Summary. Permissions in Claude Code aren’t a simple on/off. Inside permissions in settings.json, you split rules into three lanes — allow, ask, and deny — and use a matcher syntax that writes a scope inside parentheses per tool, like Bash(…), Read(…), and Edit(…). Master this syntax and you can do precise control such as “auto-allow … Read more
In short: Starting with Claude Code v2.1.186 (June 22, 2026), Claude automatically reacts to the output of any Bash command you run with !. Before, that output only landed in the conversation context and stopped there; now Claude reads the result and answers right away. If you don’t want that, add respondToBashCommands set to false … Read more
More and more AI chatbots have started, at some point, to show prompts like “Are you 18 or older?” or “Please verify your identity.” To lead with the conclusion: this is not the problem of one particular service but a signal that the AI industry as a whole is moving in the same direction — … Read more
In June 2026, something rarely seen unfolded in the AI industry. Anthropic’s top-tier model, ‘Claude Fable 5,’ had its access blocked worldwide just three days after release. This was not a model-performance problem but the result of a US government export-control order. What Happened On June 9, Anthropic released Fable 5 and Mythos 5. They … Read more
Answer Capsule Artifacts is a feature Anthropic added to Claude Code in June 2026. Per the official announcement, it is aimed at Claude Team and Enterprise plan users and turns session work into a live, interactive, shareable HTML page delivered by a single URL. Because it can connect multiple data sources and live code, it … Read more
Answer Capsule The real change in AI coding in 2026 is not one smarter model but a shift in the center of gravity toward ‘orchestration’ — weaving several AI agents together to divide up the work. Dynamic workflows like whole-codebase bug hunts and security audits are becoming standard, and for solo businesses and solo developers … Read more
The 30-second answer: MCP (Model Context Protocol) is an open standard published by Anthropic that unifies the way AI agents connect to external tools and data. It bundles the integration work you used to wire up separately for each tool into a single spec, so a connection you build once can be reused across many … Read more
The 30-second answer: An AI coding agent takes a different path every time, even given the same input. There is randomness on every call, and small differences get amplified as several tools are chained together. So you have to judge it by statistics gathered over many runs, not by a single execution — and that … Read more
Hooks are lifecycle shell commands you define in .claude/settings.json. Unlike CLAUDE.md instructions, which the LLM sometimes skips, they enforce a specific action deterministically. PreToolUse handles validation and blocking before execution; PostToolUse handles formatting and testing after execution. (Per the official docs.) I’m a solo developer who has run an automation pipeline called NABERAL on top … Read more
The 30-second answer: The three layers that extend Claude Code play different roles. Skills are a lightweight extension that pulls in task-appropriate instructions on demand; subagents are a heavier helper that runs isolated and in parallel in a separate context; hooks are a deterministic safety net enforced by a shell script rather than the LLM. … Read more