AI DUNN Right Weekly - Issue #17

When the leader cuts everyone else off

Here's what happened this week that reveals where AI competition is really headed...

Anthropic pulled the rug out from under xAI and several third-party tools.

Third-party apps that let users route their Claude Code subscriptions through alternative interfaces? Blocked.

More embarrassing... xAI engineers were using Claude Code through Cursor until Anthropic shut off access this week.

Remember when Elon said his team uses Grok internally because it's the best coding model? Yeah, that was a lie.

The internal memo from xAI's co-founder Tony Wu is telling: "I believe many of you have already discovered that Anthropic models are not responding on Cursor... We will get hit on productivity."

Here's what I think...

This drama only exists because Claude Code is the best coding agent available right now. Period.

When you're behind, you champion openness. When you're on top, things change. Anthropic just proved that competitive advantage matters more than open ecosystem promises.

For business owners, this reveals something critical about AI tool decisions in 2026...

Don't build your entire workflow on someone else's platform if you can't control access. The vendor can change rules overnight. Your productivity tanks. Your team scrambles.

The companies winning in AI right now aren't playing nice. They're protecting moats. And that changes how you should evaluate which tools to adopt.

---

What's New This Week

Google's video model just leapfrogged competitors

Google dropped Veo 3.1 this week with features that put it ahead of OpenAI's Sora and China's PixVerse.

The big upgrades:

• Native vertical video (9:16) for YouTube Shorts and mobile without cropping

• 1080p and 4K upscaling for professional-grade quality

• Ingredients to Video that maintains character identity across scenes

• Seamless blending of textures, characters, and objects
  

Available now in Gemini app, YouTube, Google Vids, and Vertex AI.

Why this matters...

While OpenAI spent two years hyping Sora, Google just embedded video generation across its entire ecosystem. YouTube creators can generate Shorts directly in the app. Enterprise teams can create videos in Google Vids.

Distribution beats features. Google doesn't need the flashiest model when they control where billions of people already create content.

For businesses creating video content, this means professional video creation is moving from specialized software to everyday tools. The barrier isn't quality anymore. It's knowing what to create.

Apple and Google just changed Siri's future

Apple announced a multiyear partnership with Google this week. Google's Gemini models will power Apple Foundation Models, including a major Siri upgrade coming later this year.

Apple also launched Creator Studio... a $12.99/month subscription bundle with six professional creative apps plus new AI features.

The bundle includes Final Cut Pro, Logic Pro, Pixelmator Pro, Motion, Compressor, and MainStage.

New AI features across the board:

• Transcript search in Final Cut Pro

• Visual search to find exact moments by describing them

• Beat detection for automatic music sync

• Generate presentation drafts from text in Keynote

• Natural language formulas in Numbers
  

Here's what this reveals...

Apple delayed their Siri AI upgrade from 2025 to 2026. Now we know why... they were negotiating the Google partnership.

Google already pays Apple billions each year to be the default search engine on iPhones. Now they're powering Siri too.

The irony? Google's search monopoly was ruled illegal last year. That threatened their lucrative deal with Apple. Instead of losing ground, they expanded the relationship into AI.

For business owners, this signals where AI assistants are headed... foundation models from Google or OpenAI, wrapped in Apple's interface, running on your devices.

The competition is not about who builds the best model anymore. It's about who controls the interface people already use daily.

Claude for Healthcare wants your medical data

Days after OpenAI revealed ChatGPT Health, Anthropic announced Claude for Healthcare.

It's different from ChatGPT's patient chat experience. Claude goes deeper... built for providers, payers, and patients.

The platform connects to medical databases:

• Centers for Medicare and Medicaid Services (CMS) Coverage Database

• International Classification of Diseases (ICD-10)

• National Provider Identifier Standard

• PubMed
  

Designed to speed up prior authorization, research, and report generation.

Why this matters now...

Doctors spend more time on paperwork than seeing patients. Prior authorization alone takes weeks and requires pulling codes from multiple databases.

Claude for Healthcare can automate this. Pull relevant codes from ICD-10. Check CMS coverage. Generate authorization requests.

The risk? LLMs hallucinate. They generate confident responses that are sometimes completely wrong. That's dangerous in healthcare.

Both Anthropic and OpenAI warn that users should see healthcare professionals for reliable guidance. The AI is supplemental, not a replacement.

But here's reality... OpenAI says 230 million people already talk about their health with ChatGPT each week.

Healthcare AI is becoming a battleground. OpenAI launched ChatGPT Health. Anthropic responded with Claude for Healthcare. Google will likely announce something soon.

The company that wins healthcare wins massive recurring revenue. But one major error, and the regulatory backlash will set the entire category back.

Wikipedia is licensing content to AI companies

Wikipedia signed AI training deals with Microsoft, Meta, and Amazon this week.

For years, Wikipedia operated as free knowledge accessible to everyone. Now they're monetizing that knowledge by licensing it to AI companies for model training.

This follows the pattern we've seen across publishing... Reddit licensing to Google. News organizations signing deals with OpenAI.

What changes...

AI models need high-quality training data. Wikipedia has curated knowledge across millions of articles. That's valuable.

For businesses, this signals that "free" content on the internet won't stay free for AI training purposes. Copyright holders are waking up to the value and demanding compensation.

The internet's knowledge commons is becoming a paid ecosystem.

