AI

Self-improving Claude Code skills reflect on sessions, extract your corrections and update their skill file with confidence levels, so mistakes stop repeating.

The best Claude Code sessions start with questions, not code. Spec-driven development forces requirements discovery upfront - interview first, spec second, code last.

Claude Code can control Chrome through your logged-in sessions, with no API keys. What it can do across tabs, how approvals work and the prompt injection risks.

Continual learning in Claude Code uses skills as persistent memory. Capture successes and failures in session retrospectives and share them across teams.

The Ralph Loop uses Claude Code stop hooks to keep an agent working autonomously for hours. How stop hooks, completion promises and max iterations fit together.

Rich Sutton's Bitter Lesson says general methods that leverage computation win by a large margin. What that means for how we build software and what we build.

Every AI-generated site looks the same. Magic Patterns bets on design instead: Chrome extension capture, an infinite canvas for variations, and Figma export.

Zed is not another Electron-based editor. It's built from the ground up in Rust, which means real performance without the memory bloat that plagues other IDEs.

Claude Opus 4.5 is the Anthropic model built for coding agents and computer use. Its pricing, benchmarks, token efficiency, effort parameter and agent features.

Coding changed more in the past two years than in the previous decade. We moved from manual typing to autocomplete, then to multi-file edits.

GitKraken with Claude Code shows agent-created branches in a visual Git graph for instant checkout, then uses AI to write and compose clean, stacked commits.

Cursor 2.0 ships Composer, the first in-house Cursor coding model, and a multi-agent interface. How Composer was trained, why it is fast and how it compares.

On October 29th, both Cursor and Windsurf dropped their first in-house models on the same day. Composer vs SWE-1.5. Here's what the benchmarks actually show.

Claude Skills load instructions, scripts and resources progressively so context stays small. A technical deep dive into SKILL.md, the API and Claude Code use.

NVIDIA Nemotron Nano 2 VL is a 12B open vision-language model with a hybrid transformer-mamba design. What it does for documents and video, and its limits.

Kimi K2 from Moonshot AI replaced Claude Sonnet 4 in my stack - an open-source 1T-parameter MoE model that is faster, cheaper, and strong at frontend code.

Emergent Labs builds full-stack apps from a conversation, then checks them with an autonomous testing agent. How the workflow, pricing and GitHub sync work.

Build a full stack AI SaaS with Next.js, Clerk billing, Convex and 11 Labs voice, then let AI coding tools speed up the build once the foundation is set.

OpenAI is turning ChatGPT into a hub. The new Apps feature lets you access external services directly inside conversations.

Claude Sonnet 4.5 in Claude Code uses parallel subagents to build a Next.js site and ten arcade games from three prompts. What worked and what still broke.