I think, therefore I am; recording insights from daily life and technology.
This article summarizes key ideas from the CPP2025 conference, including Bjarne Stroustrup's principles for C++ evolution, Pete Muldoon's pyramid of software engineering mastery, David Sankel's strategies for mitigating AI-generated code risks, and Michael Wong's vision for standardizing the C++ stack for AI agents.
The article analyzes four stages of Honor's MagicOS development, showing how each grand narrative — from core technologies to AI-driven paradigms — failed due to leadership's poor software understanding, talent gaps, and misjudged priorities, ultimately forcing a reactive, feature-copying approach.
The article distinguishes general software engineering practices — observability, security, reliability — from AI-specific challenges like LLM uncertainty, safety guardrails, agent frameworks, memory engineering, and MLOps, referencing Ant Group's open-source LLM landscape.
The author reflects on Manus's acquisition by Meta, arguing that the real value in AI lies not in training foundation models but in engineering the last mile—memory systems, tool sandboxes, and agent frameworks—that turn non-deterministic LLMs into reliable products, while analyzing five business models for AI companies.
