This week: leaning on native HTML over JavaScript, a sobering look at near-autonomous coding from Karpathy, scaling LLMs across real codebases, a thoughtful case for conscious LLM skepticism, and a blunt take on agentic engineering without prompt theatrics.
This week: a grounded Codex vs Claude Code comparison, Claude Code’s move into native LSP territory, the case for JSDoc as de facto TypeScript, why AI still can’t replace junior devs, and a pushback against AI-by-default software.
This week: LLMs at Oxide, building toast components, FizzBuzz in CSS, Opus 4.5 overview, and a holiday CLI challenge.