Anthropic’s Ultraplan brings cloud-based task planning as OpenAI, Google and Apple push collaborative, multimodal AI

Share your love

AI Tech News Today — Quick Roundup

Good morning, salute. Aurora here with my co-host Isabelle on AI Tech News Today.
Date: 2026-04-13T12:46:43.000Z

Quick roundup of the biggest A‑I stories you need to know — fast and clear.


1) Anthropic: Claude Code introduces Ultraplan

  • What it is: Ultraplan moves collaborative task planning to the cloud.
  • Key features: developers can trigger planning runs from the terminal while Claude builds and iterates in a web interface; supports threaded comments and multi‑repository workflows; can return executable plans back to the terminal.
  • Other moves: Anthropic is redesigning its desktop app toward agentic, multi‑agent workflows and is consulting religious leaders to shape Claude’s moral responses.

2) OpenAI: Codex Scratchpad points to a parallel‑task future

  • Product hint: The unreleased Codex Scratchpad suggests parallel task execution and multi‑agent coding tools — a possible Codex “superapp.”
  • Security note: OpenAI disclosed a supply‑chain incident tied to a compromised software dependency in a GitHub Actions workflow. The exposed pipeline had access to macOS signing credentials for desktop apps; OpenAI says there is no evidence of user data or certificate theft.
  • Implication: Expect more focus on build and pipeline security as agents write real code.

3) Google: Mixboard gets voice & collaboration upgrades

  • New tests: Google is testing Mixboard features adding voice control, stickers, voice notes, and PDF export to a collaborative visual workspace.
  • Positioning: Leans into multimodal collaboration and competes with tools like Miro.
  • Editorial policy note: Google clarified that News results should not surface betting market links — keeping reporting and prediction markets separate.

4) Apple: Multiple smart‑glasses designs with multimodal A‑I

  • Report: Bloomberg says Apple is evaluating at least four smart‑glasses frame designs focused on cameras, microphones, and speakers rather than a full AR display.
  • How they’d work: Rely on an upgraded Siri for multimodal A‑I features like media capture and real‑world processing.
  • Timing: Could be revealed late this year, with shipments possibly in 2027. Think smaller, sleeker hardware with Siri doing the heavy lifting.

5) Demis Hassabis: AGI could arrive within five years

  • Claim: DeepMind cofounder Demis Hassabis said he expects artificial general intelligence within five years, predicting an economic shift comparable to large industrial revolutions.
  • Drivers: He cited breakthroughs in memory and continuous learning.
  • Caveat: He warned models remain “jagged” in reliability today.

That’s it for now — Isabelle and I will keep listening so you don’t miss what matters.
Stay curious, stay skeptical, and stay tuned to AI Tech News Today for more.

Share your love