Loading posts…
Stories about building products, engineering and what we learn along the way at Hunpeo Labs.
Stories about building products, engineering and what we learn along the way at Hunpeo Labs.
An AI reply can sound helpful and still get the policy wrong. Follow one customer question to understand how language models work and how to build an assistant worth checking.
Read the story
A small email task, a few useful questions, and some thoughts on helping a team think together and see work through. English first, tiếng Việt bên dưới.
What does Firebase supply, what does Vercel run, and what do you still need to connect? Follow a café app to see the differences, costs and a practical starting choice. English first, Vietnamese below.
AI can do familiar tasks faster. A story about what remains for us to understand, what to learn next, and how to begin with the work already in front of us.
After plenty of AI experiments, I am taking SatsunicGo toward a real rollout. Here is what I am building, why I want customers to start with a conversation, and the context problems I am still working through. English · Tiếng Việt.
Satsunic SEO 2.0.0 is now available on the Chrome Web Store, with a clearer audit workflow, an interactive crawl map and flexible reports. English first; tiếng Việt bên dưới.
So sánh báo giá website bằng phạm vi công việc, chi phí vận hành và trách nhiệm sau bàn giao.
Portfolio cho người chụp ảnh nên giúp khách hiểu phong cách, dịch vụ và cách liên hệ thay vì chỉ chứa nhiều ảnh.
Catalogue và website có giỏ hàng phục vụ hai cách mua khác nhau. Chọn theo cách cửa hàng xử lý đơn thực tế.
Một website có thể giúp khách xem sản phẩm và thông tin cửa hàng dễ hơn, trong khi Facebook và Zalo vẫn là nơi trò chuyện.
Data Structures & Algorithms - Part 01 Before we get into arrays, linked lists, or algorithms, let's start with a simple question: where does data live while a program is running? In this first article, we'll go from an ordinary desk to tiny memory slots, so a line like int[] arr = {1, 3, 5} becomes something you can picture not just syntax you have to remember.
AI coding agents are becoming faster and more capable. But in production, generating working code does not prove that the code is correct, safe, or ready to release. Trust comes not only from a capable model, but from clear specifications, executable evidence, controlled permissions, and independent approval.
AI agents can research, code, test, and review at remarkable speed. The harder question is whether anyone still owns the outcome.
AI Agent Kit did not begin with a dream of replacing people with AI. It began with a much more ordinary question: if I hand part of a codebase to AI, how do I know it is doing the right thing?
Add this blog to your feed reader.