Hi {{first_name}}!

This week we’re covering several meaningful AI updates from OpenAI, Google, DeepSeek, and others.

As we kick off the year, the honeymoon phase of AI is officially over. We’re moving from simply chatting with bots to building systems that actually do things. That shift shows up everywhere, from OpenAI’s new hardware experiments to Google’s push into physical robotics.

This week we're covering:

  • New AI developments from Google and OpenAI, including industry-specific prompt packs worth testing

  • The Brain Dump Framework, how I use automation in ClickUp to make sure no idea ever falls through the cracks

  • Claude Skills vs. Projects, and why this distinction matters more than most people realize

OK, let's get into it!

New and Noteworthy

  • OpenAI Launches Industry-Specific Prompt Libraries: OpenAI just released ready-to-use prompt packs for sales, HR, marketing, and other business functions. Instead of starting from scratch every time, you can now grab proven prompts tailored to your department and start getting better results immediately.

  • Google Teams With Boston Dynamics on Industrial Robots: Google is bringing its AI capabilities to Boston Dynamics’ hardware to build smarter industrial humanoid robots. This matters because the gap is closing between AI that can reason and machines that can do physical work in warehouses, factories, and job sites. Robotics are evolving fast, and humanoid robots are quickly moving from novelty to reality.

  • OpenAI's Audio-First Device Coming: Reports suggest OpenAI is building a screenless, audio-based AI companion expected within about a year. Think less smartphone, more intelligent assistant you talk to naturally. Worth watching if you’re tired of living behind a screen.

  • Google Bringing AI Video to Your TV: Google plans to integrate Nano Banana and Veo into Gemini on Google TV. You’ll be able to generate AI videos and modify family photos using voice commands. This is rolling out first on select TCL TVs before broader deployment.

  • OpenAI's "Gumdrop" Smart Pen: OpenAI’s first consumer hardware device, codenamed “Gumdrop,” is heading into production with Foxconn. Early designs suggest a smart pen or portable audio tool with a microphone and camera that captures handwritten notes and uploads them directly to ChatGPT. If you still take notes by hand, this could bridge the gap without changing your workflow.

  • Google's Gemini and Veo Models Top the AI Rankings: Google recently claimed top spots across language, vision, and video benchmarks. If you default to ChatGPT for everything, it’s worth testing Gemini for visual analysis and video or image generation. For businesses, this means more viable options when choosing which tools to build into workflows.

  • DeepSeek's Next Efficiency Breakthrough: DeepSeek published research on a technique called “mHC” that improves training stability and efficiency without increasing compute costs. Tests showed major gains, especially in reasoning tasks. Last year, DeepSeek reached near-frontier performance at a fraction of the cost. This signals more powerful AI at lower prices for businesses in 2026.

A Simple Automation That Stopped My Best Ideas From Disappearing

Here’s a problem I used to have. Great ideas would hit me at random times. During a client call. While scrolling LinkedIn. Driving home. And by the time I sat down at my computer, half of them were gone.

I tried voice memos, notes apps, notepads, even texting myself. Nothing stuck, because there was no system. Ideas just piled up in different places with no follow-through.

Maybe this is just my ADHD brain at work, but if this sounds familiar, I’m hoping this helps you too.

A few days ago, I built something super simple, and I think I finally fixed the issue.

I created a form called “Brain Dump” with four fields:

  • What’s on your mind?

  • Category

  • Due date

  • Attachments

I added it as a shortcut on my iPhone home screen. Now when an idea hits, I open the form, type it in, and hit submit.

Here’s where the automation kicks in.

Each submission is automatically routed to the right person based on category.

  • Content idea? It goes to our content manager.

  • Potential partnership? It lands in my business development tasks, assigned to me, with a due date.

  • Website change or competitor insight? It goes straight to my assistant, often with a screenshot attached.

The idea is captured, assigned, and already in motion. I don’t think about it again.

You’re no longer relying on memory or manual sorting. The system handles the logic. If this happens, do that. If status changes to X, notify person Y. You set it up once, and it runs forever.

This is worth testing if you’re tired of good ideas disappearing before you can act on them.

I built this in ClickUp in about 15 minutes, but you could do the same thing in Monday.com or any project management tool that supports forms and basic automation. If you’re new to ClickUp, the free version covers most small team needs. Paid plans start at $7 per user per month if you want more advanced features and integrations.

Make 2026 Count: Upskill With Claude Skills

I want to highlight a Claude feature that most business owners still don’t realize exists. It’s called Skills, and it’s especially useful if you do any kind of repeat work with AI.

First, let’s clear up the confusion. Skills, Projects, and Custom GPTs all sound similar, but they work very differently.

Claude Projects

Projects are workspaces. They organize conversations and files around a single topic or client. Claude remembers context inside that project only.

Great for long-term research, client work, or anything that needs a dedicated space.

Claude Skills

Skills are reusable behaviors. You teach Claude how you want something done once, and it applies that workflow across every conversation, anywhere in Claude.

They are not tied to a project. They shape how Claude works for you by default.

The easiest way to think about it:

  • Projects are like folders

  • Skills are like muscle memory

If you’ve used Custom GPTs in ChatGPT, the difference is subtle but important. Custom GPTs are separate assistants you have to intentionally open each time. Claude Skills quietly influence every conversation without you switching tools or contexts.

Once a Skill is set up, you stop re-explaining yourself.

How to set up a Skill

  1. Go to Settings → Capabilities → Skills

  2. Enable the skill-creator skill

  3. Start a new chat and say: “Help me create a skill for [something you do repeatedly]”

  4. Claude walks you through the rest

Where Skills really help

  • Proposal writing using your exact structure and pricing logic

  • Content that consistently matches your brand voice

  • Client communication that follows your tone and standards

  • Research done with your preferred criteria and focus areas

  • This is one of those features that quietly compounds.

People who set this up now will save hours every week, while others keep retyping the same instructions.

If you use Claude regularly, this is absolutely worth doing.

That’s a wrap on the first issue of 2026. Any takeaways you’re planning to implement? I’d love to hear from you.

See you next week,

Julien

PS: If you’re finding these deep dives helpful, please share this with a friend or colleague navigating the 2026 AI landscape. They can catch up at www.ampra.ai/join-our-newsletter.

Keep Reading