AI Insights
History Rhymes: 70 Years Ago, We Didn't Trust Machines to Write Good Code Either
Seventy years ago, programmers didn't trust compilers to generate usable code either. Today's trust crisis around AI-written code is nearly the same story playing out again. From FORTRAN's origin story to the cost-driven organizational shift and the rise of software engineering itself, history has already given us the answer.
2026-08-10
$5 for a Week of Work: How We Use AI to Capture Fragmented Enterprise Knowledge
A real story from cloud migration and disaster recovery support: how AI-My-Chats uses email as a unified entry point and multimodal models to understand context, turning fragmented conversations into traceable, reusable enterprise knowledge — processing 100+ records a month for about $5 in model costs.
2026-08-02
Why 80% of a Company's Most Valuable Information Ends Up Lost in Chat Logs
A company's most valuable information rarely lives in documents or knowledge bases — it's scattered across chat logs. We re-examined the whole pipeline, from collection and comprehension to standardization and workflow integration, and open-sourced Devify to turn the real information a company produces every day into fuel that can actually drive AI.
2026-07-14
How AI Made Us Rethink Team Collaboration
A real internal experiment: the same product, built by a four-person team versus one person with AI. The results surprised everyone — and forced us to rethink how we think about division of labor, collaboration, and judgment in the age of AI.
2026-06-30
Software Engineering in the AI Era: The Bottleneck Is No Longer Development, It's Quality
Drawing from interview observations of three types of developers, this article explores how AI is restructuring software engineering: the bottleneck has shifted from generation to quality, testing is evolving into quality engineering, and product thinking is becoming every engineer's core competency.
2026-06-27
Mrs. Wang's AI Adventure
A warm and cautionary story of an elderly woman who trusted Doubao for real-world errands, revealing the everyday risks of AI hallucination.
2026-06-22
Can Your Phone Become a Primary Development Tool? A Real-World Test
A hands-on experiment using Claude Code's Remote Control feature to write code, run tests, and deploy a real project entirely from a mobile phone.
2026-06-20
When AI Writes 99% of the Code, What Value Do Developers Have Left?
As AI coding tools like Cursor's Agent mode reshape software development, a tech leader reflects on which developer skills truly survive — and which disappear.
2026-06-14