AI News: Anthropic Went Crazy This Week!

TL;DR: Anthropic shipped 74 updates in 52 days—highlighting Computer Use, Projects, and Claude Code Auto Mode—while Google countered with Gemini 3.1 Flash Live, vibe-coded browser demos, and new Lyria 3 Pro music tools, all as newcomer GenSpark promises unlimited AI for $20/month through 2026. ## Anthropic’s Blitz: 74 Drops in 52 Days Product Compass logged 74 public releases from Anthropic between mid-February and late April—an average of more than one per day. The past week alone added Cloud Tasks, Projects, Co-work Dom selection, Effort & Skills, Cloud Code updates, Computer Use, Cloud Code Schedule, Channels, Permission Prompts, and new team-plus-enterprise tiers. Mobile Claude now embeds Figma, Canvas slides, and Amplitude dashboards; Claude Code gained an “Auto Mode” that skips permission prompts for safe commands. ## Three Features You’ll Actually Use ### Computer Use (23 April) Toggle “Computer Use” in settings (any paid plan from $20/month) and Claude gains mouse-and-keyboard control. Pair it with last week’s Dispatch feature and you can text instructions from your phone while away from your desk. A five-minute demo had Claude open DaVinci Resolve, locate the Magic Mask panel, and explain its purpose—slow but fully hands-off. ### Projects Claude Co-work now hosts persistent Projects. Upload background files and write custom instructions once; every follow-up message respects that context. Fun example: instruct the model to “talk like a gym bro” and every reply starts with “dude” or “bro.” Practical use: preload product specs, style guides, or codebases for tighter answers. ### Claude Code Auto Mode Formerly, Claude stopped every few minutes to ask “Can I run this command?” Auto Mode green-lights safe terminal actions and web searches by default, letting long coding sessions run unattended—developers call it “a sanity saver.” ## The New Kid: GenSpark GenSpark, already running Super-Bowl ads, pitches itself as an all-in-one AI workspace. Inside one tab you can chat, generate slide decks, write reports, publish web pages, create images, and edit video. The $20/month paid tier offers unlimited AI chat plus unlimited image generation through 31 December 2026—priced well below stacking separate OpenAI, Midjourney, and Runway subscriptions. A free tier with starter credits is available via the link in the video description. ## Google’s Rapid Response ### Gemini 3.1 Flash Live Real-time voice, vision, and screen-share are now live in Gemini Live, enterprise APIs, and Google Search’s AI Mode. Demo: share an OBS window and Gemini names the software, explains each panel, and walks the user through setup—all without typing. ### Vibe-Coded Browser Using Gemini 3.1 Flash, engineers typed “Taco Cat Parade” and instantly received a fully styled web page; clicking nav links generated new sections on the fly. Google stresses it’s a research preview, but the zero-latency feedback loop impressed the crowd. ### One-Click Data Migration Following Anthropic’s lead, Google now offers an importer that pulls ChatGPT or Claude chat history, memories, and custom instructions into Gemini—no JSON wrangling required. ### Lyria 3 Pro Google’s answer to Suno produces radio-length tracks up to three minutes. Users can structure songs with intro, verse, chorus, and bridge prompts. Lyria 3 Pro is accessible in Vertex AI, AI Studio, Gemini API, and the new Google Vids storyboard tool. ### Suno 5.5 The latest Suno release adds voice-cloning: record (or upload) a few seconds of your own voice, train a model, then generate full songs in that vocal style. The tagline: “Can’t sing? Doesn’t matter—release the single anyway.” Source: [Matt Wolfe on YouTube](https://www.youtube.com/watch?v=OYyS0Gu5xj8)