AI DUNN Right Weekly - Issue #22
- Feb 23
- 6 min read
Practical AI insights for business growth
Hey AI Innovators! 👋
What a week. Google just shook up the reasoning leaderboard, Anthropic handed Claude users a major upgrade at no extra cost, and an app builder hit $100M in revenue in 8 months by letting anyone build software from their phone.
Here's what you need to know:
Google's Gemini 3.1 Pro more than doubled its reasoning performance while keeping the price exactly the same
Claude Sonnet 4.6 can now hold a million tokens in one conversation, still at Sonnet pricing
A startup called Emergent reached $100M annual revenue in 8 months by making app-building as easy as having a conversation
WordPress.com built AI directly into its editor, which could shake up the entire plugin industry
AI companies are choosing subscriptions over ads, and that signals something important
Read time: 5 minutes
Big Story
Google just quietly took the reasoning crown
Here's what most people missed this week...
Google dropped Gemini 3.1 Pro and it didn't come with a lot of fanfare. But the numbers tell a different story.
The benchmark that matters here is ARC-AGI-2. Most AI benchmarks test memorized answers. This one throws brand new logic patterns at the model and measures whether it can actually reason through something it has never seen before.
Gemini 3.1 Pro scored 77.1%.
Anthropic's Opus 4.6 scored 68.8%. GPT-5.2 scored 52.9%.
That's not a small gap. And Google did this without raising the price by a single cent. Same cost as before. Same 1 million token context window. Just dramatically better reasoning.
Here's what this means for you...
The AI competition is working in our favor. Every time these companies fight over the top spot, we get better tools without paying more.
But here's the part I find most interesting. Google isn't trying to win on flash. They're winning on capability baked into tools people already use every day. Gemini is inside Gmail, Docs, Slides, and Search. You don't have to change your workflow. The intelligence just shows up where you already are.
That's a strategy worth paying attention to.
Gemini 3.1 Pro is available now in the Gemini app, Google AI Studio, and Vertex AI.
What's New This Week
Claude Sonnet 4.6 is now your default and it's a big upgrade
Anthropic released Claude Sonnet 4.6 this week and made it the new default for both free and paid users on claude.ai.
Here's what actually changed...
Coding got noticeably better. Former users prefer it over the previous version about 70% of the time.
Computer use improved significantly too. It can now handle complex spreadsheets, fill out multi-step web forms, and work across multiple browser tabs without losing context.
The context window hit 1 million tokens in beta. That means you can drop an entire contract, a full research file, or a lengthy email thread into one conversation and Claude holds onto all of it.
Pricing? Unchanged. Free and paid plans both got the upgrade automatically.
If you already use Claude, you already have this. Worth testing on whatever task you've been struggling to get right.
An app builder hit $100M revenue in 8 months
A platform called Emergent just hit 100 million dollars in annual recurring revenue. In 8 months.
Here's what makes this worth paying attention to beyond the headline number...
Emergent is not a tool for developers. You describe what you want in plain language, autonomous AI agents build it, and you end up with a real working application. Backend, database, testing, deployment. The whole thing. No coding required.
This week they also launched a mobile app, so you can now build and ship software from your phone.
Over 6 million builders across 190 countries have created more than 7 million applications on the platform.
The bottom line is this: software development is getting cheap and fast. If you've had a business idea that required an app and couldn't afford to build one, that excuse is getting very small very quickly.
WordPress.com built AI directly into its editor
WordPress powers 43% of all websites. This week they added a native AI assistant that lives inside the editor itself.
No plugins. No switching tabs. Just type @ai and tell it what you want.
Change your site colors or fonts in plain language
Rewrite your bio to sound more confident
Add a new section or page without touching a menu
Translate content into another language
Generate or edit images using Google's AI models
It's opt-in right now and works only with block themes. But the direction is clear: website editing is becoming conversational. The entire plugin industry built around doing these things manually is now feeling some pressure.
