When does an agent need more context—and when does it need a smaller task?
A question that keeps showing up while building multi-stage AI systems.
Room for
the unfinished.
Experiments, questions, and things I’m still trying to understand. They don’t all need to become a company.
Enter the interactive workshopThe same interactive room that welcomes you home.Questions worth
keeping open.
AI, maths, and the occasional philosophical tangent.
What does a wrong answer tell us about how someone learns?
The thought behind exploring adaptive maths practice with Lydoo.
Is building something a way of understanding it?
It’s the thread connecting most of what I try.
Commerce experiment · Sep – Dec 2025
CruxlyTech
Build the store.
Find the customer.
I sourced smart gadgets, built a Shopify storefront, and worked on organic acquisition. It taught me how demand, margins, and distribution shape a small business.
3US sales · no ad spend
A small, real outcome from putting a storefront in front of customers.
Learning concept · Sep 2025 – Jan 2026
Lydoo
Different answers.
Different next steps.
I explored an adaptive learning concept for college-level maths: a practice path that responds to previous answers, with hints and explanations along the way.
The example here is a small local illustration, not a live AI tutor or a claim that the full product has launched.
See the demo repositoryTry the learning loopIllustrative demo
Linear equationsFind x
2x + 3 = 11
The next step changes with your answer.
A small, local example of adaptive practice. Not the Lydoo product or a live AI tutor.
Other things I’ve tried.
Water Gator
Early robotics experimentA remote-operated lily-pad cleaner made from a repurposed RC car. A first encounter with buoyancy, mechanics, and the difference between an idea and a physical prototype.