Weekly AI Highlights: April 11 – 17, 2026
What Small Business Owners Need to Know Right Now
Reader, what a week.
Started it with a quick trip to Arizona with my husband. Short and sweet, the kind of getaway that ends before you're fully unpacked but somehow still resets you. Got home, caught my breath for about five minutes, and jumped into kicking off a new CAIVA cohort. This group of VAs is sharp, and watching them dive into AI with real curiosity is honestly one of the best parts of what I do.
And while all of that was happening, the AI world did not slow down for a second.
Anthropic dropped Claude Opus 4.7. Google finally gave Mac users a real Gemini app and turned Chrome into a place where you can save prompts with one click. OpenAI came out swinging at Claude Code.
Let's get into it.
1. Claude Opus 4.7 Is Here
Anthropic released Claude Opus 4.7 on Thursday. It's a solid step up from 4.6 and it's the model you'll now see when you open Claude.
A few things that actually matter for how you use it day to day:
- It follows your instructions more literally. If you tell it "no emojis" or "keep it to three sentences," it listens. Less filling in the blanks, less drifting off-task.
- It can see images way better. You can now upload high-resolution screenshots, scanned documents, and detailed diagrams, and it will actually read them properly.
- It checks its own work before answering. Fewer "wait, that's wrong" moments on complex tasks.
Early testers included Intuit, GitHub, and Notion, and they reported real improvements on production work.
Source Link: https://www.anthropic.com/news/claude-opus-4-7
MY TAKE
I haven't gotten to test 4.7 much yet since it just dropped, but knowing Claude and how it just keeps on giving, I already know this is going to be good. And if 4.7 is the version they're comfortable letting us loose with, I can only imagine what Mythos (the bigger sibling they're keeping locked up for now) is actually capable of.
2. Chrome Now Lets You Save AI Prompts as One-Click "Skills"
Google rolled out Skills in Chrome. If you use Gemini inside the Chrome sidebar, you can now save any prompt you've written as a reusable shortcut. Next time you need it, type a forward slash, pick the Skill, and it runs on whatever page or tabs you have open.
The examples Google gave: calculating protein macros on a recipe site, comparing product specs across shopping tabs, summarizing long documents. Google is also shipping a Skills library with pre-built ones if you don't want to start from scratch.
Source Link: https://blog.google/products-and-platforms/products/chrome/skills-in-chrome/
MY TAKE
I really like this one. I use Chrome every single day, and anything that saves me from retyping the same prompt over and over is a win. This is great.
ACTION STEP (DO THIS WEEK)
Think about the one prompt you've typed into Gemini more than three times this month. Save it as a Skill. That's the whole homework.
|
3. Gemini Finally Has a Native Mac App
Google shipped a real Mac app for Gemini this week, available globally for anyone on macOS 15 or later. It's free to download at gemini.google/mac.
You trigger it with Option + Space for a quick chat, or Option + Shift + Space for the full window. You can share any window on your screen with it, drag in local files, and keep working in other apps without tab-switching. ChatGPT and Claude have had Mac apps forever. Gemini was the last holdout.
Source Link: https://blog.google/innovation-and-ai/products/gemini-app/gemini-app-now-on-mac-os/
MY TAKE
Finally. Now I'm curious how soon Google ships their own version of Claude Cowork, because if they do, it could be just as good as Claude Cowork or even better. Looking forward to that one.
4. OpenAI Gives Codex Desktop Control to Go After Claude Code
OpenAI announced a major Codex upgrade on Thursday. Codex can now open apps on your desktop, click, type, and run multiple agents at the same time while you work on other things. It also got an in-app browser so it can operate inside web apps, and a new "memory" feature that recalls past sessions and learns how you work.
This is OpenAI directly chasing Claude Code, which Anthropic updated last month to remotely control your Mac.
Source Link: https://openai.com/index/codex-for-almost-everything/
MY TAKE
You're probably not writing code, so this might feel like someone else's battle. Watch it anyway. The agent-controls-your-computer pattern being built for developers is the same one that's about to show up in your marketing tools, your inbox, and your CRM. The plumbing is being laid right now.
PROMPT OF THE WEEK
Audit Your Own AI Setup (The Declutter Prompt)
What it's for: If you've been using Claude, ChatGPT, or Gemini for more than a couple of months, you've probably been adding custom instructions, saved prompts, project rules, and context files every time the AI did something annoying. "Be more concise." "Use a casual tone." "Don't use em dashes." Three months in, some of those instructions are contradicting each other, and your AI is trying to follow all of them at 60%.
When to use it: Any time the outputs feel off, bloated, or nothing like what you asked for. Also worth running as a spring cleaning every few months.
What to expect: A list of instructions to cut, any conflicts between rules, and a cleaned up version of your setup. I tested this on my own Claude project and deleted about a third of my instructions. Everything since has sounded more like me, not less.
The prompt:
|
Read my entire setup before responding. Check my custom instructions, every saved prompt, every project file, every context document, and any other instruction files you can find.
Then go through every rule, instruction, and preference. For each one, tell me:
- Is this something you already do by default without being told?
- Does this contradict or conflict with another rule somewhere else in my setup?
- Does this repeat something that's already covered by a different rule or file?
- Does this read like it was added to fix one specific bad output rather than improve outputs overall?
- Is this so vague that you'd interpret it differently every time? (example: "be more natural" or "use a good tone")
Then give me:
- A list of everything you'd cut with a one-line reason for each
- A list of any conflicts you found between files
- A cleaned up version of my main instructions with the dead weight removed
|
How to run it:
- In Claude: paste it into a Project that has your custom instructions and any knowledge files loaded.
- In ChatGPT: run it in a chat where you've got custom instructions enabled, or inside a Project.
- In Gemini: run it inside a Gem where your saved context lives.
Don't blindly delete what it flags. Read the reasons first. Then cut the ones that make sense, run your three most common tasks, and see if anything breaks. If it does, add that one rule back. You're looking for the smallest set of instructions that still gets you the output you want.
Source: The AI Solopreneur
|
Comments
If any of this made you think differently about how you want to work with AI, hit reply and let me know. I read every message. Your questions shape what I cover next.
Have a great weekend,
Founder, Front Row AI Club
PS. Was this newsletter forwarded to you? Sign up to get it in your inbox.
You can find my previous newsletter articles here: The Front Row AI Report
© 2026, Front Row AI Club. All rights reserved.