Section

7 pages

Posts

Posts about development and architecture

AI Agent Linter Ecosystem: Quality, Cost and Safety Control

Five open-source tools working together: archlint checks architecture, promptlint routes to the right model, costlint tracks spending, seclint filters content. All Go, no LLM, under 10ms per request.

Three blockers that prevent technical leaders from growing

Leadership growth is not about technology - it's about thinking. Three patterns that keep strong professionals stuck: territory defense, craving recognition, and the experience trap. With real examples.

Featured image of post Three blockers that prevent technical leaders from growing

Фаулер про LLM и разработчиков: cognitive debt, supervisory programming и кто на самом деле под ударом

Мартин Фаулер собрал senior-разработчиков и обсудил будущее профессии в эпоху LLM. Разбираю ключевые тезисы: cognitive debt опаснее technical debt, mid-level разработчики под большей угрозой чем джуны, DevEx и Agent Experience - это одно и то же, а supervisory programming несет с собой менеджерское выгорание.

Featured image of post Фаулер про LLM и разработчиков: cognitive debt, supervisory programming и кто на самом деле под ударом

Архитектура памяти для AI-агентов: как я научил Claude Code помнить

AI-агент без памяти - как архитектор с амнезией: каждый раз начинает с нуля. Я построил систему памяти для Claude Code - с индексами, контекстами, кросс-ссылками и жизненным циклом знаний. В статье показываю архитектуру решения и провожу параллель с тем, как устроена человеческая память.

Featured image of post Архитектура памяти для AI-агентов: как я научил Claude Code помнить

Spec-Driven Development: Controlling AI Code Generation

10 specifications, 20 minutes, 85% reproducibility. How formalizing requirements turns AI code generation into a controlled process.

Featured image of post Spec-Driven Development: Controlling AI Code Generation

Structural and Behavioral Architecture: Graph-Based Approach to Complexity Control

AI agents generate code quickly but often create architectural chaos. After two weeks of vibe-coding, the project turned into an unmaintainable mess. It became clear: a formal architecture model is needed. This article shows how to automatically build two types of architectural graphs: structural (from source code via AST) and behavioral (from acceptance test traces). In future articles, I'll cover architecture validation based on these graphs.

Featured image of post Structural and Behavioral Architecture: Graph-Based Approach to Complexity Control

Working with Chaos in Architecture

Most of the time you work not with a clean slate, but with history, compromises, and unfinished ideas. At first, it seems messy. But over time, you understand something important: clarity isn't given — it's created. Sharing experience working with chaotic architectures and how to learn to find patterns where at first glance there's only chaos.

Featured image of post Working with Chaos in Architecture