Engineering Methods on 乘物游心录 · Technical Notes

From 'It Runs' to 'It Can Be Maintained': Three Foundations of Reliable Systems

Wed, 15 Jul 2026 09:30:00 +0800

A system being able to start only proves that it passed the earliest gate. Once it reaches production, teams care about three questions: Can we locate a problem quickly? Will a local fault spread? Can every release be controlled?

Reliability is not a one-time acceptance check before launch. It is an engineering capability that keeps shortening the time from discovery to containment to recovery. These three foundations connect in order: make the system explain what happened, keep a fault inside its boundary, and give every change a safe way out.

Connect an AI Assistant to the Development Workflow: Optimize the Feedback Loop First

Fri, 10 Jul 2026 14:00:00 +0800

When people discuss AI-assisted programming, attention often goes to how much code a model can write in one pass. In a real project, a more useful measure is how long it takes to move from a proposed change to trustworthy verification.

Split work into verifiable units
#

A good task usually satisfies three conditions at the same time:

Technical Writing Is Not Filing Information; It Is Building Searchable Judgment

Fri, 03 Jul 2026 20:10:00 +0800

Saving links preserves information, but rarely preserves judgment. A few months later, what matters is often not what was read, but why a solution was chosen and under which conditions it remains valid.

Four kinds of information an article should keep
#

Context
#

Explain which system, scale, and constraints the problem came from. Without context, an experience is easily reused in the wrong setting.