In partnership with
Hey folks,
Remember when NotebookLM could turn your notes into a podcast? Now it turns them into a documentary. Cinematic Video Overviews replace the old narrated slideshows with fully animated explainer videos generated from your research. Three AI models split the work: Gemini 3 acts as creative director, deciding narrative structure and visual style; Nano Banana Pro generates the images; Veo 3 handles video and audio. Upload a PDF, get back something that looks like it had production input.
Today in AI:
→ GPT-5.4 lets you steer its thinking mid-response
→ Google puts app building inside Search
→ How to build your first app with Gemini Canvas
Let’s dive in..
🧠 GPT-5.4: The Model That Shows You Its Plan Before It Starts
GPT-5.4, which OpenAI launched as its new flagship comes in three flavors: standard, Thinking, and Pro. The standout is GPT-5.4 Thinking. Before it dives into a complex task, it shows you an upfront plan of how it intends to approach your request. If you don't like the direction, you can steer it mid-response without starting over. Instead of getting a 2,000-word answer and realizing it went sideways at paragraph two, you can course-correct in real time.
Here's what actually changes for day-to-day work:
Steerable thinking plans: The model outlines its approach before executing. You review, adjust, and it proceeds. No more burning through tokens on a direction you didn't want
Computer use, built in: GPT-5.4 can operate your computer natively. It navigates websites, fills out forms, chains actions across apps. It writes code via Playwright and issues commands from screenshots
1M token context window: The largest ever. Feed it an entire codebase or a multi-hour transcript and it holds the whole thing in memory
No single feature here is the point. It's that GPT-5.4 treats you like a collaborator instead of a requester. You brief it, it proposes a plan, you approve or redirect, then it executes.
For your team: If you've been frustrated by AI going off-track on complex tasks, the steerable thinking plan is worth testing. It changes the workflow from "prompt and pray" to "brief, review, approve."
Together with Hubspot
The Future of AI in Marketing. Your Shortcut to Smarter, Faster Marketing.
Unlock a focused set of AI strategies built to streamline your work and maximize impact. This guide delivers the practical tactics and tools marketers need to start seeing results right away:
7 high-impact AI strategies to accelerate your marketing performance
Practical use cases for content creation, lead gen, and personalization
Expert insights into how top marketers are using AI today
A framework to evaluate and implement AI tools efficiently
Stay ahead of the curve with these top strategies AI helped develop for marketers, built for real-world results.
🎨 Google Just Put App Building Inside Search
Google rolled out Gemini Canvas in AI Mode. Describe an idea and Canvas generates a functional app, game, or prototype right there in a side panel. You type "make me a project tracker with drag-and-drop columns" and it builds one.
The wild part is where it lives. It's in Google Search. The same place 4 billion people go every day. Google is betting that the next generation of app builders won't open VS Code, they'll open a browser tab.
The workflow is simple: enter AI Mode in Google Search, tap the Canvas option from the (+) menu, describe what you want, and watch it generate. You can test the prototype and refine it by chatting with Gemini in the panel.
It's early, and the apps you'll build aren't production-ready. But for internal tools, quick prototypes, or testing an idea before investing engineering time is good for now.
For your team: If product managers or designers on your team regularly need quick prototypes to validate ideas, point them at Canvas. It won't replace your eng team, but it could eliminate time of "can you build me a quick mockup?"
🎯 How To Build Your First App With Gemini Canvas
Canvas lets anyone build a working prototype from Google Search. Here's how to try it:
Open Google Search — Go to google.com on desktop (Canvas works best on desktop, mobile support is limited)
Enter AI Mode — Type a question or creative prompt into Search to activate AI Mode. You'll see the AI Mode interface appear
Open Canvas — Click the (+) tool menu in AI Mode and select Canvas
Describe your app — Type what you want to build. Good starting prompts: "Build a habit tracker with streaks," "Make a quiz app about world capitals," or "Create a kanban board for project tasks"
Test and refine — Your app appears in the side panel. Click through it to test functionality. Toggle the code view to inspect what's under the hood. Chat with Gemini to make changes: "Add a dark mode toggle" or "Make the cards draggable"
🐝 AI Buzz Bits
🔧 Google open-sourced a CLI for its entire Workspace suite. One command-line tool gws for Drive, Gmail, Calendar, Sheets, Docs, and 50+ other APIs. Ships with a built-in MCP server, 100+ agent skills, and a --sanitize flag that scans responses for prompt injection.
🏛️ Pentagon formally brands Anthropic a "supply chain risk." First US company to receive the designation, previously reserved for adversary nations like Huawei. The irony: the military is still using Claude in its Iran campaign. Meanwhile, over a million people are signing up for Claude every day.
⚖️ A NY judge ruled Claude-generated legal docs are fully discoverable. In a fraud case, the defendant fed privileged info from his lawyers into Claude, generated 31 strategy documents, and tried to claim attorney-client privilege. Judge Rakoff's ruling: Claude isn't a lawyer, Anthropic's privacy policy allows data disclosure, and nothing you tell an AI chatbot is confidential. If you're using AI to prep legal strategy, this ruling says none of it is confidential.
🛠 Tool Spotlight
Unblocked — Context layer that gives AI agents team-specific understanding from your code, docs, planning tools, and conversations. Your agents generate code that matches your standards instead of generic boilerplate.
AgentCenter for OpenClaw — Mission control dashboard for monitoring and managing OpenClaw agents in production. Real-time visibility into runs, workflows, and performance. Add a few API calls to connect existing agents, core logic stays untouched.
Scoutflo — AI SRE that investigates incidents and fixes issues in production. Knows your services, infrastructure, and dependencies. Designed for on-call engineers who'd rather sleep than debug at 3am.
For a full list of 1500+ AI tools, visit our Supertool Directory
👉 Know someone drowning in AI news? Forward this to them or send your unique referral link
Cheers, Tim
