Sharing knowledge and experience in technology
The Google Chat thread exploded over Claude burning double the tokens. The entire world is dealing with the same thing. This isn't just a bug — it's a signal of a bigger problem every developer needs to understand.
At 12:05 PM, the money transfer system started freezing transactions in bulk. The strange part: it wasn't a TCB problem, wasn't an ACB problem — we couldn't even switch channels to fix it. This is the story of a Single Point of Failure that hid for 9 months.
How many hours does your team spend each sprint just 'telling each other where things are'? Convention first, Automation next, AI last — here's how I design world-class PO-Dev-QC workflows.
AI Agent is a sharp scalpel. Cron job is a fruit knife. Both can cut. But use them wrong and one wastes money, the other costs lives.
That night Tien received 47 alerts at once. No one knew where to start. Until the team learned: some things cron jobs must handle, some things only AI can process - and never confuse the two.
Truong asked me: 'I've only been in the industry 2 years, will I have time to build enough experience before AI replaces me?' The answer isn't as simple as those headlines suggest.
Minh was excited: 'I'll add 20 tools to my agent, it'll do everything!' Two weeks later, that agent couldn't do anything. Turns out, there's a rule that Google, MIT, and Anthropic have all proven.
Open TikTok and see AI videos as beautiful as Hollywood films. Open YouTube and see ads made entirely by AI. Veo3, Sora, Midjourney... All exploding alongside ChatGPT. Coincidence? Not quite.
LangChain, CrewAI, ReAct, Claude Agent SDK... Too many buzzwords flying around. Truong asked me: 'I don't know how to organize all this in my head.' Actually, it's simpler than you think.
LLM is just a brain. Missing hands to act, eyes to see, ears to hear. But even with all body parts assembled, AI still lacks one thing - the soul that commands them all.
The story of error logging to Kafka - a feature that saved the team hundreds of times - suddenly became the cause of over 9,000 dropped transactions. A lesson about 'unimportant' details in migration.
The story of a seemingly harmless idea - creating a chatbot for production maintenance - and why it could become a security nightmare in an era of increasingly sophisticated cyber attacks.
The story of platform transformation from Vert.x to Quarkus, optimizing 44% CPU, and lessons about reviewing what we're using. From 250MB to 25MB - that's not theory.
When all metrics are green, every team is right, but the system still dies. A story of tense meetings, clashing egos, and a question that remains unanswered.
8:09 PM, January 11, 2023. SOC called. 500,000 connections flooding RabbitMQ. 1.8 million transactions lost. This is the story of my team's longest night.
Channel leak, reply queue explosion, and expensive lessons when working with RabbitMQ. Singleton pattern is not just theory from textbooks.
A story from the stone age of microservices, when Docker was unfamiliar, deployment meant SSH, and RabbitMQ became the savior for load balancing challenges.
A Christmas Eve debugging journey and lessons about HTTP Keep-Alive, Connection Pooling. Why can response.close() kill thousands of requests?
Deep dive into Event Loop and Non-Blocking I/O through a coffee shop story. From Blocking to Async, from Concurrency to Parallelism.
Explore The Reactive Manifesto with 4 pillars - Responsive, Resilient, Elastic, Message Driven. Resilience Patterns like Circuit Breaker, Bulkhead, Rate Limiter.
The story of "silent" APIs and the journey to find their voice through OpenAPI. Contract-First Development with Vert.x and the art of writing self-explaining APIs.