My first real week building in Bangkok: AI bugs, secret project and perfect weather
End of November, 2025. The setup phase is over. Time to actually work.
This week marked my transition from "getting operational" to "being operational". Client work, personal projects, and a weather shift that changes everything.
Client work: when AI gets too creative
I’m currently finishing a project involving an image search system for a French client. I write the specifications, a web agency develops the platform. The concept: type a complex text query like "Find me all photos of Jean Dujardin with George Clooney in tuxedos at their latest film premiere" and get exactly that from a massive database of celebrity photos.
The tech stack? OpenCLIP for visual embeddings, Amazon Rekognition for face detection, Elasticsearch for search. Sounds solid on paper.
Reality? AI models have opinions
The system works beautifully... until it doesn't. False positives everywhere 🤯 Ask for two specific celebrities together, get photos with one but not the other. Ask for "in tuxedos", get formal events that match visually but aren't the right context. The AI sees patterns humans don't, and misses patterns humans do…
The challenge isn't making it work, it's making it work accurately. There's a massive difference between "technically functional" and "actually useful for real users".
With the dev team we’re currently fine-tuning thresholds, adjusting similarity scores, and testing different combinations of face detection + visual similarity + text understanding. It's the unglamorous part of AI development that nobody talks about: endless iterations to get from 60% accuracy to 95%.
But that's the job. Client projects help me build my runway so I can eventually work on my own projects.
My secret SaaS: the plot thickens
I can't tell you what I'm building. Not yet 🤫
What I can tell you: it's an SEO tool. Python-based. Heavily AI-assisted development. I'm using Claude Opus 4.5 to write code, debug issues, and architect solutions faster than I ever could alone.
The project needs data. Lots of it. I'm spending most of my time figuring out how to collect, structure, and process information at scale.
The technical challenges are real, but that's exactly what makes it interesting. Every day brings new problems to solve, new edge cases to handle, new optimizations to implement.
Why the secrecy? Because ideas are worthless until execution. I've learned this the hard way with my previous startups. Talk less, ship more.
The beta version is taking shape. Slowly. Very slowly. But it's happening.
More details coming when I have something to show. For now, just know that I'm coding every day, learning constantly, and building something that I know others need 🤓
Bangkok winter: the game changer
Official start of "winter" in Thailand.
And it's glorious ☀️
Humidity dropped from 80% to 45% overnight. The air feels breathable again. Temperatures hit 18°C at night. Cold by Thai standards, perfect by mine!
Here's what changed practically:
Air conditioning usage dropped by 70%. I'm running it maybe 2 hours a day instead of 12. My electricity bill will thank me. Laundry actually dries now. In the rainy season, clothes would stay damp for days, developing that weird mildew smell. Now? Dry in 4 hours, fresh and clean.
Mornings and evenings are comfortable for walking. The oppressive heat that made every outdoor movement feel like swimming through hot soup? Gone. I can actually explore the neighborhood without melting 😂
The city feels more energetic. People are out. Rooftop bars are packed. Street food vendors stay open later.
The downside? Tourist season has officially started. Sukhumvit is noticeably more crowded. Popular restaurants need reservations. The BTS is packed during peak hours.
But honestly? I'll take the crowds if it means breathing comfortably and sleeping without AC blasting all night.
Bangkok in winter is what everyone talks about. Now I understand why!
One sad note though: my favorite British couple, Danielle and James, are currently in Japan. No more spontaneous beer meetups or restaurant recommendations for the next few months. I'm not crying, you're crying 😢
This is what building a SaaS business from Bangkok actually looks like. Client work that pays the bills. Personal projects that take forever. And weather that finally cooperates.
No comments yet. Be the first to comment!