My checklist for a personal AI
A peek into how LLMs act as my personal AI today.
Welcome to the 12th edition of Perplexed by Jaga.
I built this digital garden to share everything that has perplexed me. If it’s your first time here, you can find older posts here and read more about me here.
I have always dreamed of owning a personal robot, and speaking to an LLM feels like we’re almost there.
Here’s my four-step checklist for the ideal personal AI 👇🏻
First - understand the “young me.”
Where did I grow up?
What was my childhood like?
Where did I go when I was younger?
What did I experience when I was growing up?
Who were my friends, parents and peers when I was growing up?
There is some proof that your childhood experiences shape you as an adult.
If that is true, I want my personal AI to understand the
The different ways in which it provides me with an advantage
How does it make me blind to opportunities & aspirations?
and how it biases my goals
Second - understand the “current me.”
Where do I want to go?
What do I want to achieve?
What kind of relationships do I want?
What kind of things do I want to build?
What kind of places do I want to visit?
What kind of family/friendships do I want to build?
Now that AI understands the past and future, I want it to suggest a pathway for the present.
Third - suggest action items.
Suggest diet plans that will help me reach my goals
Suggest improvements in how I approach both personal & professional relationships
Suggest areas of improvement and prevent me from self-sabotage (it is a thing, apparently)
I don’t just want my personal AI to tell me more about what I already know.
I want it to help me discover the things I don't even realize I should know—my unknown unknowns.
If a certain kind of diet will help me reach my fitness goals faster and I don’t even know about it, I want my AI to tell me. Humans are bad at thinking from first principles because it has a higher cognitive load.
Fourth - refine and suggest new action items.
I want my AI to ask me questions daily, observe and refine my daily life.
For example, my personal AI should understand why I keep missing those fitness sessions even when I keep saying fitness is one of my top priorities.
Then, it should either suggest action items to make the habit practical or have a challenging conversation with me to explain how fitness is not my biggest priority (at least right now).
💡 Suggest action items and keep me focused, or bring up non-goals and nudge me in a new direction.
This is where I see a hardware play that tracks everything I say, eat, and see. It functions like a ghost that encapsulates the nuances of my life, including things that I fail to mention to my AI or bring up things I am not even aware of.
Today, my personal AI’s feedback cycle is limited to what I tell it. But humans are imperfect at reflecting and self-correcting.
That’s why we have a saying in behavioural psychology - “Look at what they do & not what they say”.
My experience with LLMs as a personal AI
I’ve been using AI to improve my life both personally and professionally.
Today, I use AI as a thought partner, and the outcomes I can extract from a Grok or a Claude are limited by my ability & patience to prompt well.
Of course, there is a privacy concern. There is also concern about LLMs misrepresenting facts, hallucinating, and giving you the wrong advice.
I am wise enough to understand that it is also a technology prone to mistakes.
But LLMs as a personal AI have solved a nuanced problem for me. Earlier, if I had a problem, I first had to 👇🏻
Google the problem
Read through all solutions
Find the right solution which fits my context
Sometimes, I might also have to spend time understanding the problem well.
The process is time-consuming, and most answers were non-specific. Also, once I exit the search, the trail is lost. Google doesn’t store my nuances or context; it only helps me find answers.
But an LLM stores all this context inside that chat window. The constraints and nuances are also considered for any follow-up questions I ask.
Plus, if I can prompt it right, I can get the LLM to think from first principles and offer solutions I haven’t even considered yet.
This is the actual delta with a personal AI.
Even for writing this post, I have a chat saved with a writing expert. I shared the draft with it, and it tells me where my writing is weak and good.
Interestingly, it pulled references from an older input and added that to the output.
It feels delightful - an actual Delta 4 experience.
Again, I understand that errors are common and LLMs often hallucinate. But I have seen 10+ cases where LLMs already feel like personal AIs.
Writing & content creation
Navigating through personal struggles
Navigating through career choices and business questions
I constantly open Grok, go to a specific chat window, and converse with AI like a mentor or friend. It has my context, is nonjudgmental, and is super quick. It feels like we are already in the age of personal AI.
The only thing missing is the feedback cycle, which depends on me to input context into a specific chat window.
The ideal personal AI
A software layer that captures everything I do on my smartphone, laptop or digital device
A hardware layer that captures everything I do and say offline - speech, steps, habits, consumption & everything else
I want an LLM that is stored locally and doesn’t transfer my data back to one of the big tech companies for product development/feedback, etc.
Fin.