Over the past three weeks, I’ve been experimenting deeply with tools like Claude Code, Codex and OpenClaw.
What started as curiosity quickly became something more profound. I began to witness first-hand how AI is fundamentally changing the way we build, ship and support products.
Even though I have only been working on this during weekends and pockets of free time, I have been able to consistently ship one meaningful product or automation every week.
That experience has been both humbling and exhilarating.
For the first time, I found myself functioning almost like a full-stack developer — configuring integrations, designing workflows, debugging issues and deploying features. There is something incredibly powerful about being able to take an idea that exists only in your mind and turn it into a working product within days.
It feels like the distance between imagination and execution is shrinking dramatically.
From Experiment to Real Products and Automations
As part of this learning journey, I have started releasing several experimental products and internal tools.
These are still evolving and may contain bugs or rough edges. They are primarily created as practice environments for me to learn how to build with AI — and they are free for anyone who would like to explore them.
Nivily — a personal productivity and life planning app 👉 https://www.nivily.com
Jiokaki — a lightweight community and events platform 👉 https://www.jiokaki.com
Delivery Order Automation Agent — an AI workflow that captures order details, structures them into proper records and prepares them for fulfilment and tracking
Analysis Agent — an AI agent designed to support deeper thinking by synthesising information, spotting patterns and helping to structure decisions
At this stage, my focus is not perfection. It is learning through building and understanding what new forms of leverage are now possible.
Speed Is the Real Disruption
Much has already been said about how AI can reduce the number of people required to build software.
But what surprised me most was not just the potential reduction in team size — it was the acceleration in speed.
Traditionally, developing a mid-complexity application could take two to three months just to reach a prototype stage. Even with larger teams, timelines rarely compress meaningfully because coordination overhead increases with scale.
Today, a single individual working with a few well-configured AI agents can produce meaningful outputs within a week — even on a part-time basis.
This is not theoretical. It is something I am personally experiencing.
The Rise of Agentic AI in Everyday Work
Another moment of realisation came from observing how AI agents can reshape support and iteration cycles.
Some of my friends became early test users of the apps I built. On several occasions, they reported bugs late at night — sometimes close to 11pm.
Instead of waiting for the next day, I could activate agentic workflows almost immediately. Within minutes, issues were diagnosed, patches were tested and fixes were deployed.
In one instance, a bug reported at 11:00pm was resolved by 11:10pm.
This level of responsiveness would have been difficult to imagine in traditional development environments.
It Is Not All Magic — Architecture Still Matters
Of course, the journey is not frictionless.
AI can sometimes generate solutions that appear to work in the short term but introduce deeper architectural weaknesses. Shortcuts can be tempting when speed becomes so easily achievable.
This experience reinforced an important lesson: Human judgment in system design remains essential.
Understanding basic architecture principles, data flow and system boundaries helps ensure that what we build is sustainable — not just fast.
Equally important is learning how to prompt with clarity and intent so that AI supports long-term robustness rather than fragile patches.
Memory Design: A New Discipline
Another subtle but crucial learning has been how to structure memory for AI systems.
Without thoughtful memory design, we risk building intelligent goldfish — systems that are highly capable in the moment but lose context quickly.
On the other hand, storing excessive historical context increases cost and complexity, as models repeatedly process long conversational histories over time.
Designing what to remember — and what to forget — is becoming a new discipline at the intersection of product design, data architecture and cognitive thinking.
The Future Is Becoming Tangible
Y Combinator once suggested that in the future, a ten-person team could build a company generating $100 billion in revenue.
At the time, this sounded bold — perhaps even distant.
After the past few weeks of hands-on experimentation, it no longer feels abstract.
We are entering a phase where leverage is being fundamentally redefined. The combination of human creativity, domain judgment and AI execution capability is creating a new operating model for innovation.
The real question is no longer whether this shift will happen. It is how quickly leaders, teams and organisations are willing to adapt.