this turns Cowork into a lightweight automation layer that handles routine knowledge work without requiring technical setup.
What else you need to know:
OpenAI appointed Arvind KC as Chief People Officer to lead hiring, development and workforce adaptation efforts, drawing on his prior leadership roles at Roblox, Google, Palantir and Meta.
Alibaba's Qwen team released the Qwen 3.5 medium model series including Flash, 35B-A3B, 122B-A10B and 27B variants that outperform larger predecessors while using significantly less compute.
MiniMax launched MaxClaw, an always-on AI agent built on OpenClaw and its M2.5 model that works natively across Telegram, WhatsApp, Slack and Discord with no deployment or extra API costs.
Anthropic acquired Vercept, a startup focused on AI perception and interaction, to advance Claude's computer use capabilities following Sonnet 4.6 reaching 72.5% on the OSWorld benchmark.
Google expanded Flow into a full AI creative studio with a redesigned interface that lets users draft, visualize and refine stories using images, videos and natural language prompting in one workflow.
That’s it for today’s edition of The AI Night.
Our goal is to cut through the noise, surface what actually changed, and explain why it matters.
If this was useful, you’ll get the same signal here tomorrow.