Agents Are Making Top-Down Software Development Real
Agents make a new top-down companion workflow possible: start with a detailed plan, turn it into rough working software, then shape the result by judgment, pruning, and proof.
Practical guides for Claude Code — CI/CD automation, SEO workflows, skills, plugins, hooks, and building real products with AI-assisted development.
Agents make a new top-down companion workflow possible: start with a detailed plan, turn it into rough working software, then shape the result by judgment, pruning, and proof.
A real iteration log: designing a macOS app icon with Claude Code and the logo-designer-skill. 20 SVG iterations, 6 parallel subagents, and the surprisingly hard problem of curving a sound wave around a cylinder.

A short tour of five macOS menu bar apps I built to scratch my own itch, and why AI agents, XcodeBuildMCP, XC testing, and remote Mac runners make this kind of tiny native tooling much easier to ship.
A practical Claude Code skills hub for reusable workflows: logo design, interviews, Socratic reviews, AskUserQuestion, SEO research, and repo setup.
AskUserQuestion is the Claude Code tool that turns vague requests into structured decisions. Use it inside skills to force clarification before an agent writes code.

Dogfooding notes from a week of Mac menu bar apps, iOS workflows, BugDrop, issuectl, and small agent-powered developer tools.
A step-by-step guide to using Claude or ChatGPT for AI tax filing. Collect your documents, redact sensitive info, clean up the noise, and let AI help you fill out every form — without uploading your SSN to the cloud.
Stop re-authenticating every Playwright MCP session. This Claude Code plugin saves storageState authentication profiles per user role — log in once, keep your session persistent across sessions.
It's 2026 and big checks are still being cut to pre-product AI companies. Here's my standard reply to the recruiters.
The brute-force framing misses the point. For well-defined closure tasks — plan refinement, prototype validation, implementation verification — the Ralph Loop doesn't stumble toward correctness. It converges.
Nine layers of guardrails that make AI coding agents safe, predictable, and actually useful — from CLAUDE.md to CI pipelines to specialized review agents.
The old question was 'can you code?' The new question is 'can you build and ship?' A framework for evaluating engineers who use Claude Code, Cursor, and other AI-assisted coding tools.
Test Driven Development is dead. Quality control can be largely automated now. That frees you to focus on what actually matters: building things people want.
A reusable Claude Code skill for logo design. Point it at your repo, generate SVG logo concepts, refine the strongest direction, and export production-ready icons.

Anthropic just raised $30B. A small fraction could fund the Claude Code showcase the ecosystem actually needs.

The real barriers to AI coding tool adoption aren't technical — they're management problems, employee problems, and human nature problems.

Claude's default mode is eager compliance—give it vague instructions and it'll happily make assumptions. These workflow skills force systematic iOS HIG verification, catching platform convention violations you'd never think to check.

Claude's default mode is eager compliance. These interview skills flip the script—using AskUserQuestion to force systematic questioning before any code gets written.
Four Claude Code skills that generate and execute user workflow tests using Claude-in-Chrome and iOS Simulator MCPs. Catch bugs and edge cases while you're still building.
Use Claude Code to query your GSC data directly from the terminal. Find ranking opportunities, see what's working, skip the manual exports.

CLI tools for posting to Twitter, LinkedIn, Reddit, and more—designed to work with Claude Code so you can capture and share ideas without context switching.
Use Claude Code and gh CLI to choose optimal GitHub repo topics with real volume data. Data-driven tag selection in 30 seconds.
Use Claude Code with Google autocomplete for fast keyword research at an industrial scale - for blog SEO, product naming, or just understanding what people actually search for.

Claude Code uses gh CLI commands to debug and fix failing GitHub Actions workflows

You don't need custom slash commands or skills—just point Claude Code at your CLI tools and describe what you want

How I stopped babysitting GitHub Actions and let Claude Code monitor, debug, and fix CI failures autonomously

How AI pair programming helped me finally ship a stalled project
Automatically generate visual storyboards for demo videos using Claude Code and Playwright.
Turn any image sequence into an interactive treasure map using Claude Code
Generate, evolve, and iterate on SVG logos using natural language and Claude Code.
Why JSON task management wastes context and corrupts data in Claude Code sessions, and how todoq's CLI approach solves both problems with targeted commands that only load individual tasks.
How I use TFQ (Test Failure Queue) to avoid hitting context limits when debugging test failures with Claude Code. A simple CLI tool that queues failing tests and fixes them one at a time in separate headless sessions.