I think, therefore I am; recording insights from daily life and technology.
The Pareto frontier defines the boundary of optimal trade-offs where improving one objective requires sacrificing another; applied to large model design, it reveals why no solution can simultaneously improve capability, cost, and latency beyond this frontier.
The article contrasts research (money to knowledge, long-term, idealistic) with engineering (knowledge to money, pragmatic, short-term), argues that software engineering is a practice-driven discipline where hands-on experimentation outweighs theoretical debate, and warns against replacing real-world practice with abstract models.
This technical briefing curates ten deep-dive articles covering Linux kernel memory leaks, Transparent Huge Pages latency pitfalls, linked list performance on modern hardware, Android AutoFDO optimization, memory consistency debugging, PageCache internals, DMA-BUF zero-copy architecture, and mutex/futex priority inheritance traps.
The article presents two perspectives on software architecture: the view perspective, where the system is the integral of all historical design decisions, and the design decision perspective, where each decision acts as the derivative defining the architecture's evolution direction and speed.
The article revisits Dijkstra's 1977 critique of natural language programming, arguing that formal symbols are essential for precise computation, and questions whether modern AI code agents can truly bridge the gap between ambiguous natural language and rigorous formal systems without reintroducing the same problems.
The article argues software engineering is an experimental discipline, critiques hindsight bias, and highlights OpenAI's harness engineering approach emphasizing executable architecture constraints, agent-readable code, observability integration, and human creativity as essential for reliable AI-assisted development.
Jeremy Howard argues that AI tools only interpolate existing code patterns and cannot perform true software engineering, which requires understanding component interactions and system design; overreliance on AI erodes developer skills and creates unmaintainable technical debt, especially harming mid-career engineers.
This article summarizes key insights from a design book covering design definition (Idea, Implementation, Interaction), conceptual integrity, design process models (Rational, Co-Evolution, Bazaar, Spiral), collaboration challenges, cognitive aspects (rationalism vs empiricism, user models, constraints, elegance), and practices for cultivating great designers through critique, varied experience, and protecting focus.
