Steal 6 of My Claude Skills: AI Update #2
Here's everything that mattered in AI last week + 6 of my Claude Skills
Last week, I launched AI Update as a new section with a simple promise: cut through the AI noise and deliver what actually matters.
You guys loved it. The response showed me something important:
You don’t just want to know what happened. You want to understand which developments actually change how things work.
So this week, I’m doing a deep dive on the AI news that matters most, Claude Skills, then covering the news you should know.
Dazl: The New AI Prototyping Tool Purpose Built for PMs
I’ve been advising this small company in Israel from a former Wix co-founder, and they’ve just gone live on ProductHunt today. They’re currently #1. Vote for them.
And check them out. Unlike other tools, they allow you to edit what you’re seeing:
Get 1 month free with code: DazlXPG
When I talk to people using Claude, the consistent pain point I hear is:
I know Claude skills exist. But I don’t know what skills to actually build. I need examples that show me what’s possible, and how to use them.
So, I’m sharing 6 skills I built that you can steal. Welcome to: skills that actually save you time. If you want to build your own skill or access existing ones, last week’s deep dive covers everything.
I’m doing 2 things with this:
Showing you what’s possible when you stack skills together
Giving you the exact skills so you can start using them today
This is going to be broadly applicable to anyone who does repetitive work.
Now then - let’s get into it!
1/ LinkedIn Post Writer
Takes your rough idea and turns it into something 10% away from publishable. Uses my advice on LinkedIn post writing.
2/ Prompt Engineer
This skill uses research-backed techniques (Chain-of-Thought, Multi-Shot, structured formats). Shows you before/after. Explains why the new version works better.
3/ Agent Workflow Skill
You can automate your boring routines. Just describe the task. It figures out which tools to use. Then it builds the entire workflow strategy based on my advice.
4/ Idea Validator
Enter your idea and get a clear analysis of whether it solves a real problem, whether there’s demand, whether it’s feasible to build, and what could go wrong.
5/ PRD Writer
Simply paste an existing PRD and it finds the gaps. It tells you what’s broken and how to fix it based on my PRD philosophy.
6/ Product Designer
This skill makes your design consistent. It finds out what’s off, what’s cluttered, what needs work, and then shows you how to fix it.
How to actually use these skills together
Most people jump straight to building. These skills help you move faster through the entire cycle.
Start with validation:
Idea Validator stress-tests if it’s worth building
Then move to execution:
PRD Writer defines what you’re building
Prompt Engineer builds the system instructions
Agent Workflow automates the process
Product Designer ensures it looks professional
Then share:
LinkedIn Post Writer turns your work into posts your network sees
And they all work together in Claude.
What takes 8 hours now takes 2. You stop asking “which skill do I use?” and start asking “what’s my biggest bottleneck?” Then you use the skill (or stack) that eliminates it.
That’s it for today’s deep dive, now here’s what mattered most in AI this week:
Top News
Canva just built its own AI model
Canva dropped a massive redesign this week. But buried in the announcement is the real story: they built their own foundation model specifically trained on design.
This isn’t some off-the-shelf LLM. Canva Design Model understands layout, color balance, typography - the stuff that matters for actual design work. They’re also shipping faster video editing, a new marketing platform called Canva Grow, and smarter data collection for forms.
Why this is different: Canva could have used an existing LLM. Cheaper. Faster. But they didn’t. Because generic models don’t understand design the way domain-specific models do.
The shift: Canva moved from “tool you use” to “infrastructure you depend on.”
With 180M+ users, they now own the design AI layer. Every design decision, every user preference, every failure teaches their model. That moat is impossible for competitors to catch.
What this signals: Domain-specific AI is winning.
The Rest of What Mattered in AI This Week
News
Perplexity will pay Snap $400M to power search in Snapchat
Anthropic is projecting $70B in revenue and $17B in cash flow by 2028
Google’s AI Mode got new agentic capabilities to help book event tickets and beauty appointments
Resources
Walkthrough of Cursor 2.0
Get great results with Veo 3.1
Every single product OpenAI is working on
New Tools
Monologue: effortless voice dictation by every
MagicPath: design UIs with AI, then edit and export code
Fundraising
Inception raised $50 million to build diffusion models for code and text
AI startup Metropolis, biggest parking lot network in U.S., raised $1.6 billion for major retail expansion
That’s all for today. Reply with your thoughts, and see you next week,
Aakash