Good Inside: What If a Parenting Book Could Talk Back?

I recently became a father, and like most new parents, I found myself desperately searching for guidance. A friend recommended Good Inside by Dr. Becky Kennedy, and it quickly became one of those books that shifts how you see your kid — and yourself.

The core idea is simple: your child is good inside, even when their behavior says otherwise. Tantrums, defiance, whining — these aren’t signs of a bad kid. They’re windows into what a child is feeling and struggling to express. Dr. Becky reframes the parent’s job: instead of fixing behavior, build connection. Reduce shame. Tell the truth. It sounds obvious on paper, but in the middle of a meltdown at 2 AM, none of it comes naturally.

The book is packed with practical frameworks — how to handle hitting, sibling rivalry, separation anxiety, picky eating, lying — all grounded in the same principle: connection first, behavior second.

The idea that got me excited

Here’s what kept nagging at me as I read: this book is incredible, but no sleep-deprived parent is going to flip through 29 chapters to find the right advice in the moment they need it most.

What if you could just tell something about your situation — “my toddler won’t stop hitting his sister and I’m losing it” — and get back a thoughtful, Dr. Becky-style response? Not a generic parenting tip, but something grounded in the actual frameworks from the book, tailored to your specific problem.

Large language models are just starting to feel capable enough to pull this off. ChatGPT has been out for a little over a year, and people are beginning to experiment with giving LLMs domain-specific knowledge. The idea of building an AI parenting coach — one informed by a specific, evidence-based philosophy — feels both novel and genuinely useful.

I’m thinking about a tool where new parents could describe their challenge in plain language, and an LLM grounded in the principles from Good Inside would respond as a coach. Not a replacement for reading the book, but a complement — an interactive way to access its insights when you need them, in the context you need them.

This still feels like a pretty unconventional idea. RAG pipelines are clunky, prompt engineering is more art than science, and most people haven’t yet thought of LLMs as something you can point at a specific body of knowledge and get reliable, nuanced advice from. But I think there’s something genuinely interesting here at the intersection of AI and parenting.

I’m not sure if I’ll ship it, but the idea is exciting — building tools that meet people where they are, not where a textbook assumes they’ll be.