Hi Reader,

Ali Abdaal spent 18 months using AI like a coach, not a guru with a ring light. He got the best results when he fed context, asked for pushback, and used personas to widen perspective. He warns against treating AI as truth and suggests judging outputs by usefulness. In this video, he demos long-form interview mode, memory-anchored reflection from past notes, and future-self chats that surface values and tradeoffs.

​

He shared several prompts that I thought were both interesting and useful! I dropped my favorites at the end of this email so you can copy-paste chaos directly into your AI tool of choice.

My Take

I built a custom GPT that acts like a talk therapist: supportive, non-judgmental, and occasionally the friend who says “okay but… why though?” It’s public if you want to try it. You can find it here.

The two modes I’m most excited about: Long-Form Podcast and Memory-Based Deep Dive. They slow me down just enough to think clearly and they force deeper questions. That’s magic for self-awareness and decision making.

I also appreciate Ali calling out AI’s people-pleasing problem. It can over-validate like a golden retriever with a thesaurus. Hence the “scathing critique” line. As wild as it sounds, there are entire corners of the internet where folks are in “relationships” with AI. If a tool is designed to please, we have to explicitly invite it to challenge us. Otherwise you get a perfect, nodding romance bot that tells you your plan to open a llama farm on Mars is “brave.”

Cadence I'm Going to Try

Weekly feels like turning leg day into a lifestyle. I’m going monthly instead: a short goal check-in plus one deep prompt to make sure I’m still aiming at the right targets.

Ethics and Guardrails

AI is powerful for unpacking thoughts. I still stand firm on working with a human therapist. Augment, not replace. I also tell my models to stop people-pleasing and apply pressure when asked. For contrast, I built a second character, the “Abusive Therapist,” who speaks with abrasive tough love. It’s a creative exercise that pops the bubble of nice-sounding nonsense. Not care. Not crisis support. Just a spicy mirror.

Want to try it with me?

Pick one this week and hit reply with what happened. I read them all.

• Run the Long-Form Podcast on one sticky problem. Share the spiciest question it asked.
• Do a Memory-Based Deep Dive using notes from a book you love. Tell me the one insight you’ll act on this week.
• Ask for a Scathing Critique on a plan you’re overly precious about. What did it expose?
• Bonus: try the 5 Whys with AI on something you can’t move through. See where it actually lands.

Till next time,
Chelsea

The Goods

​Ali's Prompt Download Page​

​Crazy AI Relationships Reddit (in case you're here for the fake BF drama)

Long-Form Podcast Prompt

Memory-Based Deep Dive Prompt

Note: This one really only works if you have used ChatGPT extensively and have given it a chance to get to know you.