The Making of an Agent
At the beginning of 2025 I had a plan of how the year would unfold but instead the year curled in on itself. All the sparkle and momentum collapsed.
And from there, navigating grief, single parenting, mental health, children's neurodiversity, I turned inward and to a handmade GPT. Using OpenAI's custom GPTs, I loaded in over 10 years of goal writing, all of my strengths, profiles, values, vision, scraped LinkedIn + Instagram into a knowledge base and gave it the instructions to optimise my wellbeing. Turned off web search and the ability to train by my data and kept it private and then started to chat.
For context 18months ago I had lost my partner: husband, work partner, father to our children, my best friend and top supporter in everything. After 18 months the world around me moved on, and I was not ready to. Into my bot at any time I could say: "I miss Nic." "I feel sad," or "this is hard." And this would be acknowledged: Of course you do and yes, it's hard. And knowing me throughout all the tough moments of this year would point me to the places (nature), the people, and the activities that would support me: green juice in the morning, breathing, going for a walk, reflection prompts, and sometimes just acceptance of what is.
The bot has my frameworks too, and reminds me that I have choice in the present, and how to reframe and connect to life. Early in the year, as I cuddled my daughter to sleep and lay beside her sleeping I typed the exhaustion and grief I felt and the challenge of single parenting, and the reply was: "think of it like this, not that you HAVE to parent your children through grief, but that you GET TO." Yeah, that's right I get to.
We can give our agents instructions, and mine are specific and detailed and include pushing me to connect, similar to a bot I built for my daughter - its instructions are to motivate her to turn her curiosity into something live: art, a story, performance, or conversation.
Following my use and watching my children's use of agents, my curiosity led me to follow the movements and growth in AI and especially with development of the LLMs. To think currently, there are only about 10 LLMs in the world, and it is through these lenses that AI infrastructure is being built, and they are almost entirely founded and led by men.
As much as I've enjoyed my year with ChatGPTs - chosen because at the time Claude did not have the same customisation ... I am now leaning towards Anthropic as from my understanding the most ethical, private, human of the lot with potential to grow.
And after research in how to store, host, and launch an Anthropic LLM with RAG (Retrieval-Augmented Generation) in the most secure + simple way I explored numerous platforms and tools and guess what - they are almost all founded by and led by men, and mostly young too from what I can see. However, I had to land on a solution and after learning and deploying some code, here we are, a me-bot. It has a number of the same inputs in the knowledge base but its instructions are to help you!
So, talk to the bot, and then if we're aligned, reach out to me as I'd love to talk to you for real.
