A few weeks ago, I picked up a Mac mini for a very specific reason: I wanted a box I could leave running, close enough to “infrastructure” to be dependable, but still personal enough to feel like mine. That machine ended up becoming the home for OpenClaw, and OpenClaw has ended up being one of the more interesting shifts in how I think about local AI agents. Not because it’s perfect. It absolutely isn’t. ...
Last night, I was sprawled on the sofa watching some low-quality UFO’s-are-real show on Prime when an idea hit me: what if I built a simple app to experiment with paywalls and tiered subscriptions? Not because I had a grand business plan, but because I was curious how long it would take to spin up something functional from absolute zero. Three hours later, Paint Vault was live on Replit—a minimalist hobbyist tool with Stripe-powered subscriptions and Clerk authentication. ...
Touchdown back in the UK after a week of navigating the neon-soaked chaos of Las Vegas. If re:Invent 2025 proved anything, it is that the industry has finally stopped talking about what AI might do and started grappling with what it actually does when we let it off the leash. A quick heads-up: These are just my personal thoughts based on the sessions I managed to squeeze into and the big announcements that everyone was buzzing about. It’s not a list of every single update, but rather a look at the “agentic” shift I felt everywhere from the Venetian to Caesar’s Forum. ...
Back in Rhodes, piña colada in hand and my feet dipped in the pool, I wrote about building yet another algo-trading framework. That post chronicled the shift from overengineered C++ and Go experiments to a more pragmatic, productive Python setup, powered by GitHub Copilot and AI pair-programming. It was lightweight, testable, and already backtesting a simple RSI trailing stop strategy via Alpaca’s market data API. Since then, that experimental repo has evolved into something far more structured: a FastMCP 2.0-compatible server hosting two composable tools: get_historical_data and run_backtest. These tools, abstracted from my core logic, form the backbone of a modular and agent-friendly trading system. They allow for decoupled execution paths, better observability, and—most importantly—plug-and-play access from LLMs. ...
As I write this, the Aegean sun glints off the water and my daughter cannonballs into the pool for the third time this morning. We’re on a family holiday on the beautiful Greek island of Rhodes—an ideal mix of history, relaxation, and, somewhat unexpectedly, a productive coding sprint. Because yes, once again, I’ve found myself writing yet another algo-trading framework. I think this must be the third or fourth time I’ve gone down this rabbit hole. ...