Over the past few years, I’ve spent most of my time focused on family life. Raising young children, managing a household, writing, reading, and trying to navigate a period of significant personal and professional transition.
At the same time, I’ve kept one ear to the ground on developments in AI and software. Rather than chasing every new tool or trend, I’ve been more interested in finding opportunities to explore these technologies in grounded, immediately practical ways. Most of the experiments collected here began not as business ideas or product concepts, but as ordinary frustrations, curiosities, or small inefficiencies encountered in daily life.
The projects below are a collection of these explorations (I will expand on their details in separate posts over time):
- We track our toddler’s sleep using Cubtale and have become quite dependent on it. The problem was not collecting the data, but making use of it. Simple calculations like wake windows either had to be done manually or required paying for premium features. Using OpenClaw and a few AI-assisted workflows, I created a local mirror of the sleep data by translating screenshots into structured records. Once the data was under our control, I could calculate wake windows, experiment with different analyses, and explore questions that the original application was never designed to answer.
- Every so often, a tweet I vaguely remember liking becomes relevant to a conversation. Despite having “saved” it years ago, finding it again usually meant scrolling endlessly on my phone or relying on crude browser searches. I first solved this by extracting nearly 30,000 liked tweets from an existing archiving tool using a small AI-assisted script, then making them searchable locally with a lightweight command-line workflow. Later, after running into limitations with that approach, I used AI-assisted coding to build my own capture pipeline directly. What began as a search problem gradually turned into a broader exercise in data ownership and portability.
- I wanted a way to capture conversations from my messaging platforms (e.g. Google Chat) that preserved quoted messages, threads, attachments, and context. The official export processes are often slow and pulls in more than I need, while screenshots lose too much information and are just cumbersome from a data perspective. AI dramatically lowered the cost of exploring the problem, allowing me to prototype a browser-based export tool that converted otherwise ephemeral conversations into durable knowledge assets.
- While reading to my son, I wanted to upload the book to ChatGPT in order to critique and discuss it more precisely (later in my own time, not as I’m reading to him 😆). A digital copy existed, but the platform intentionally provided no way to download it. AI helped me quickly find a way to extract the content and made it available for further analysis.
- I wanted to add a simple image carousel to a Notion page in a note that I shared with some friends and family. For some reason, this is not native to Notion. Existing third-party solutions either required ongoing subscriptions, came bundled with features I didn’t need, or demanded far more setup than the problem justified. Rather than adopting an entire platform, I used AI to create a focused solution tailored to the specific experience I wanted.
- I have been writing a lot and discovering a writing process that fits me. One of the techniques I’ve been playing with is mind-mapping. While organizing ideas in Obsidian (my note-taking platform), I wanted to transform a nested outline into cards in a mind map. Manually creating dozens of cards would have interrupted the flow of thought far more than the actual problem warranted. AI made it practical to build a one-off script that converted structure into visualization and let me stay focused on the thinking itself.
- I became curious about how much useful work could be done from my phone while away from my desk. One experiment involved transcribing Vietnamese video content using specialized language models… while walking my dog. Another involved configuring and debugging software systems through an AI agent over chat. The interesting result was not the transcription itself, but discovering how much the boundary between intention and execution has begun to dissolve.
None of these problems are particularly large. Most are things that people simply tolerate. What interests me is that recent advances in AI have changed the economics of problem-solving. The value is not merely that software can be produced faster. It is that many ideas that were previously not worth pursuing can now be explored, tested, and validated before they are forgotten or abandoned.
While these experiments are personal, I suspect the underlying patterns extend far beyond my own life. Every organization accumulates friction. Information gets trapped. Processes become repetitive. Small annoyances compound into meaningful costs. The same instincts that help uncover opportunities in a household, a personal workflow, or a side project can often be applied to larger operational and organizational challenges as well.