---

Tool of the Week

Personal Intelligence makes Gemini actually useful

Google launched Personal Intelligence for Gemini this week.

It connects Gmail, Google Photos, YouTube, and Search with a single tap. Once enabled, Gemini knows you... your preferences, past trips, schedule, even license plate numbers from photos. (scary, but useful)

Real example from Google's VP Josh Woodward:

He needed new tires for his 2019 Honda minivan. Standing in line at the shop, he asked Gemini for the tire size.

Gemini didn't just give specs. It suggested options... one for daily driving, another for all-weather. Then referenced his family road trips to Oklahoma found in Google Photos. Pulled ratings and prices for each tire type.

When he needed his license plate number, Gemini grabbed it from a photo. It identified his van's specific trim by searching Gmail.

That's the difference. Not just answering questions. Taking action with your actual data.

Other use cases:

• Meal planning with automatic ingredient lists added to Amazon Fresh

• Life admin that extracts details from documents and adds appointments to calendar

• Recipe management with dietary customization

• Entertainment discovery that recalls selections on your Fire TV
  

Privacy approach...

Connecting apps is off by default. You choose to turn it on, decide which apps to connect, and can turn it off anytime.

Google doesn't train on your Gmail inbox or Google Photos library. They train on limited info like specific prompts and responses to improve functionality.

This is Google's answer to OpenAI's ChatGPT memory and Anthropic's context windows. Every AI assistant is racing to become personal.

Google has Gmail, Photos, Calendar, Search, Maps. That's a massive data advantage over competitors.

If Personal Intelligence works as promised, Gemini becomes your operating system for daily tasks. Not just a chatbot. An assistant that knows you.

Rolling out this week to Google AI Pro and AI Ultra subscribers in the U.S. Coming to more countries and the free tier later.

Quick Hits Worth Your Time

→ Shopify CEO Tobi built an MRI viewer in minutes using Claude Code that looked better and worked better than expensive commercial software sold to hospitals.

→ Linus Torvalds (creator of Linux) started vibe coding for his hobby guitar pedal audio software, noting "Is this much better than I could do by hand? Sure is."

→ AI drones found a missing 66-year-old man in the Italian Alps by analyzing thousands of images across a massive search area, though unfortunately too late to save him.

→ Launch Lemonade is making AI agent building accessible to non-technical users with no-code workflows that hook into Claude, Gemini, and ChatGPT.

→ AI influencers are booking real brand deals now, working 24/7 without contracts or scandals, creating new entry points for influencer marketing.

---

Prompt of the Week

90-Day AI Implementation Roadmap

Stop experimenting forever. Actually implement AI in your business in 90 days.

Act as an AI transformation specialist. Create one 90-day implementation roadmap for adding AI to [MARKETING AREA] that delivers results fast.

Essential Details...
- Marketing Area... [CONTENT/EMAILS/ADS/SOCIAL/ANALYTICS]
- Current State... [WHAT YOU'RE DOING NOW]
- Goal State... [WHAT YOU WANT IN 90 DAYS]
- Budget... [TOTAL AMOUNT AVAILABLE]
- Team Skill Level... [BEGINNER/INTERMEDIATE/ADVANCED]
- Top Priority... [#1 THING TO ACHIEVE]

Create one roadmap including...

Days 1-30 (Quick Wins)
- Which AI tool to start with
- First 3 tasks to automate
- Team training basics
- Success metrics to track

Days 31-60 (Build Systems)
- Full tool integration
- Workflow changes needed
- Advanced training
- Quality checks

Days 61-90 (Scale Up)
- Optimize what's working
- Expand to more tasks
- Measure ROI
- Plan next phase

Also include...
- Weekly team check-ins
- How to handle resistance
- Monthly progress reports

Why this works...

Most companies stay stuck in pilot mode forever. They test AI tools but never actually implement them into workflows.

A 90-day roadmap forces action. Quick wins build momentum. Systems thinking prevents tool sprawl. Clear metrics show whether it's working.

Get AI working in your marketing in 90 days, not someday.

My Take

The biggest shift this week is not a new tool or flashy model.

It's watching competitive moats replace open collaboration.

Anthropic blocked xAI from using Claude Code. Wikipedia signed exclusive licensing deals. Apple chose Google over OpenAI for Siri's foundation.

This is what happens when AI moves from experimental to essential.

When tools become critical infrastructure, vendors protect access. When content becomes training data, publishers demand payment. When market position matters, companies choose partners strategically.

2025 was "look what AI can do." 2026 is "who controls access to what works."

That changes everything for how you should evaluate AI tools:

• Can you afford to lose access if the vendor changes terms?

• Do you have alternatives if your primary tool gets shut down?

• Are you building on someone else's platform without control?
  

The operators who win this year will be the ones who understand that AI tool decisions are infrastructure decisions.

You're not just picking software anymore. You're choosing which ecosystem to depend on.

And unlike traditional software, AI vendors can change access overnight. Your productivity tanks. Your team scrambles. Your workflows break.

The smart play? Diversify your AI stack. Don't put everything on one platform. Build workflows that can adapt when vendors inevitably prioritize their own interests over yours.

Because that's what's happening. And it's only getting more competitive from here.

That's it for this week!

What AI tool are you most dependent on right now? Hit reply and let me know.

Jackie