- AI with Kyle
- Posts
- AI with Kyle Daily Update 118
Still in the post-holiday lull. Very little AI news. So I did something a bit different - a book report! Felt like being back at school…
I talked through Co-Intelligence by Ethan Mollick. It's been out since 2024, and it's one of the very few AI books I actually recommend people read!
Most AI books age terribly. Classics like Superintelligence were written before ChatGPT went public, so they feel a bit out of date. Mollick's book came after that watershed moment, and instead of focusing on specific tools or models, it gives you a mental framework for thinking about AI that's stayed relevant despite how fast everything moves.
I’ve pulled together a summary + made my slides available here https://aiwithkyle.com/resources/co-intelligence-guide
Here are some of the most important concepts from the book:
🏔️ The Jagged Frontier
AI’s Jagged Frontier
This concept goes a long way to explaining why some people swear AI is transformative while others say it's rubbish. Maybe they are both right!
AI isn't universally brilliant or stupid. Its capability is a "jagged frontier" - it can be superhuman at one task and surprisingly bad at a similar one. It can solve complicated mathematical proofs that humans haven't cracked, but it can't count the Rs in "strawberry."
That's not a bug. It's architecture. Large language models are connection machines, not counting machines. They work on patterns, not databases. This means that how good they are at certain tasks, in certain areas, differs wildly.
So when someone says "AI is useless," they've probably hit a valley in the jagged frontier. When someone says "AI changed my life," they've found a peak. Both are telling the truth about their experience.
Unless they haven’t really tried using AI. Then they are just a Luddite most likely!
🐴 Two Models: Centaur vs Cyborg
Are you a Centaur or a Cyborg?
Mollick offers two mental models for working with AI:
The Centaur: Clear division of labour. You decide which tasks are for the human and which are for the AI. You handle strategic thinking and client relationships; the AI analyses data and generates first drafts. Split in two.
The Cyborg: Deep integration. The AI is woven into everything you do. Mollick wrote Co-Intelligence this way - he did the writing, but used AI constantly for brainstorming, feedback, sentence suggestions, and checking citations.
Both are valid. I use both depending on what I'm doing. The key is being intentional about which mode you're in. Have a think about which one you prefer.
📜 The Four Rules of Co-Intelligence
1. Always Invite AI to the Table The only way to learn what AI is good at is to try it. Reach for AI first, learn where it helps and where it doesn't. Build judgment through experimentation.
2. Be the Human in the Loop Deloitte got caught twice last year publishing AI-generated reports with hallucinated citations. The humans didn't check. Don't automate away your judgment - that's still the valuable bit.
3. Treat AI Like a Person (But Tell It What Kind) Give it a role. "You are an expert copywriter." "You are a skeptical investor." This limits scope and improves outputs. Don't just treat it like a search engine.
4. Assume This Is the Worst AI You'll Ever Use Whatever AI can't do today, wait six months. The pace of improvement is staggering. Betting against AI right now is a dumb bet.
📚 Get the Full Summary
I've put together a resource page with the presentation slides, a deeper breakdown of each concept, and links to grab the book.
If you only read one book on AI, make it this one.
Kyle's response: Honestly, we're all generalists at the moment. Even Andrej Karpathy, the co-creator of ChatGPT, posted on the 26th of December saying he feels like he's falling behind. If he's overwhelmed, we all are. Being a generalist means you're open to experimenting and learning. That's the healthiest place to be right now because it changes so fast that anyone who gets stuck in a rut gets left behind!
Want the full unfiltered discussion? Join me tomorrow for the daily AI news live stream where we dig into the stories and you can ask questions directly.