🤖 AI Weekly Recap (Week 8)

Happy Sunday! We just had another crazy week in AI. Google dropped Nano Banana 2 with dramatically improved text rendering inside images, Anthropic turned Claude into a mobile-controlled coding terminal, and Alibaba released open models that rival top proprietary systems.

And that's not all, here are the most important AI moves you need to know this week.

1. Google releases Nano Banana 2

Google unveiled Nano Banana 2, its latest image generation model, powered by real-time web search data for stronger world knowledge and more realistic outputs. The biggest leap? Clean, accurate text inside images.

• Dramatically improved text rendering, making AI images harder to spot and usable for print

• Better instruction following, character consistency, and production-ready outputs (512px–4K)

• Can generate infographics, diagrams, and data visualizations with structured accuracy

• Available across Gemini, Search, AI Studio, Google Cloud, and Google Ads

Try it now → https://gemini.google.com/app

2. Anthropic Brings Claude Code to Mobile

Anthropic introduced Remote Control, a new mode that connects your local Claude Code terminal session to the Claude mobile app. It’s rolling out first as a Research Preview for Claude Max ($100–$200/month) users, with Pro access coming later.

• Lets you start a coding task on desktop and control it from iPhone or Android via synced mobile session

• Runs locally on your machine, no inbound ports or file uploads, just encrypted API bridging

• Auto-reconnects if your laptop sleeps or drops connection

• Replaces developer “hacks” like SSH tunnels, Tailscale, and custom WebSocket bridges

Try it now → https://code.claude.com/en/remote-control

3. QuiverAI Launches Arrow 1.0 SVG Model

QuiverAI introduced Arrow-1.0 in public beta, positioning it as a frontier model for vector-first AI design. Instead of generating pixels, Arrow produces structured SVG code, making outputs fully editable and production-ready.

• Generates SVG vector graphics directly from text or image inputs

• Built on research around “visual code generation” (treating visuals as structured code, not pixels)

• Enables editable, composable, and workflow-ready design outputs

• Raised $8.3M seed led by Andreessen Horowitz (a16z) to scale research and product

Try it now → https://app.quiver.ai/

4. Alibaba Drops Qwen3.5 Medium Open Models

Alibaba’s Qwen team unveiled four new models, three of them Apache 2.0 open source, delivering performance that rivals or beats top proprietary systems while running efficiently on consumer hardware.

• Qwen3.5-35B-A3B supports 1M+ token context on 32GB GPUs with near-lossless 4-bit quantization

• Hybrid MoE architecture (35B total, only 3B active) slashes inference cost and latency

• Benchmarks show it outperforming GPT-5-mini and Claude Sonnet 4.5 in key reasoning tasks

• Qwen3.5-Flash API priced at $0.50 per 1M tokens total, among the cheapest globally

Try it now → https://chat.qwen.ai/

5. Anthropic Adds Scheduled Tasks to Claude Cowork

Anthropic has officially rolled out scheduled task automation inside Claude Cowork, turning the AI from a reactive assistant into a proactive agent that executes multi-step workflows on a set timetable.

• Users can schedule recurring tasks (daily, weekly, weekdays) like email summaries, spreadsheet updates, or team presentations

• Works with integrations including Gmail, Slack, Google Drive, Asana, Canva, and Notion

• No coding required, just type /schedule inside any Cowork session

• Available on paid plans (Pro, Max, Team, Enterprise) via the Claude Desktop app

Try it now → https://claude.ai/

6. Perplexity Launches a “Computer” AI Agent

Perplexity released Perplexity Computer, a general-purpose AI agent that can create and execute full workflows across tools, files, the web, and code, running for hours or even months with minimal input.

• Works like a digital employee: operates browsers, tools, files, coding environments, and research workflows

• Orchestrates 19 AI models (including OpenAI, Google, and others) by assigning tasks to the best model for each job

• Lets users specify an outcome, then auto-splits work into agents and sub-agents that run in parallel

• Launches web-only for Max users ($200/month) with per-token billing and bonus launch tokens

Try it now → https://www.perplexity.ai/