Grok Imagine Video 1.5 Lite ships: audio clips from $0.02/sec

xAI's text/image-to-video does native sound ~65% under Veo 3.1. Plus Claude Code 2.1.296, Harness buys Augment Code, and persistent Gemini agents.

Nowline OCT 10 5:00 PM banner

Top AI stories from the last hour

Top AI stories from the last hour

Copy markdown

  • Text-to-video with sound, from 2¢ a second

    xAI's Grok Imagine Video 1.5 Lite makes 1–15s clips from text or an image with native audio and seven aspect ratios — $0.02/sec at 480p, $0.03 at 720p, $0.14 at 1080p, roughly 65% under Veo 3.1's $0.40. It's live on the xAI API, fal, Vercel AI Gateway, and OpenRouter, with a 10s 1080p clip in about 60 seconds. Reach for storyboards, b-roll, and social drafts; skip it for tight lip-sync.

  • Claude Code doubles MCP tool-description limits

    Claude Code 2.1.296 (Oct 9) raises the default MCP tool-description cap from 2,048 to 4,096 characters, so richer tool specs stop getting truncated. It also adds a per-subagent autoCompactWindow for earlier compaction, CLAUDE_CODE_WORKFLOW_SUBAGENT_MODEL to pin every workflow agent to one model, and a tunable backoff for 529 'overloaded' retries.

  • Harness buys Augment Code's Cosmos agent

    Harness acquired the assets behind Augment Code — its Cosmos coding agent and context engine — to wire AI coding agents directly into CI/CD and deployment. If you use Augment or Auggie, expect it folded into Harness's delivery platform; the marquee 'learn from failed deploys' capability isn't shipping yet.

  • Gemini agents get their own inbox — and can run on Claude

    Google Cloud unveiled persistent Gemini agents that run multi-day tasks and get their own Gmail, Calendar, and Drive storage, with support for multiple models including Anthropic's Claude. It's pitched at enterprise work, but long-lived agents with real storage and a choice of model is the direction every agent platform is heading.