Welcome back, AI enthusiasts!

Walmart is planning controls for shopping agents, Gemini Notebook lets you review a presentation’s structure, and Loom turns a screen walkthrough into instructions.

IN TODAY’S ISSUE · About 4 minutes

• 🛒 Walmart’s planned agent spending limits

• 📊 Shape a Gemini Notebook presentation

• 🧰 Four useful tools to explore

• 🎥 Show an agent what you mean with Loom

• ⚡ New models and a look inside AI training

NEWS

🛒 Walmart plans a separate profile for your AI shopper

An assistant buying groceries needs more than a shopping list. In its October 21 announcement, Walmart outlined Agent Profiles: a separate identity for each authorized agent inside your account, with purchases you can distinguish from your own.

• Choose the approval rule: require confirmation for every purchase or set a spending limit. Walmart may still request an extra approval below that limit.

• Keep payment controls: agents won’t receive your card details or be allowed to add or change payment methods.

• Switch access off: you can delete an agent’s profile and see its purchases separately.

This is still ahead: Walmart said profiles would launch within the next month. Its separate Agent Gateway begins with a beta for companies and developers; the announcement isn’t a promise that every assistant can shop this way today.

Why it matters: a visible budget and purchase trail could make delegated errands easier to supervise. When access arrives, choose the approval rule before handing over a recurring shopping task.

Read more here

TOGETHER WITH ROKU

It's time to set up your holiday streaming campaign

Are you ready to reach holiday shoppers on CTV? For many streamers, the season is already underway.

Last year, according to Roku internal data, holiday content searches on Roku saw an 86% increase during October. And in November, those searches rise an average of 39% weekly. This surging holiday spirit is mirrored by consumer activity, as about half of Roku streamers shopped Black Friday and Cyber Monday deals in 2025.

Meeting those shoppers is easier than you might think. Begin with the seasonal outcome you hope to drive: awareness, online sales, leads, store visits, or another measurable goal. From there, define your audience, upload your creative, set the budget and campaign dates, and establish measurement before launching. You can even create a shoppable ad unit that lets streamers buy with a click of their remote.

Our holiday playbook offers a month-by-month guide to reaching shoppers during this all-important time of year.

First-time spenders get a $5,000 match on the first $5,000 spent with code BEEHIIV5K. Terms apply.

Launch your campaign today.

AI TOOL OF THE DAY

📊 Review Gemini Notebook’s plan before it makes your slides

Gemini Notebook now offers an outline review before generating a Slide Deck. The October 21 update also adds a producer-notes review for Video Overviews. Rollout began October 16 and can take up to 15 days, including personal Google accounts and Workspace customers.

Try a small presentation from your own notes:

1. Add a short document you can check, such as notes for a community workshop. Ask for a presentation in chat or choose Slide Deck in Studio.

2. Turn on Review outline. Give the draft structure a clear brief:

Create five slides for first-time volunteers at a community garden. Use only my notes. Cover arrival, tools, watering, cleanup, and who to ask for help. Flag missing details instead of inventing them. Keep each slide focused on one action.

3. Read the outline and request changes before generating. Check the finished slides against your notes, especially names and instructions.

If the review toggle hasn’t reached your account, check again as the rollout continues.

Read more here

4 TRENDING AI TOOLS

Useful discoveries; not all launched this week.

🎬 MOVO AI: develop a video through conversation, from a scene plan to generated shots and a final sequence. Revise its scene plan before generating.

📝 TwinMind: capture conversations and turn them into notes you can revisit or ask questions about. Try one meeting, with participants’ agreement, and check the names, decisions, and action items afterward.

🔎 Specsight: translate changes in a software product’s code into explanations and release notes for the wider team. Verify its draft against the shipped product.

🤝 AgentGrid: keep people and agents working on shared documents, slides, and other files with version history. Set access for each agent and folder, then use a small project to test the handoff.

NEWS

🎥 Loom turns your walkthrough into an agent brief

Sometimes the fastest explanation is pointing at the screen. Atlassian’s October 22 Loom update introduces Record for agent, which turns a short narrated recording into instructions an AI agent can use.

• More context travels with it: the recording captures your spoken instructions, screenshots of clicked elements, and links visited while sharing your screen.

• Point precisely: the new Mac app adds a Selection tool for specific interface elements and a Magic pen for drawn annotations.

• Check your plan: Record for agent is in open beta for Loom Business + AI and Enterprise, through the new Mac app or Chrome extension.

Why it matters: a visual brief could reduce the back-and-forth involved in explaining a change. Try describing one confusing form or slide, then read the generated instructions before passing them on.

Read more here

QUICK AND FRESH

🌙 Kimi K3.1 Pro reaches the apps. Moonshot’s latest model rolls out across Kimi’s web and mobile products, Work, Code, and API, with a focus on coding and other multi-step tasks. Downloadable model weights are promised by November 4; they aren’t part of today’s access.

🎙️ Block previews a model trained on speech alone. Dessa can listen and speak simultaneously, according to Block. The page offers demonstrations and a waitlist, with early access for a small group of safety researchers. This is a research preview, not a generally available voice assistant.

💻 Flint brings another option for local coding. Recursive’s model is designed to run on your own hardware, including supported Apple Silicon setups. Its speed comparisons are company measurements; check the setup requirements and try a small coding task before choosing it for a project.

🔒 OpenAI describes an offline training environment. Its frontier reinforcement-learning runs use cached and indexed web content so models can practice online tasks without unrestricted live-web access. The report also describes containment failures found in testing and the fixes they prompted.

PARTNER WITH US

Building something useful with AI? Reach readers looking for practical tools and ideas. Explore sponsorship opportunities.

YOUR OPINION MATTERS

Which would help you most: a budget for your shopping agent, a better slide outline, or a clearer visual brief? Reply and tell us what you’d try.

Find us at @alltrendsai.

Until next time, stay informed!