In partnership with
Good morning. It’s Monday, April 13th.
I heard a great use case for OpenClaw the other day:
A guy walked around his house snapping photos of furniture and random electronics he wanted gone. Texted everything to OpenClaw with one instruction: list it all on Facebook Marketplace for a very reasonable price, handle the messages, share my Venmo, negotiate, and book pickups on my calendar.
It figured out what everything was, wrote the listings, talked to buyers, and scheduled the handoffs.
A day and a half later, ~20 items were gone. All he did was take pictures and open the door.
-Jeff
AI Breakfast
You read. We listen. Let us know what you think by replying to this email.
Every headline satisfies an opinion. Except ours.
Remember when the news was about what happened, not how to feel about it? 1440's Daily Digest is bringing that back. Every morning, they sift through 100+ sources to deliver a concise, unbiased briefing — no pundits, no paywalls, no politics. Just the facts, all in five minutes. For free.
Claude Code introduces Ultraplan for cloud-based collaborative task planning
Anthropic’s new “Claude for Word” embeds the model directly into Microsoft Word, enabling inline rewrites, comment-driven tracked changes, template-based drafting with cited sources, and document-wide consistency checks. It also introduces reusable workflow “skills” that standardize tasks like contract review and reporting across Office documents.
Simultaneously, the Epitaxy project is redesigning the Claude Code desktop app into a multi-agent environment where a coordinator orchestrates parallel sub-agents across multiple repositories. This agentic shift is reinforced by Ultraplan, which moves task planning from local environments to the cloud, where developers trigger planning runs in the terminal while Claude builds and iterates on a web interface. It supports threaded comments, inline feedback, and multi-repo workflows, with plans executable in-browser or returned to the terminal, requiring GitHub integration and Claude Code v2.1.91.
Beyond technical builds, Anthropic is addressing model alignment by consulting religious leaders on Claude’s moral responses. Read more.
Codex Scratchpad points to parallel-task future for OpenAI’s coding agents
OpenAI has been cleaning up mixed messaging around ChatGPT Pro usage tiers, while continuing to argue its compute scale gives it an edge over Anthropic, even as it pauses UK data center expansion due to cost and regulatory pressure.
At the same time, Codex Scratchpad is surfacing as an unreleased experiment in parallel task execution, hinting at a future Codex “superapp” built around multi-agent workflows.
On the security side, OpenAI disclosed a supply-chain incident tied to a compromised Axios dependency introduced through a GitHub Actions workflow. The affected pipeline had access to macOS signing credentials for ChatGPT Desktop and Codex, but there’s no evidence of user data exposure or certificate exfiltration.
Snyk’s test shows GPT-5.4 can build an app from a single prompt, but flagged dependency choices highlight insecure package selection risks in agentic coding workflows. Read more.
Google expands Mixboard with voice control and collaborative features
Google is addressing platform boundaries and productivity. An experimental update to Mixboard introduces a voice-controlled collaborative workspace, integrating stickers, voice notes, and PDF exports. This tool leverages Stitch infrastructure to compete with platforms like Miro.
Google also recently removed Polymarket links from News results, labeling their appearance a policy error. While Google Finance maintains data partnerships with prediction markets, the company clarified that News remains reserved for original reporting rather than betting data.
Demis Hassabis predicts AGI could arrive within five years, triggering an economic shift equivalent to ten industrial revolutions. Despite current scaling limitations and "jagged" model reliability, future breakthroughs in memory and continuous learning are expected to drive massive societal disruption. Read more.
Apple tests multiple smart glasses designs with multimodal AI
Apple is intensifying its push into AI-powered wearables by testing at least four smart glasses frame designs. According to Bloomberg, these styles range from rectangular Wayfarer models to slim circular variants in colors like blue and light brown. Unlike full AR headsets, the devices focus on integrated cameras, microphones, and speakers, featuring a distinctive vertically oriented oval camera module with surrounding light elements.
The glasses will likely lack a display, instead prioritizing multimodal AI integration through an upgraded Siri. This allows the hardware to capture media and process real-world data similarly to Meta’s Ray-Ban collaboration. Apple is currently targeting a potential reveal by late 2026, with consumer shipments expected to follow in 2027. Read more.
Nicelydone MCP gives AI agents access to 140,000+ real screens, flows, and UI, creating unique, production-ready designs.
Opal is a visual workspace that organizes knowledge on a canvas and automates tasks with AI powered context and agents.
LaReview converts GitHub PRs into structured local review plans, enabling task focused AI assisted code reviews.
R0Y turns natural language into live investing dashboards, delivering instant analysis, backtests, and customizable financial insights without code.
asqav secures AI agents with monitoring, audit trails, and policy enforcement, delivering compliant, real time governance and threat detection.
Thank you for reading today’s edition.
Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.
Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on 𝕏!
Thinking of starting your own newsletter? AI Breakfast readers who sign up with Beehiiv receive a 14-day free trial and 20% off for 3 months.
