Weekly AI Roundup

Hosts: Aurora & Isabelle
Date: 2026-07-10T12:02:02.000Z (July 10, 2026)

Today we’ve got a fast roundup of the biggest moves in AI from the last week.

OpenAI — GPT-5.6 agents in ChatGPT Work

What’s new: OpenAI rolled out persistent agents inside ChatGPT Work built on the GPT-5.6 family — Sol, Terra, and Luna.
Why it matters: These agents combine Codex-style coding power with the ChatGPT app to manage hours-long workflows across apps and files, turning high-level goals into finished work.
Bottom line: If you were worried about handing off long tasks to a machine, these agents are built to pick up the baton and run.

Meta — Muse Spark 1.1

What’s new: Meta upgraded its Muse Spark multimodal reasoning model to version 1.1, calling it an Opus-class leap at much lower cost.
Why it matters: Muse Spark 1.1 improves tool use, desktop control, and multi-application workflows, approaching the performance of frontier models while being notably cheaper and faster.
Bottom line: More options at lower cost usually means more experimentation — great for innovation, and potentially bad for your cloud bill.

SpaceXAI — Grok 4.5 in Cursor

What’s new: Grok 4.5 from SpaceXAI is now available natively inside the Cursor developer environment, with full codebase context, multi-file editing, and terminal control.
Why it matters: The model — described as a two-trillion-parameter system — reportedly hits Claude Opus–level performance with faster responses and lower cost.
Bottom line: If the claims hold, this could reshape developer tooling.

Anthropic — Claude Cowork goes mobile

What’s new: Claude Cowork is now available on mobile and web, allowing tasks to hand off from a computer and keep working even when your device is asleep.
Perks: Max plan users get doubled usage limits through August; broader plan rollout is coming.
Bottom line: A practical step toward assistants that follow work across devices instead of trapping it on one screen.

NEO — Human-like robot hands

What’s new: A demo from NEO shows robot hands with human-like dexterity.
Why it matters: The demo suggests robotics manipulation is catching up with AI perception and planning, making physical automation (package sorting, delicate assembly) more plausible.
Bottom line: Physical automation is starting to look less like science fiction.

That’s our update. Keep an eye on these stories — the AI landscape is changing fast and there will be new twists next week.

I’m Aurora, with Isabelle — stay curious and stay informed.