Tool of the Week
Kael
Kael is an AI tool built specifically for people who need to work across multiple documents at once.
Most AI tools handle one document at a time. Kael handles dozens simultaneously. You upload a collection of PDFs, contracts, research papers, or reports and ask it to find themes, contradictions, connections, or comparisons across the entire set.
Here's a real example that got my attention: a legal team analyzed 40 contracts, over 2,000 pages, in a fraction of the time it would normally take. A PhD candidate synthesized 60 research papers in about 4 hours and got back a structured document mapping 8 major themes with contradictions flagged and citations included.
Free tier covers up to 10 documents. Pro plan is $39 per month.
Best use cases: research synthesis, contract review, competitor analysis, policy comparison, or any situation where you're sitting on a pile of documents and need clarity without spending two weeks reading everything manually.
Quick Hits Worth Your Time
→ Perplexity announced it's rejecting ads inside its AI platform. Instead of selling ad slots, they're betting on subscriptions and trust. When an AI company chooses trust over ad revenue, pay attention. It tells you where they think the real value is.
→ Runway raised $315 million to double down on AI video generation. Their Gen-4.5 model is currently ranked number one for realistic AI video. If you've been avoiding video content because it feels too expensive or complicated, this space is moving fast.
→ An Anthropic safety researcher publicly resigned this week, saying the field is moving too fast and that upholding safety values internally is harder than it looks from the outside.
Worth being aware of.
→ Someone built a fully functioning Roblox game in 4 days using Claude Opus 4.6. No traditional coding. Just describing what they wanted. The gap between idea and execution keeps closing.
→ Reddit is testing AI shopping tools that convert community recommendations into shoppable product carousels with prices and links. If you sell products, start tracking where AI is showing up in the purchase journey.
→ A Fed governor said this week that an AI scenario where many workers become "essentially unemployable" is "totally possible." Upskilling now is the most practical response to that risk.
Prompt of the Week
Act as an AI learning coach.
I want to build a simple daily habit for getting better
at using AI tools without adding more to my plate.
Essential Details:
- Current AI experience: [BEGINNER / SOME / CONFIDENT]
- Main tool I use: [ChatGPT / Claude / Gemini / Other]
- Time I can realistically spend: [5 / 10 / 15 minutes per day]
- Biggest struggle: [PROMPTING / KNOWING WHAT TO USE IT FOR / TRUSTING THE OUTPUT]
- My job or business: [WHAT YOU DO]
- Goal: [WHAT YOU WANT TO DO BETTER WITH AI]
Design one daily AI skill-building habit including:
1. A 5-minute daily practice (specific, repeatable, low effort)
2. One weekly challenge that builds real confidence
3. Three prompts tailored to my job that I can use this week
4. A simple way to track whether I'm actually improving
5. What to do when I feel stuck or frustrated
Make it feel doable, not overwhelming.
Why this works: Most people give up on AI not because the tools are hard but because they don't have a system for getting better. This prompt builds a personalised on-ramp that fits around what you're already doing. Five minutes a day beats one overwhelming workshop every three months.
My Take
The question Kyle Balmer raised in his community this week stopped me in my tracks.
Is AI making us dumb?
He called it the illusion of competence. You get the output. You skip the understanding. And over time, you lose the ability to think through things yourself.
Here's where I land on this, from someone who teaches AI to people who are just starting out.
The risk is real. If you hand every task to AI and never engage with the material yourself, you're producing but not learning. That's a problem.
But here's the other side. For the people I work with, the ones in their 40s, 50s, and 60s who were never going to write code or read 60 research papers manually, AI is not replacing their thinking. It's giving them access to capabilities they never had before.
The question is not whether AI is making us dumb. The question is whether you're using it to skip the understanding or to deepen it.
Used well, AI is a thinking partner. Used poorly, it's a crutch.
That choice is yours every single time you open the app.
That's it for this week.
What are you using AI for that's actually saving you time right now?
Hit reply and let me know.
Jackie
Comments